Free ChatGPT is Dead. The $200 Subscription PANIC

You probably think ChatGPT is a magical free assistant getting smarter every single day to benefit humanity. It's not. It's actually getting worse. OpenAI is burning through billions of dollars. And their brilliant survival plan involves downgrading free users, locking the real model behind a $200 paywall, and turning your conversations into ad space. So, if it's felt off lately, it's probably not in your head. They hope most people wouldn't notice the shift, and that's the point. But that free AI assistant, it's already dead.

Chapter 1, the $5 billion dumpster fire. OpenAI is in trouble. Documents show the company missed nearly every internal target it set for itself this year. Revenue goals missed. Weekly active user benchmarks missed. Even the IPO timeline, once framed as a victory lap, now looks more like a high-pressure sprint with no clear finish line. According to reporting from the Wall Street Journal in April 2026, the numbers paint a far less polished picture than the public narrative suggests. A $5 billion loss against $3.7 billion in revenue for 2024. The math doesn't lie. They earned $3.7 billion and they lost $5 billion. For every dollar the company brings in, it spends $2.35. The Wall Street Journal also indicates CFO Sarah Frier has flagged serious concerns about funding the next wave of multi-billion dollar data center commitments, especially if growth even only slightly wobbles. But according to OpenAI, it's all part of the scaling toward AGI. And despite the red flags, Wall Street is still largely on board. $5 billion in losses works out to roughly $13.7 million per day spread out across an entire year. That's about $9,500 every minute. Blink and roughly $160 bucks is gone. Executives aren't volunteering their stock options. Venture capitalists who've already poured in tens of billions aren't eager for another top-up round. Microsoft has even started [music] asking pointed questions on earnings calls, which leaves one increasingly exposed group. You.

Chapter 2. Labbotomizing the free tier. ChatGPT was never free. It was subsidized heavily and deliberately by one of the biggest funding machines in tech. The backers behind it read like a who's who of late-stage venture capital: Microsoft, SoftBank, Thrive Capital, plus Saudi sovereign funds. Combined investment reportedly exceeds $60 billion. And at the current burn rates, that doesn't last decades, it lasts years. Eventually, someone has to decide what refilling the tank actually costs. The panic is real. The IPO clock is ticking, and the survival plan that they've cooked up is desperate. It was to make ChatGPT seem the same while quietly changing what powers it. Essentially, it became dumber. There's a mechanism that users don't really see called capacity-based [music] fallback. During peak hours or after invisible daily quotas, users [music] get silently rerouted from the flagship model to cheaper mini variants. The handoff is invisible. The quality drop is not. After roughly 10 messages every 5 hours on the strong model, your conversation collapses [music] mid-sentence into generic autocomplete mode. User telemetry, internal strategy notes, and a parade of Reddit threads all confirm the same thing. Routing queries to smaller models is a deliberate cost control lever. Your AI assistant has been intentionally swapped for a budget version while you stare at a familiar brand name and assume nothing has changed. [music] The savings are huge. Routing queries to smaller models can cost a fraction of a cent versus several cents on a flagship. Multiply that across 800 million weekly active users and the math starts to [music] make sense for you. It boils down to paying full attention to a conversation partner who suddenly forgets your last three messages and starts answering in fortune cookie aphorisms. Performance gaps are not subtle. Independent testing suggests fallback models struggle with tasks the flagship handles in a single step. Multi-step reasoning breaks down. Code generation starts introducing subtle, hard-to-spot bugs. Long context summarization begins dropping key details. One day your assistant solves a calculus problem and the next day, after lunchtime traffic kicks in, it can't count the syllables in a haiku. [music] So this doesn't look random. It looks designed. A free tier that felt like a miracle was in reality a loss leader and [music] it was propped up by marketing capital and long-term bets that were never meant to last forever. The next shift is already priced in and it isn't subtle.

Chapter 3, the elite paywall. Internal pricing memos confirm OpenAI's $20 per month Plus tier, once the crown jewel of consumer AI, has been quietly demoted to neglected middle child. Plus subscribers still get something usable, but the Frontier Intelligence is behind a velvet rope. Meet ChatGPT Pro, $200 a month. It has unlimited access to GPT-4 and GPT-4 Turbo mode, higher rate limits than Plus, and priority compute when servers are under load. There are also lower error rates on harder reasoning tasks compared to standard tiers. The takeaway is simple: the smarter model now sits behind a paywall. The math gets uncomfortable the longer you stare at it. $200 a month is $2,400 a year per user. A family of four wanting access to the same smart tier? That's $9,600 annually. That is more than many households spend on groceries in several months. Reported uptake of the $200 tier sits in the lower single digits of total users, roughly 5 million subscribers worldwide. That's about 0.06%. And meanwhile, frustration from Plus users is building. The strategy is becoming harder to miss and it's about to reach into places you probably never expected. [music]

Chapter 4, your chats, brought to you by Charles Jr. The next version of ChatGPT isn't just smarter or more expensive. It's monetized in a new way. [music] Ads are coming to your AI assistant, and that is not rumor or speculation. It's an OpenAI's own announcement. Advertising is now an official product line, starting with the free tier and a new $8 ChatGPT Go plan rolling out in US beta. Advertisers reportedly face minimum commitments north of $200,000 just to get in the door. Early pricing was high, around $60 per thousand views, before being lowered to actually attract buyers. Ads will appear at the bottom of responses and be clearly labeled, for now, because the same company once told a podcast audience that ads in ChatGPT would be a last resort. But apparently, that last resort has arrived early. That last resort got a launch plan. The mechanics here are simple and deliberate. Free users see ads. $8 users see fewer or differently formatted [music] ads. $20 users are spared, for now. And $200 users, no ads at all. Same product, different rules. The more you pay, the cleaner and more capable it gets. The real play isn't just ads, it's targeting. Type "wedding" and you'll be swamped by venues, rings, and honeymoon packages for [music] the next 6 months. Because unlike social media, people tell this thing everything. Industry analysts estimate conversational targeting can massively outperform traditional ads, sometimes by multiples, for one simple reason. [music] Nobody tells Instagram their problems, but they tell ChatGPT. Reported figures suggest the ad business is being modeled to generate over $25 billion in annual revenue by 2029. That is a number that would on its own exceed OpenAI's entire current revenue. They have access to your medical questions, relationship doubts, [music] tax confusion, and 3:00 a.m. searches. That's a marketer's dream wrapped up in a friendly chat bubble. The reason any of this is actually happening comes down to one rival quietly devouring the only segment of the market that actually prints money.

Chapter 5, the Anthropic panic. Industry reporting shows something most people won't see in a headline. Rival firm Anthropic is now generating roughly the same level of annual revenue as OpenAI. Some estimates even put it ahead. The smaller company, the one Sam Altman's former colleagues founded, now generates more annualized revenue than the firm everyone calls the leader. What used to feel like a one-horse race is not anymore. The difference between these two firms is sharp, and it explains the panic. Anthropic runs more like a luxury boutique than a mass-market platform, serving just over 1,000 enterprise clients. Many of them pay in the seven figures annually, some in the eight. Its flagship model, Claude, has become a major force in enterprise coding, and it's used by Fortune 500 engineering teams, Wall Street, and biotech research groups. These aren't customers sensitive to pricing. At $50 a seat, they don't [music] blink. And that's because the productivity gains often pay for themselves before lunch on day one. OpenAI, meanwhile, is babysitting hundreds of millions of free-tier users. Anthropic extracts more from a single mid-sized enterprise contract than OpenAI does from a million casual ChatGPT users combined. Anthropic's niche market creates a very different kind of business. There are higher prices, longer contracts, and lower churn rates. The usage patterns are stable enough to actually forecast. Finance [music] teams can model it. Compare that to OpenAI's consumer scale. It's a constant surge of unpredictable demand from hundreds of millions of users. A system where a small price change could trigger mass cancellations overnight. They are two very different machines. One is optimized for predictability. The other is built for scale [music] and volatility. And in this market, it's not about branding. It's about what gets used. This is why the consumer product starts to look deliberately carved up. The free tier gets restricted to cut compute costs. The middle tier stops being the focus and becomes a funnel into the higher plans. The $200 Pro tier is aimed at users who treat it like business infrastructure, not a subscription. And the ads are built for everyone who will never pay at that level. And every decision traces to closing the gap on Anthropic, hopefully before the investors [music] start to ask uncomfortable questions about the story. Because if that gap doesn't close, the numbers won't either. And that pressure is starting to show.

Chapter 6, the [music] safety team bails. Key members of the safety team, people tasked with preventing the system from behaving unpredictably or dangerously, have been leaving, not quietly, and not slowly, at a pace that should concern anyone treating this like a stable long-term bet. Jan Leike's resignation post is short and very public. His exact phrasing reads like a warning: "safety culture and processes have taken a backseat to shiny products." Like he was the co-lead of an alignment team, a man whose entire job was preventing the technology he was building from causing catastrophic harm. And he was telling everyone that the people who write his paychecks have stopped caring about the part where humanity does not get hurt. But Ilya Sutskever's exit was the warning shot. Sutskever was OpenAI's chief scientist and one of its original founders. He left, founded his own venture aimed squarely at safe superintelligence, and took a chunk of the technical brain trust with him. Leike's departure, which followed shortly after, confirmed the warning shot had not changed a thing. The superalignment team, created to ensure future, more powerful AI systems remained under control, has been effectively dissolved. It was originally promised a significant share of internal compute resources, around 20%, to focus on safety research. That didn't last. Members have since left. The computing power has reportedly been redirected toward product development. And [music] what was framed as a core mission now looks from the outside like something that's been sidelined. There's a consistent pattern across this exodus. Engineers who joined to build safe, aligned AGI now describe a very different reality: a company whose product decisions are driven by commercial teams, where safety reviews are sped up to meet launch deadlines, and where internal concerns about whistleblower protections reportedly surfaced during the same restructuring period that positioned OpenAI for a potential IPO. The guardrails are gone. The promises around transparency and caution start to feel harder to pin on any one group. What used to be handled by dedicated oversight is now handled by process under pressure. Former employees estimate that the number of people [music] focused on alignment work has dropped sharply over the past 18 months. And the public-facing side of this company, it still moves fast. But some of the most impressive demos, the ones that made the headlines, are now more tightly controlled, delayed, or quietly pushed aside in favor of commercial priorities. And nothing was safe.

Chapter 7, the Sora bait and switch. Sora went viral in 2024 for its hyperreal clips of things like woolly mammoths walking through snow and eerily realistic city scenes, but even it had a shelf life. In April 2026, it was no longer available as a standalone consumer product. What once looked like a public glimpse of the future has been pulled out of public hands. Internal communications cite excessive compute costs and a strategic pivot toward enterprise tools, agents, and robotics. It all comes down to economics. A single Sora video generation reportedly cost over 1,000 times more in compute than a typical text query. Multiply that by millions of curious users and it becomes a problem. Hollywood studios, ad agencies, and enterprise media partners still get it at pricing that makes the economics work. But if you are just someone who saw a demo and thought you might like to make a birthday video for your niece, you don't. Industry reporting suggests studio licensing deals are in the $5 million to $20 million annual range, depending on the partner. The consumer-facing version that once drove the hype now exists mostly as clips, demos, and memories. [music] And it's not just Sora. Voice features that once felt unlimited are now throttled. Custom GPTs come with tighter limits, and advanced [music] reasoning that was shown freely during hype cycles is increasingly reserved for higher price tiers. Over time, the pattern becomes hard to miss: monetization not as a side effect, but as the structure itself. It all leads toward one scheduled outcome, and that outcome already has a timeline attached.

Chapter 8, the IPO extraction machine. Filings reveal OpenAI is restructuring its governance setup as it moves toward a potential public offering on a scale financial press called unprecedented. The original structure, designed to limit investor profit and protect the mission, is being dismantled. The nonprofit board that once had real control over the company is being pushed to a more advisory role. On paper, the mission hasn't changed. It's still about safe AGI for humanity. But in reality, the system holding that mission in place is being rebuilt from the ground up. The trillion-dollar valuation target, repeated in every leaked memo, is the only number that matters. The data doesn't just disappear. Every prompt you've typed, every conversation you've had since 2022, it all feeds the system. [music] And as this moves toward a public market story, that activity gets reframed. In an IPO document, it shows up as engagement or momentum. Strip away that language and it begins to look familiar. It's the same late-stage shift you see across tech platforms. The user stops being just the customer and starts becoming part of the product. Not someone being served, something being used. And in the end, the uncomfortable realization: you were never the customer.

Chapter 9. Welcome to the AI underclass. So where does that leave you? The golden age of free, powerful, broadly available AI is over. It lasted about 24 months, and it's not coming back. Your daily digital workflow is about to get more expensive and more constrained. What used to feel like free access to intelligence is turning into a tiered system where better capability sits behind higher prices. And the price for anything that actually works keeps going up. This is the pattern every tech wave hinted at but never made visible. Most people get the base version, and the base version doesn't get better first. It [music] gets cheaper, simpler, and more limited. Access stops being equal. It all becomes allocated. Different people and different levels depending on what they can pay. A version of technology that was supposed to expand access is quietly splitting into tiers. Some people get full capability. Most people get enough to stay inside of the system, and over time, even that shifts. The promise was a tool that would lift everyone. Now the future of human intelligence is behind a subscription service, and we are being priced out.

But inside the companies building this future, something else is happening. The people closest to the system are starting to leave, and it is sending serious shockwaves through Silicon Valley. For the last 2 years, you've been told that your salary is the biggest liability on your company's balance sheet. You've been told that AI is coming for your job. A digital worker is infinitely cheaper than a human one. But inside the boardrooms of Microsoft and Uber, something [music] just changed. Behind the scenes, Microsoft is pulling the plug on its own internal AI tools, and Uber just burned through its entire 2026 AI budget in just 4 months. The great AI replacement has hit a brick wall. For the first time in history, you are now the budget option.

In late 2025, Microsoft gave its engineers access to Claude Code, [music] one of the strongest AI coding tools on the market. Thousands of developers, managers, and designers were asked to hand over more of their work to Claude. As a result, AI coding became a core part of how the teams did their work. The results showed up right away. Use of the tool spread across the company faster than most people expected. Then, in May 2026, the company began pulling those licenses back. The Verge reported that Microsoft canceled most of its direct Claude Code access, steering engineers to its own cheaper tool, GitHub [music] Copilot CLI, by June 30th, 2026. The cuts fell the hardest on the team behind Windows and Microsoft 365. Officially, it was housekeeping. The tool had simply become too popular. The more the engineers leaned on it, the bigger [music] the bill grew, and the bigger it grew, the faster Microsoft locked its own people out. The tool was simply too good, and far too costly. The reversal ended right at the close [music] of the fiscal year when companies go hunting for costs to cut. Cutting the engineers off did nothing to Microsoft's bet on [music] the same technology. In November 2025, Microsoft had agreed to put up $5 billion into Anthropic, the maker of Claude. In return, Anthropic agreed to spend $30 billion running on Microsoft's Azure cloud. The deal never changed. That wasn't a company losing faith in AI. It still believed in the product. It still wanted to reap the windfall. It just didn't want to carry the cost itself.

And Microsoft wasn't alone. The same thing was happening over at Uber. In April 2026, Uber's Chief Technology Officer, Pravin Nag, admitted the company had already spent its entire 2026 budget for AI coding tools. A full year of money gone in 4 months. The rollout had reached roughly 5,000 engineers in late 2025, and adoption climbed fast. In February 2026, about 32% of engineers were using Claude Code. By March 2026, it was 84%. Uber had encouraged it, even building leaderboards that ranked teams by AI use. Usage turned into a contest, and then the money ran out. A typical engineer ran up to $150 to $250 a month. The heaviest users hit $500 to $2,000 each. Nag himself torched $1,200 of tokens in a single 2-hour demo. As the leaderboards pushed every team to climb, the monthly bill climbed with him. Nobody slowed down to ask where the ceiling was. Uber's CEO says AI agents now write about 10% of its code. It's performing real work, but there's no real savings. It's like a family blowing a 25-year mortgage fund on groceries in a single weekend. That is the shape of Uber's 4-month collapse. Uber's Chief Operating Officer, Andrew McDonald, described the figures as "head exploding." For all the tokens his engineers burned, he couldn't point to anything customers could feel. We're told that AI saves money by replacing pricey humans. But the reality is much different. A higher-performing artificial intelligence agent can rack up massive hidden costs per finished task compared to the junior developer it replaces. These are the richest, most advanced firms on the planet, and they cannot keep their own AI switched on. An entire industry gaslit itself into believing this would be the future. Instead, they're holding the bill.

Through 2025 into 2026, the biggest names in tech turned AI use into a sport. At Amazon, staff were pushed to "token max," slang for burning through as many tokens as possible. And that is a problem. More tokens means a bigger bill at the end of the month. Amazon's message wasn't "use AI wisely." It was "use more." At Meta, an employee built a user board called Claudionics, ranking workers by how much they spent on AI. Heavy use became a badge of honor. The engineer firing off the most prompts didn't look wasteful. He looked like the future. Meta would later see the danger and axe the board. It created a tragedy of the commons. Everyone was burning through the resources for self-gain. To any single engineer, one extra prompt felt like nothing because it nearly was. It was a fraction of a cent. So, nobody hesitated. Nobody asked if that prompt was needed. And everybody reached for the machine every time. That won them leaderboard points. No one stopped and looked at the bigger picture. The bigger [music] total. And the total was brutal. One harmless prompt. Thousands of workers. Millions of tasks each day. It all grew into a figure no one could defend.

But the chatbots were only the warm-up. A worker overusing a chat window is an expensive habit, not a threat to the company's survival. The real danger arrived the moment those same workers stopped chatting with the AI and [music] just turned it loose. That changed everything. A normal chat prompt is a single, neat exchange. You ask, the model [music] answers, the meter clicks once. An agent is a different beast. It doesn't answer a question, it chases a goal. Hand it a coding job and it works on its own, reading the code, drafting a plan, writing an edit, running the tests, and trying again when something breaks. The loop continues until it decides the job is done. Every step costs money, [music] so a single agent task can swallow far more computing power than a simple chat. A typical job runs 5 [music] to 30 times the tokens. In the worst cases, reports say it can spike past 1,000 times. [music] And the priciest part isn't the work that lands, it's the work that fails. The SweetBench is an industry test built from real software bugs pulled from code that human engineers once had to fix [music] by hand. The best agents currently solve less than 50% of the test. The agent finds the bug, proposes a fix, runs the tests, and fails. So, it rethinks, it tries a new path, and fails again. A difficult bug can mean 5 to 10 full attempts, [music] each costing about as much as the first, before the agent just gives up, or a human gets called in to fix it. Imagine a plumber that charges full rate every time he drops his wrench, bills you for each trip to his van, and still leaves the pipe leaking. You wouldn't call that a timesaver. You would call it a scam. And at the scale of a big company, that failed agent task is exactly that: full price for the dropped wrench with no promises of fixing the [music] leak.

Even if the price of a single token keeps dropping, and it will, the problems don't change because it was never just about one token. The chips that power AI burn electricity, a lot of it, and that's a cost nobody can just wish away. If a single chat prompt is like a desk lamp switching on, an AI agent is like lighting an entire stadium for one person. Brian Kenzaro, vice president of applied deep learning at Nvidia, the company whose chips power the AI boom. Nvidia has every single reason to insist the money isn't an issue. And yet, in an interview with Axios, Kenzaro said the cost of compute for his team is "far beyond the cost of the employees." The chips and the power to run the AI revolution now cost more than the people working on it. The machine is not the cheap part anymore. The people are. The industry knows it, which is why Microsoft is busy designing its own chips. Its newest one, the Maya 200, was built to drag the cost of running those models down. On Microsoft's April 2026 earnings call, CEO Satya Nadella said the Maya 200 produces "over 30% better tokens per dollar" than the older chips in Microsoft's rack. It's already running in data centers in Arizona and Iowa. It is genuine progress, but there is a catch. Every bit of savings the hardware makes gets eaten up by the AI agent workload stacked on top. You make each token cheaper, and the agents just demand more of them. The savings vanish long before they reach the bottom line.

Research firm Gartner predicted that by 2030, running a top model could cost about 90% less than it did in 2025. A huge drop in the price of raw thinking power. And still, Gartner says company AI won't actually get cheaper. The [music] reason is the same force squeezing Microsoft and Uber. Agents burn so many more tokens per task that rising use outweighs the falling price. AI companies have little reason to hand the full discount to customers anyway. One Gartner analyst framed it as a way of executives banking on "cheap automation." "Cheap everyday tokens," he said, "are not the same as cheap access to the most powerful AI." The cheap part and the powerful part are different products. The powerful part stays expensive. Human wages, meanwhile, have barely even moved. The cost of high-end AI thinking hasn't. So, the price of running the digital worker soars past the salary of the human it was built [music] to replace. At that point, things start to fall apart.

Goldman Sachs ran the numbers and found that Agentic AI could push global token [music] use up 24-fold by 2030. It would rise to roughly 120 quadrillion tokens every month. Big tech is racing to build for that demand, pouring fortunes into data centers and chips for a future of AI agents. But the very demand those companies are spending hundreds of billions to capture is the same demand destroying their profit margins. They're building, at huge cost, machines that lose money on every job. None of this is new. It just has reached full scale. Go back to 2023 when Microsoft was [music] selling GitHub Copilot for $10 a month. Wall Street Journal found that Microsoft was losing more than $20 per user every month. Its heaviest users cost as much as $80 a month each. $10 coming in, $80 going out. It was [music] the business equivalent of selling $20 bills for $10 and promising to make it back on volume. The more you sell, the more you lose. Those early Copilot losses were a warning, [music] one that hasn't been heeded. A single agent can burn through in one task what a chat user spent in a month. The Goldman Sachs [music] report described it as "too much spend, too little benefit." The man behind that report, Jim Colloo, has been the loudest skeptic in the room. He notes big tech is on the course to spend around $1 trillion building out AI. But what expensive problem does all of that money solve? Years in, and there's still no definitive answer. The spending is enormous. The payoff is marginal and hard to measure. The market isn't closing the gap between what AI costs and what it's worth. It is running in the opposite direction and picking up speed. An entire industry is spending trillions to build a future where its own flagship products are guaranteed to lose money. And every one of those failures traces back to one root cause no amount of cash can fix.

As AI systems become more capable, their costs don't rise in a straight line. They curve upward. Moving from chat to a true agent means adding layers of reasoning and retries, each compounding on the last. Humans behave differently. A worker who becomes more skilled typically gets faster and more effective at roughly the same cost. [music] The two aren't competing on the same trajectory. They are following different curves. A junior worker doubles their output and their pay barely moves. An agent doubles its output and its token bill can jump many times over. At the prices of 2026, the fully self-running agent is not a replacement for workers. It is a luxury. A [music] company can afford it in small, high-value corners where the numbers work, but they can't turn it loose across an entire workforce. Despite the hype, the dream of the self-running worker was never going to survive its own costs. It was never going to work. Humans aren't being spared out of kindness or even saved by a sudden corporate change of heart. They're protected by something colder and more reliable: simple arithmetic. The smarter the system gets, the more it costs. The great replacement has run straight into a reality no clever engineering will get around in this decade.

So what does all of that mean for your job? Through the whole AI boom, "human in the loop" sounded like a soft idea, a feel-good safety feature, a box to tick to keep regulators calm. Not anymore. It's become the only way that this works. Big tech will keep using AI, aiming it at these small, low-cost jobs where the cost justifies the means. But the dream of cheap AI replacing everyone is already over. It died inside of those canceled licenses and burned budgets long before it reached the rest of the workforce. In its place, a new job is becoming the most valuable seat in the building: the verification specialist. Someone who watches the agent and catches it the moment it slips into an expensive failure loop. The person who checks the [music] output and kills a runaway process before the cost bankrupts the project. They are the one thing standing between the company and a budget [music] that bleeds out cycle after cycle.

Even the industry's loudest cheerleaders describe this future without meaning to. Nvidia CEO Jensen Huang likes to say that one day "100 AI agents will work alongside every employee." Alongside every employee, not replacing them. The agents multiply, sure, but they multiply around one worker at the center. Are the machines creating value or just running up a bill? Take that worker away and nobody knows the answer until that bill arrives. And AI doubt is spreading. Language learning app Duolingo leaned hard into the "AI replaces everyone" story and then walked it back. The CEO admitted the tech won't take over the work like people do. The company even scrapped a rule that it had tied job reviews to how much AI staff used after workers pointed out that it rewarded busy work instead of results. So, the future of work is a blend, one that was always coming. Through the whole boom, we were told AI was the worker and the people were just a cost to trim. That's changed. Humans are the cheapest thinking engine on the earth. A flexible mind that runs on a sandwich and a night's sleep and never bills the company 1,000 times over just to stop [music] and think. The great replacement wasn't defeated by fear. It was defeated by a spreadsheet. And in those numbers lies a surprising triumph for ordinary workers.

It's not just Microsoft that sees the writing on the wall. The whole industry is starting to get nervous about what's coming next. A $3 trillion market cap, $32.9 billion in cloud revenue. But could Microsoft's empire collapse because of one startup? [music] Nearly half of Microsoft's future cloud empire depends on a single startup, one that is burning $12 billion every quarter. I'm Josh, and on today's episode of The Infographic Show, we'll reveal the massive Microsoft divorce that could bankrupt OpenAI and ChatGPT forever.

Microsoft doesn't just invest in startups. It captures them. They hand founders up to $150,000 in free Azure cloud [music] credits, not cash, digital vouchers. These small companies spend months building their products on Azure, mapping every database and workflow to Microsoft's proprietary formats. By the time the free credits run out, they're stuck. Tear out the backend and their apps crash. So, they [music] start paying real money. And now they're stuck in the architecture. They turn to corporate credit [music] cards, pay-as-you-go tiers. And just like that, Microsoft turns free credits into real cash flowing straight into its books. But it's not just small startups that get caught up in Microsoft's digital web. Microsoft plowed $13.8 billion in direct funding into OpenAI, but almost none of that money actually left Microsoft's coffers. [music] Instead, the company handed Sam Altman customized digital vouchers. OpenAI then used those vouchers to rent Microsoft servers. Every dollar spent legally counted as Azure revenue growth on Microsoft's books. OpenAI [music] was backed into a corner. What does this mean for Microsoft's balance sheet? Corporate accountants have a secret weapon: [music] a metric called the remaining performance obligation. It tracks guaranteed future revenue, and Microsoft is currently sitting at a staggering $625 billion. [music] Wall Street treats that number as cash in the bank. Analysts feed it into discounted cash flow models, using it to justify Microsoft's share price all the way to the end of the 2020s. 45% of Microsoft's guaranteed $625 billion is locked in fueling OpenAI's machines. The future of its cloud empire hinges on one startup, and Wall Street [music] expects it'll pay. Microsoft's balance sheet shows $40.3 billion in debt. Wall Street accepts that, [music] but off the books, there's something hidden: a $662 billion trap. Shadow leases and custom deals keep OpenAI servers running, [music] and Microsoft isn't alone. In the cloud world, physical hardware hides behind complex [music] lease structures, and that is the problem. OpenAI doesn't have the cash to cover this hidden debt. Financial analysts at [music] Deutsche Bank crunched the numbers. They projected OpenAI will burn through $143 billion before ever turning a real profit. [music] A company setting billions of dollars on fire every 12 months just handed its largest investor the biggest profit [music] spike in recent corporate history. OpenAI is an unemployed tenant facing eviction. Microsoft [music] is the landlord holding the keys. Microsoft prints fake IOUs and hands them to OpenAI. Those IOUs pay for renting the servers. Microsoft [music] legally reports that rent to Wall Street as cloud revenue growth. It is a flawless infinite money loop until the servers actually turn on.

Why can't OpenAI [music] just build their own infrastructure? Well, the math doesn't add up. Sam Altman saw the problem. Azure's credits could never fuel [music] the endless compute he needed. So, he engineered an escape route called Project Stargate. He pitched a $500 billion master plan. He wanted independent data centers, 10 gigawatts of dedicated power. He flirted with sovereign wealth funds and foreign telecom giants. He plans to bypass the Azure ecosystem entirely to sever the partnership. SoftBank and Oracle entered negotiations to provide alternative capital and infrastructure. Construction crews mobilized in Allen, Texas. OpenAI prepared to build a 1.2 [music] gigawatt facility. One site serving as the beachhead for a sprawling $665 billion infrastructure rollout through 2030. All they needed was the financing. The banks opened the disclosures, ran the numbers on the deficit, logged delays on permits, and tallied the engineer shortage to cool the massive racks. Wall Street refused the $500 billion gamble. Private investors [music] wouldn't touch it. OpenAI quietly scrapped their master plan. They slashed the projected independent [music] compute speed. They retreated to the existing infrastructure. They lack the capital to build their [music] own fortresses and the margins to keep renting Microsoft servers. Azure's credit can't sustain the burn. Microsoft is left fueling a captive entity that cannot repay the principal.

How does the physical hardware accelerate this crisis? Microsoft [music] spreads its massive server costs over a 6-year accounting window. This keeps their quarterly spending low on paper, but AI doesn't wait. Frontier training models make top-tier GPUs obsolete in just 36 months. Every chip you buy today is tomorrow's legacy hardware. Imagine a delivery company buying a brand new fleet of trucks. They have to replace that entire fleet every 18 months because the old trucks suddenly cannot deliver packages [music] fast enough. That's the economic reality of artificial intelligence hardware. Financial models from analysts expose $176 [music] billion in hidden GPU depreciation actively decaying across the tech sector. Microsoft is booking record profits today by ignoring the physical decay of its own hardware. When the actual replacement cycle hits the balance sheet, the capital expenditure bill will explode. The models burn cash at a high velocity. OpenAI generates $12 billion in quarterly losses. Forbes [music] estimates that the Sora video generation model alone consumes $15 million in hard cash every single day. Don't forget to like, share, and subscribe. The AI takeover isn't coming. It's already here, and we'll try [music] to keep revealing the true story.

Video generation isn't just text on steroids. Every single pixel must [music] be calculated and rendered in sequence. It demands an exponential jump in raw computational power. Microsoft's [music] capital expenditures surged 66% to $37.5 billion in a single quarter to feed this. The tech giant is purchasing land and pouring concrete to meet a theoretical demand that OpenAI literally can't afford to use. Taiwan Semiconductor Manufacturing Company (TSMC) operates as the physical brake on global artificial [music] intelligence. They produce 90% of advanced silicon on Earth. Corporate demand outpaces their physical factory capacity by a factor of three. You can't speed up the extreme ultraviolet lithography processes. [music] You can't skip the chemical etching. Each chip must be born inside hyper-specialized clean rooms with perfect vacuums and extreme atmospheric control. Tech giants are sitting on billions in cash, unable to spend it while their current servers lose value every single day. TSMC is racing to expand. They're planning a $52 to $56 billion expansion in 2026, but even that can't catch up. And it gets worse. A single gigawatt data center requires thousands of miles of thick copper wiring. Copper is running out. Mines in South America are struggling to meet demand. The cables, the power, the cooling, they [music] all have to be perfect. One slip, one missing component, and the whole operation stalls. It's a billion-dollar waiting game. [music] And that's not the only problem. Across the industry, data centers swallow 449 million gallons of water daily. Hyperscale facilities [music] drain up to 5 million gallons of potable water every 24 hours just to stop the server racks from literally melting. Standard air cooling maxes out entirely at modern rack densities. The facilities need to pipe in cold water directly to the silicon chips to maintain operational temperatures. Local governments are starting to panic. Municipal water supplies are dropping while server farms keep expanding. In Florida, regulators stepped in with strict new rules to stop residents' utility bills from spiking. The legislation targets the massive energy and water demands of new data center construction. The Midwest [music] faces water stress. Local city councils are passing emergency moratoriums on new facility permits. They are choosing drinking water for their citizens over artificial intelligence infrastructure. The hyperscalers are being locked out of prime real estate because the local aquifer cannot support the thermal load. The hyperscalers are scrambling. They're abandoning traditional air cooling and moving to direct-to-chip liquid systems. [music] That means ripping out entire air cooling units and bolting in metal cold plates straight onto the hottest silicon chips. Engineers need to thread miles of pressurized coolant pipes directly over racks holding billions of dollars of active hardware. The sheer material cost of this plumbing destroys the baseline construction budgets. If a single leak in a coolant line drips onto a motherboard, it destroys millions of dollars in silicon instantly. The capital required effectively doubles the initial build cost. Microsoft [music] is footing this bill entirely upfront.

What happens when the hardware reaches its absolute limit? Generative AI only works as a business if the margins are massive. Those fat profits are supposed to pay for the mountains of steel, silicon, water, and electricity working away behind the curtain. OpenAI historically charged premium prices for Application Programming Interface (API) access. They utilized their monopoly position to drain enterprise budgets. The API market is turning into a commodity battlefield. OpenAI launched the GPT 5.2 Frontier model and priced it at $1.75 per million input tokens. They guessed Fortune 500 companies would just absorb the cost to maintain access. They assumed the dominance would hold. They didn't expect what came next. Chinese competitors [music] arrived. They didn't try to outspend OpenAI. They didn't build trillion-dollar server empires from scratch. They used model distillation. Instead of training artificial intelligence from scratch, they bought API access to OpenAI's best model. They asked millions of advanced math and coding questions. They captured the answers and [music] then they trained smaller, leaner architectures on those outputs, all without paying for the training costs. Deepseek released their V3.2 architecture and matched the performance of the American Frontier models. They dropped their exact same token package to $0.28. Overnight, the premium collapsed. A global price war erupted. Gross margins evaporated. Developers are actively routing their daily tasks to cheaper alternatives. They use OpenAI strictly for the most complex reasoning tasks. [music] They funnel 90% of their standard workloads to foreign models or localized open-source alternatives. This strips away the high-margin volume OpenAI desperately needs to survive.

How does Microsoft respond [music] to this revenue collapse? Microsoft is watching on as this collapse unfolds. They know OpenAI can't generate enough revenue to pay off their hidden debt. [music] The answer is to extract the value themselves. They integrated OpenAI's tech into their own products. Copilot becomes a core part of Word, Excel, and Teams. [music] They charge a flat $30 a month per enterprise user. 15 million people already subscribe. Wall Street assumes every $30 is pure profit. Standard software is like a printing press. Build it once, distribute it forever, and profit margins soar toward 90%. Generative AI obliterates that [music] model. Every time someone clicks the Copilot button, a supercomputer fires up and devours energy. [music] That $30 flat monthly fee doesn't even cover the basic electrical cost. Microsoft eats the cost to keep the customer locked into the ecosystem. They're actively subsidizing the enterprise workflows of the [music] largest corporations on Earth. Year-over-year cloud growth slowed to 39% in the second quarter of 2025. [music] That's below the 40% growth Wall Street demands to justify the $3 trillion valuation. Profit margins are under pressure, sliding from a strong 46.7% [music] to unsustainable levels. Microsoft is utilizing the most expensive computational infrastructure in human history to draft basic corporate emails. If they raise the price of Copilot, clients will turn to cheaper alternatives. If they keep the price at $30, the power users consume the server capacity. If they restrict the Azure capacity for Copilot, the software experience degrades [music] instantly. Where does the breaking point occur?

Standard venture capital can't touch this [music] deficit. Silicon Valley doesn't have the liquidity to cover a hole this massive. They need a single entity capable of writing a $50 billion check in one afternoon. In January 2026, Sam Altman [music] flew to the United Arab Emirates. He pitched an $830 billion corporate valuation to sovereign wealth funds in Abu Dhabi and requested $50 billion in hard cash. It was a Hail Mary to keep the existing Azure servers running and bypass [music] the domestic banking system entirely. The Committee on Foreign Investment in the United States watches every move. The Pentagon classifies frontier AI as critical national security infrastructure. It's seen as a weapon system. Middle Eastern funds are blocked instantly. The government previously forced a Saudi fund to completely divest and exit [music] an Altman-backed artificial intelligence chip startup. OpenAI is being starved of domestic liquidity and barred from accepting foreign sovereign bailouts. Every lifeline has been cut. The fallout [music] in the bond market was swift. Microsoft officially carries $100 billion in debt. Investors recalculated the risk premium. Azure's growth slowed and 45% of future revenue is locked [music] into a single unprofitable tenant. They watched $122.7 billion vanish in shareholder [music] payouts while the data centers burned cash at record rates. Treasury yields climbed to [music] 4.08%. Suddenly, cheap capital disappeared. Every new facility [music] had to prove immediate profitability. Something OpenAI couldn't guarantee. And just like [music] that, expansion froze. The data halls began to implode. The sheer scale of the operational bleed has caused an internal panic. Compute access has been throttled to survive the cash crunch. ChatGPT responses lag for everyday users. The revenue curve has flatlined. They were cornered. So in late February, OpenAI did the unthinkable. They betrayed Microsoft. In a desperate bid to keep the lights on, Sam [music] Altman secured a $110 billion bailout led by Amazon, Nvidia, and SoftBank. But this isn't a victory for Microsoft. It's a hostage situation. To get Amazon's $50 billion, OpenAI had to agree to plow huge amounts of money into Amazon Web Services. They are cannibalizing Azure. The flawless infinite money loop is officially broken. Microsoft is left holding the bag on billions in decaying hardware while their unemployed [music] tenant packs up and moves across the street. Microsoft fed OpenAI billions in fake digital credits. [music] OpenAI returned the favor by walking away, leaving Microsoft with billions in real [music] physical liabilities. The accounting tricks were merely smoke and mirrors. OpenAI betraying Microsoft and pulling them down is just the first domino. Wall Street is looking at software, but the companies building the physical AI hardware are hiding a completely different, much deadlier financial secret. The cracks in that foundation [music] are already tearing open right here. The divorce might not be finalized, but they are spending time apart. Now, Microsoft is left thinking about [music] what could have been.

But OpenAI's troubles are far from over. Windows 11 runs on roughly 50 million lines of code. Print that out and it would stretch 131 miles end [music] to end. Sounds impressive, but this isn't some modern marvel. It is a 40-year-old disaster in the making. The constant glitches, the blue screens, and the 240 million PCs that became instant landfill waste. And if you look closely enough, you'll see Microsoft's biggest mistakes weren't accidents. They were calculated strategic sabotage.

Chapter 1, the 50 million line time bomb. If you're using Windows, there's a good chance you're actually running Windows NT under the hood. And when that first launched in Windows NT 3.1, it already contained 5.6 million lines of code. At the time, that was massive. But it wasn't built for consumers. NT was designed for workstations and enterprise servers, where stability mattered more than simplicity, and backward compatibility mattered more than elegance. But then something changed. It began powering just about everything. Over time, it became the foundation for everything Windows runs on today. And with each new generation, it

Windows has only become more complex. It's become harder to maintain and more prone to performance issues, slowdowns, and unexpected glitches with every major release. It reached a breaking point in 2021 with the launch of Windows 11, the most bloated and intricate version Microsoft had ever released. And this time, users couldn't ignore the elephant in the room because for the first time, their computer simply couldn't handle it.

Windows 11 now comes pre-installed on all new Microsoft devices, and it's offered as a free upgrade for existing PCs, but only if they meet the system requirements. That list is thinner than most people expect. You need a compatible 64-bit processor, at least 4 gigs of memory, and 64 gigs of storage. On paper, that sounds reasonable. In reality, it quietly excludes a massive number of older machines that are still perfectly functional. Anyone with an older workhorse computer has to upgrade or be left behind. And for millions of devices, that doesn't mean a slow transition. It means obsolescence. Millions of computers are destined for the landfill. It's turning into the biggest upheaval in Microsoft history.

The term backwards compatibility is familiar to gamers. And for a long time, it was a frustration point. When a new console launches, support for older systems doesn't last long. New games stop releasing for the older hardware, and eventually upgrading becomes the only way to keep playing. But in earlier generations, it went even further. New consoles often couldn't even run old games at all. This was often due to incompatible hardware, especially during the cartridge era. But there's also another reason backwards compatibility is hard to come by. It simply does not make the company money.

Windows 11 is seen by most users as a bloated mess of a user interface, and that's not surprising. At this point, Windows NT is built on a 40-year-old foundation of code that never quite stopped being built. What is left is unwieldy and prone to glitches, but it's paying dividends. When Windows 11 dropped, it's estimated that around 50 to 55% of existing devices were incompatible with it. Those users all had to buy new devices, a massive windfall from Microsoft. Was it a lack of foresight or by design? And if it was the latter, which other massive Microsoft screw-ups were by design?

Chapter 2. The sacrificial lamb, Windows MI.

In 2000, Microsoft was still the dominant force of home computing long before Apple's resurgence. Windows 98 was everywhere. It ran on most home PCs, defining an entire era of personal computing. But it was aging and Microsoft desperately needed a replacement. So, they introduced a new system, Windows MI, also known as Windows Millennium Edition. It would become one of the worst ways to start a new millennium.

Windows MI promised a new interface, quicker startup times, and built-in internet features for the emerging web era. The consumer versions could theoretically handle up to 1 and a half gigs of RAM, but most users had limited need for heavyduty computing. So, when it launched, the system gained applause for its enhanced security systems. And then reality set in. Windows MI was plagued with glitches. People reported that the security and maintenance system named PC Health wouldn't work properly. The computers that upgraded were experiencing frequent malfunctions and slowdowns. Computers wouldn't even shut down properly. With overheating already common on older desktops, people began to worry that they installed a Trojan horse onto their systems.

Microsoft would respond sooner than expected. As the negative word of mouth continued to grow for Windows MI, the company released some limited fixes. Given the scale of the issues, many critics went as far as calling it unusable, even a downgrade from Windows 98. So, it wasn't surprising when Microsoft announced a replacement less than a year later. Windows XP, released just 13 months after Windows MI. It introduced what now feels like basic features like a built-in zip folder support and far more stable foundation. Windows MI had been a disaster, but it was also one Microsoft was eager to move past and one the industry quickly forgot.

Windows XP was the first consumer-focused operating system built on the Windows NT kernel, a far more stable foundation than what came before it. Today, it's often seen as the start of modern Windows. On the surface, it looks like poor planning, releasing a major new operating system only to abandon its predecessor less than a year later. But it wasn't. It was deliberate. Leaked internal memos from 1999 mention a project code named Neptune. It was an early attempt to bring NT architecture to consumer machines, but the project was eventually scrapped and folded into something larger, something that became Windows XP. Meanwhile, Windows Mi still shipped to market, retailing for around $29 for the full version, or between 59 and 109 for qualifying upgrades. And it delivered what many users remember as instability, frustration, and constant technical issues. This was seen as one of Microsoft's big early screw-ups, but it wasn't. It was a disaster by design, and it did exactly what the company intended, a rushed stop gap to squeeze one last paycheck out of a dying technology. And it wouldn't be the only time Microsoft played its loyal customers like a fiddle.

Chapter 3. The $2,100 email machine.

The next major shift came in 2007. Windows Vista was Microsoft's first major successor to Windows XP, released 5 years after its hugely successful predecessor. And a lot had changed in that time. Home computers were faster and more capable. Customers weren't sure if their older computers were going to be capable of handling the new program, which contained much more code than the previous version. When it was announced, Microsoft claimed that nearly every PC on the market could at the time be able to run Windows Vista. They could, just not very well. A lot of PCs at the time ran on Intel processing chips, and their current graphic card chipset was seen as underpowered for running the slick new 3D interface that rolled out with Vista.

Seeing the concern, Microsoft promised they would classify PCs as Vista capable if they were usable with the new program. This led to many users upgrading their PCs after years of holding out just to run the new system. A large number of them chose the popular, relatively affordable Intel powered machines. It turned out to be a costly mistake. There's a difference between running a program and running it in a way that is usable. Those who tried to run Windows Vista on these new PCs, some of which could run up to $2,100 at the time, would find that many key functions were missing. The computers didn't handle the complex interface, and that left them to be able to do little more with their new PC than send emails, and Microsoft found itself besieged with complaints.

This eventually led to a class action lawsuit where many documents from the development period were unsealed. After all, if a lawnmower manufacturer slaps Ferrari capable on one of their products, it doesn't make it so. Again, that was another deliberate strategy. Internal emails showed that Microsoft very well knew that these Vista capable systems were not up to par. Corporate vice president of Windows product management Mike Nash saying, "I now have a $2,100 email machine." Former Microsoft co-president Jim Alin even raised concerns about misleading customers. Internal emails also revealed frustration from Huelet Packard, which found itself caught between Microsoft and a wave of unhappy customers. The situation escalated into legal action, including a class action lawsuit that was eventually descertified, but not before internal documents painted a clearer picture of what was happening behind the scenes. Microsoft had engineered the fiasco to ensure their partner Intel met its quarterly earnings goal. The customers were screwed over for corporate gamesmanship and it wouldn't be the last time.

Chapter 4, the Trojan horse, Windows 8.

The releases were coming faster now, and in 2012, customers were greeted with Windows 8. This would turn out to be one of the most controversial Windows releases yet, and not because people were finding themselves locked out of features. Rather, everyone was on the same page on this one, and that wasn't a good thing. This was the first Windows model to feature a new user interface, Metro. At the time, mobile devices were exploding in popularity, and Apple's iPad had already become the dominant way to browse the internet. Microsoft wanted to mimic that success, whether the users wanted it or not.

Metro used a simplified format that involved shortcut icons to apps, minimalist design, and an optimized desktop. This would mean the apps people clicked on most often were the first ones that they saw when they opened it. On mobile devices, that seemed to be a success, as app-based navigation was what most people used. But on the desktop, that was a different story. The full screen start screen worked well on tablets, but on desktop PCs, it quickly became a problem. It looked clean, but it made precise workflows harder. And the shift to full screen apps made multitasking feel clunky, even restrictive. For many users, especially in business environments, it did not feel like an upgrade. It felt like a step backwards.

When Windows 8 launched in 2012, that tension became impossible to ignore. The interface was designed around touchscreens, but most users were still on traditional desktops. It quickly became a nightmare. The full screen apps made multitasking frustrating, if not impossible, and the PCs were seen as a poor fit for business use. A failed attempt to copy the iPad or something worse. The Metro user interface, which would later be renamed modern, was designed for everyday users, but everyone else was pulled into it. whether they wanted it or not. UX designer Jacob Miller, who worked on Metro, later commented in a 2014 Reddit discussion, clarifying some of the design intent behind it. He explained that the interface was built with casual users in mind, who Microsoft believed represented the majority of its customer base. The hardcore users would just have to deal with it. It was a mercenary decision designed to boost Microsoft's long-term sales prospects. The professional users were the heaviest buyers, but they were relatively few in numbers. Content creators and tech experts were heavily outnumbered by grandmas and gamers, but the casual users didn't like when new features were too complex for them to use. So, Microsoft decided to move toward a simplified Windows interface that everyone would have to use, expanding its user base in the process. The assumption was that the users with the most invested in the platform would adapt over time rather than abandon it and start fresh with new systems. The users weren't customers, they were the product, and the worst was yet to come.

Chapter 5, the Great PC Purge.

Windows kept on rolling out new releases for the next few years, with all models being seen as minor updates. That was until Windows 10 dropped in 2015. This one was promised as the biggest upgrade to Windows in years, a unified version optimized to run across all devices. The user interface was heavily designed for mobile and incorporated new gaming technologies. When it was announced, it was seen as a promising update for a brand that was starting to be seen as outdated. There was just one problem. Your computer might not run it.

Windows 11 was positioned as a fresh start for Microsoft. Windows 10 support was gradually phased out, and Windows 11 became the only real way forward. But the shift came with a catch. The system requirements were the strictest Microsoft had ever set. modern 64-bit processor, multiple cores, 4 gigs of RAM, 64 gigs of storage, DirectX12 graphics, a highdefinition display. Individually, these all sound reasonable. Together, they quietly excluded a huge portion of older but still functional machines. But the real breaking point wasn't performance, it was security. Windows 11 required a TPM 2.0 chip known as the trusted platform module. It was designed to store encryption keys and biometric data in a way that protects the physical device from tampering. And unlike CPU or memory requirements, it wasn't something that you could just upgrade easily. Because TPM 2.0 only became common in consumer PCs around 2015, it meant millions of computers were suddenly locked out. It would lead to the greatest PC purge in electronics history.

With one fell swoop, as many as 60% of PCs used for Windows computing were instantly removed from the upgrade path, unable to use the newest software. Some stubborn users powered through with the old technology, but the vast majority just decided it was time to get a new computer. According to Candalyst Research, that upgrade barrier led to a massive coal of technology, sending an estimated 240 million PCs to the landfill by the end of 2025. And that won't just impact your pocketbook. It will impact the world. That's around 105 million pounds of electronics waste. It weighs about the equivalent of 320,000 cars. Folded and stacked flat on top of each other, they would be more than 370 mi taller than the diameter of the moon. It raised serious concerns about an environmental ripple effect, and millions of still working computers became obsolete overnight, feeding into an already growing e-waste problem that's difficult to recycle or safely dispose of at scale. But Microsoft maintained its position. This wasn't an accident. In their view, it was a necessary reset, one that would clear the foundation for the next phase of their product ecosystem. Microsoft had experienced a historic hardware boom during the CO9 pandemic. The strict hardware requirements of Windows 11 promised to force another massive wave of new PC sales to keep the momentum going. Not only would it give them the biggest new product launch of the new era, but it would lead to a buying spree of new PCs by people who needed to meet the system requirements. And it definitely worked. In May 2021, the stock was sitting at just below $250 a share. It surged to 555 a share before settling at its May 2026 point of over 400 a share. Windows 11 has been a winner for the company despite and maybe because of the anger that forced many people to start a new with a fresh PC. But it's not just about the new computer. It's about what's going into them.

Chapter 6. Telemetry debt. The spy in the shell.

One of the biggest topics in computing lately is security. Not just from outside hackers, but from the company itself. Computers are more connected to the internet and to third party applications than ever before. Smartphone users are familiar with this, but until recently, desktop computers were seen as more secure. That might be changing. There's been a recent focus on Windows 11 telemetry. That's the automated process of collecting and transmitting data from your computer to exterior sources. It turns out your computer is always talking. An analysis by the PC security channel showed that Windows 11 is very active in collecting diagnostic usage and performance data as you work. Users quickly began looking for ways to disable it. And while some of it could be limited, a baseline level of telemetry is effectively unavoidable. But Windows 11 goes further than most users expected. Even before the user opens a web browser, the computer is active and analyzing your data. Then it delivers it to the Microsoft servers so it can ostensibly optimize your performance and to sell you things.

According to the critics of the modern OS, Windows 11 is little more than an ad delivery system that you are paying extra for. The software has been known for its display ads and promos as soon as you turn it on, offering you suggestions for new Microsoft products and features based on your ongoing activity. It's not exactly new behavior, but for most people, it is just more hassle than it's worth to turn it off. Digging through the systems and turning off features like personalization as well as upgrading privacy settings can minimize the number of ads that show up. In recent updates, users have reported seeing fewer ads, but users have seen behind the curtain now. Even if they're not seeing the ads, they know that Microsoft is watching them and taking notes. It's a far cry from the clean, easy to use Windows 7. But some worry the privacy issues are just beginning because Microsoft is out to win the next war.

Chapter 7, the final Windows 11 has had several updates since it launched in 2021, but none as controversial as Windows recall. It was a part of Microsoft's broader copilot system, an AI powered assistant rolled out across its products in 2024 through a partnership with OpenAI. Instead of building its own model from scratch like some competitors, Microsoft integrated existing AI technology directly into Windows. The chatbot was supposed to be a built-in assistant that could help write text, automate everyday tasks, and offer real-time suggestions across the system. But what they didn't know was it was spying on them all the time.

Windows Recall was positioned as a flagship feature, but most users would never interact with it directly or even know it existed. That is unless they were watching the press conference where it was unveiled in May 2024. It sounded deceptively simple. It would automatically take a screenshot of the desktop every few seconds, creating a record of everything you do. If you lost any information, it'd be easy to look it up. Your computer now knew what you were doing all the time. And so did any hacker or malware that managed to get into your system. As soon as people heard about this feature, the backlash was immediate. People didn't want their privacy logged to this extent, especially as no one knew about the security of the data, making it worse, the feature was going to be turned on by default. This wasn't so much a backup or security feature as automatic spyware that comes pre-installed on your computer. The criticism from both security professionals and casual users who didn't like AI was so swift that Microsoft postponed the roll out and added an opt-in feature. But the biggest problem was yet to come. It soon came out that everyone's worst fears were accurate. The initial data was being saved on a plain text database with no encryption. This created a comprehensive, easily hackable database of your life directly on your own hard drive. It was like a bank building a glass vault in your living room and then forgetting to put a lock on the door. According to cyber security expert Alexander Hagen, users could have all their activity compromised in a shocking breach of customer trust. And the motivation for this sloppy rollout, Microsoft just wanted to beat Apple to the AI punch. But Apple had its own major AI problems, missing several deadlines and underdelivering. Ironically, that might have been the winning hand because Microsoft's rollout was costly and disastrous. And it all comes down to one question. Does Microsoft even know what it's doing anymore?

Chapter 8. The crumbling foundation.

Windows loyalists have lived through what seems like a neverending parade of software screw-ups, failed launches, and glitches that play casual users against professionals. With each new release, the code gets more and more complex, leading to the current bloated 50 million line monstrosity. So, what could fix it? Can it even be fixed? More and more users are looking at the string of disasters and saying the problem isn't individual releases, it's the foundation. Windows keeps building on itself, but the base is rotting. With each new release, Windows seems to become more complicated and less userfriendly, and many of the new features seem like more trouble than they're worth.

Analysts say the best way to fix Windows might be to simply start fresh, create a new model that would be more suited for the tastes of today's users. But to do that, they would need to ditch the backwards compatibility that's become their bread and butter. They would have to compete with all the established rivals for the market without the security of being able to transition people from one version of Windows to another. They're never going to give up that monopoly. To start over, Microsoft would have to risk everything. So, the safer option is to continue releasing flawed models that aim to improve the problem with the last Windows while creating new ones. The user loses, but Microsoft stock price wins.

Microsoft just made a $3 trillion bet on an AI tool. almost nobody wants. They're charging you $360 a year for it while forcing it onto your computer. And in legal terms, it's not even considered essential software. It's for entertainment purposes only. Meanwhile, millions of users are searching for one thing. How do I delete this? This is why no one is using Copilot. For 40 years, one company built a massive wall around your professional life. If you work at a desk, you live in Microsoft's world. Their strategy is simple. Create a bunch of apps so essential that you were essentially buying the very air an office needs to breathe. But now those walls are cracking. Out of 450 million business professionals who use Microsoft 365, only a tiny sliver has agreed to pay for Copilot, their new AI assistant. The numbers coming out of company headquarters are a disaster. About 15 million people have signed up for the paid service, 3.3%. It's like inviting 100 people to a revolution and only three show up with a weapon.

Copilot was marketed as the new nervous system for the global economy. It was the biggest gamble in Microsoft's history. And they made the wrong bet. A growing number of users are treating it less like an upgrade and more like a digital parasite that they never asked for. It is a staggering disconnect between what the boardroom wants and what the workers are actually doing. This rejection is fueled by a phenomenon known as feature fatigue. For decades, Microsoft users have been conditioned to expect bloatware features that they never asked for that slow down their systems. But Copilot is bloatware on steroids. It eats RAM. It consumes bandwidth and constantly demands attention. Copilot's icon on your taskbar has become a symbol of something that takes up space but gives nothing back. And it's leading to a tipping point. The great UI rebellion. Traffic to debloating websites has surged. Users aren't just avoiding the new features. They're actively searching for ways to remove them entirely. The very features Microsoft spent billions building. For $30 a month, businesses expect a tool to work. They expect accuracy. They expect a program that can handle their data without making things up. And that's not the worst part. The real problem is hidden in plain sight, buried in a document that no one has even bothered to read. In October 2025, Microsoft's legal team stepped in with a bombshell update to the terms of service for individual users. In the fine print, Microsoft explicitly states, "Copilot is for entertainment purposes only. Use Copilot at your own risk." Sure, the $360 a year enterprise tier comes with better legal cover, but under the hood, it is the exact same engine, and that is the part that they don't want to say out loud. The game is up. Millions of people are being forced to use a tool on their own computers that is legally treated like a toy. They are effectively telling you don't trust this with your career and users are listening.

According to data from Recon Analytics, the net promoter score, which measures customer satisfaction, collapsed in just 2 months. It plummeted from -3.5 to a devastating -4.1. In the world of business metrics, an NPS of -4 is a code red. It means the product is actively creating detractors and the percentage of people who want to abandon it heavily outnumbers the people who actually like it. The reason for this anger is the workday test. In a professional setting, any tool meant to save time has to prove it won't waste time. This is where the AI fails completely. Creatives say the text it produces feels hollow, robotic, and it lacks the nuance of human communication. It produces uncanny valley pros that takes more time to edit than it would have just taken to write it from scratch. Technical staff find the code suggestions are often wrong or it uses outdated libraries, forcing them to doublech checkck every single line. It's the hidden cost you pay when you use an assistant you can't actually rely on. If you have to watch your assistant's every move, you aren't using an autopilot, you're just babysitting a machine.

This toxic relationship with co-pilot peaked with the introduction of the recall feature. Microsoft marketed it as a photographic memory for your PC, designed to take a screenshot of your desktop every few seconds. The public saw it as something much darker. It was a gift to hackers and a total end to privacy. Security researchers quickly proved that this snapshot of your entire digital life was stored in plain text. This meant that anyone with access to your computer, no matter how small, could steal every password, every private chat, and every bank statement you ever looked at. The backlash was so violent that Microsoft had to pull the feature and retool it in a desperate attempt to save face. But the damage was done. When you force a surveillance tool onto 450 million people and then disguise it as an assistant, you don't get engagement. Users have started viewing their own operating system as an enemy. And it's why the delete button is winning. People are wondering if the machine is working for them or if it's reporting on them. This is where the software failure turns into an industrial death spiral.

To understand why Microsoft is panicking, you have to look past the code and the icons. You have to look at the physical world. Every 90 days, Microsoft reports its capital expenditure or capex. This is the money that they spend on physical parts, land, buildings, and hardware. In the most recent quarter, that hit 37.5 billion. It's a 66% jump in spending from just a year ago. They are burning through cash. Microsoft is spending enough money to build 100 Burge Khalifas every single year. Microsoft isn't building skyscrapers. They're building a new civilization of computers. They're buying hundreds of thousands of GPUs. And these are expensive, $30,000 each. They're hard to find, and they also run so hot they can melt if they aren't cooled by massive industrial fans and millions of gallons of water. In almost any other business in history, you build the supply because the demand is already there. But Microsoft has flipped the logic on that. They're building the largest factories in human history for an audience that has already tuned out. We're witnessing the creation of computer ghost towns, giant expensive warehouses filled with humming machines that are just waiting for users who are never going to show up.

But Microsoft can absorb that cash. It's the world that's paying for their mistakes. Data centers are the hungriest machines ever built. In Northern Virginia, the data center capital of the world, the local power grid is starting to fail. The grid was built for homes and hospitals, not for millions of GPUs hallucinating poems. Because of this, people are seeing their utility bills spike as power companies struggle to upgrade infrastructure for the AI gold rush. Microsoft has become so desperate for power that it signed a deal to restart the nuclear power plant at 3-M Island, the site of the most significant nuclear accident in US history. This one company now consumes more electricity than many small nations. And for what? A study by Recon Analytics looked at 150,000 office workers in the United States who have a paid license for co-pilot. The results were a nightmare for the board of directors. Only 35.8% 8% of those people actually open and use the tool on a regular basis. In the tech world, this is called shelfware. It's a product a company buys, puts on the digital shelf, and never touches again. Compare that to Chat GPT, where over 83.1% of people who pay for an enterprise license open the app every single day. Microsoft has the products. They own Windows. They own Word and Excel. They put the C-pilot button right in your path. It doesn't get any easier than that. And yet, nobody has time for it. Workers are bypassing the button and opening a separate browser tab to use free tools from other companies. The default placement was supposed to force people to use it. Instead, they created a quiet rebellion.

Behind the scenes at Microsoft headquarters, the panic has reached a breaking point. They're realizing that their own AI might not be smart enough to win. For the last 2 years, the story was simple. Microsoft and OpenAI were the leaders. They had the exclusive edge. But that era is officially dead. Microsoft has officially started integrating Anthropics Claude into Copilot. While the PR team claims it's just about customer choice, it's because their own systems based on OpenAI's GPT are failing critical accuracy tests. By bringing in Claude, Microsoft is hoping that someone else's brain can save their brand. It's like Coca-Cola admitting their secret recipe just doesn't taste good anymore. So, they started mixing in a little Pepsi just to keep people drinking it. In Copilot's researcher agent, GPT now drafts responses while Claude reviews accuracy, completeness, and citations. Microsoft claims this Frankenstein setup delivers a 13.8% improvement in deep research. Yet, even with the patch, the core problem remains. A standalone co-pilot simply could not cut it for professional work. This multimodel approach is the smoking gun. Microsoft spent years touting exclusive OpenAI access as their unbeatable edge. Now they're diluting it with Claude Sonnet 4 and Opus 4.1 across Copilot Studio and Researcher. They're hoping a rival can salvage the product that they're charging an extra $30 for. It's a quiet admission that the quote most important product in company history isn't good enough on its own. The workers already know it when given a choice only 8% choose C-Pilot. They're clicking on that little blue icon less than ever, still reaching for alternatives. The crown is slipping and no amount of model mixing can hide it. The revolution Microsoft bet $150 billion a year on is being rejected one frustrated click at a time. Microsoft is losing 11 to1 in the very office suites they created and sell. The more people used it, the less they liked it. This new toy lost its shine pretty quickly. What was left was a tool that added extra steps to every single job. This is why the 8% number is so terrifying for Microsoft. These are the people who are supposed to love technology. If even they are walking away, the product is in deep trouble. The multimodel surrender is a desperate attempt to fix this. But this creates even more confusion. One day your AI sounds like one person and the next day it sounds like another. The tone changes, the quality changes, the trust drops even further. So many are starting to feel the end of the AI honeymoon.

For 2 years, high-priced consultants told every CEO on Earth that they just had to buy into this revolution or watch their companies die. They spent millions of dollars on licenses. They forced their staff to sit through endless meetings. They bought into the hype. But the data is hitting home. Researchers at MIT recently completed a massive study on these big company AI projects. Their report called the Gen AI divide reveals a brutal reality. 95% of these enterprise AI projects failed to create a single dollar of extra profit. The math behind this failure is simple. Any speed a worker gained by using AI was immediately eaten by the time they spent fixing the machine's mistakes. Teams spent their mornings writing prompts and their afternoons acting as high paid editors for a machine that couldn't pass a basic quality check. The Rand Corporation found the same pattern. Their scientists discovered that these tools don't actually fit into the way any real work happens. Real jobs are messy. They require deep context and absolute trust. AI provides neither. It provides a generic average of the internet, which is rarely what a highstakes professional job requires.

When you compare Copilot to its rivals, the reason for its collapse becomes even clearer. Most professionals now treat Copilot like the internet explorer of AI. They use it only because it's already there. usually because their IT department forced it on them, but when they want real results, they leave the Microsoft ecosystem entirely. Claude is often described as feeling more human and accurate in its reasoning. Chat GPT Enterprise offers a level of data control and speed that Microsoft's forced integration can't match. Workers are choosing to open up a separate browser tab and pay for their own tools rather than use the free one that Microsoft placed on their taskbar.

But the most extreme danger is hidden in the financials, and it should keep every Microsoft investor awake at night. Nearly half of Microsoft's cloud future now depends on a company that is not Microsoft. Around 45% of its remaining cloud commitments are tied to OpenAI, effectively turning one of the world's largest tech giants into a highstakes partner in a single startup success. Microsoft has effectively bet half of its house on a partner that is burning through cash at a record pace. If OpenAI stumbles or if the market realizes that these models can't produce profit, the Microsoft valuation won't just dip, it will crater. Microsoft has tied its reputation, its hardware, and its cloud future to a button on your keyboard that most workers are trying to hide or delete. They are resting a $3 trillion empire on a foundation of entertainment software. Every time a worker right clicks on that co-pilot icon to hide it from their screen, they're participating in the largest consumer rejection in tech history. The 3.3% conversion rate is a wall that Microsoft can't climb. The 95% failure rate is a hole they can't fill. The great AI revolution was supposed to be the moment we moved into the future. Instead, it became a 37 billion lesson in what happens when a company is too big to listen to its own users. The fans in the data centers are still spinning. The servers are still humming, but the offices are quiet. The users have moved on. They have taken their computers back. And as the numbers come in, the only thing left to see is how long the bubble can stretch before it bursts. The era of forced AI is over. And when users walk away, the real fallout starts behind closed doors. Contracts get questioned, partnerships get strained, and the biggest alliance in tech suddenly looks a lot more fragile than it did on the way up. So, what happens when the company funding the AI boom starts to rethink the deal?

This is Josh, and today on the infographic show, we're going to talk about why OpenAI will run out of money, but not for the reason you think. The creator of Chat GBT looks like the king of tech with $20 billion in revenue, but internal spreadsheets reveal something startling. Starting in 2026, they face projected losses of$ 14 billion annually. By 2029, cumulative spending could hit 115 billion. The product works, but the bills are tied to expensive realworld constraints. Here's the thing that most people miss. The massive losses lie in a simple fact. AI is not just another app and it behaves unlike any software we have ever built. In the traditional software world, if you want to make a better app, you hire better engineers. You write cleaner code. It's a human cost. But AI doesn't work like that. It works on something called scaling laws. These are mathematical rules that govern how AI gets smarter. And they are incredibly expensive. The rules are simple. If you want a model to be, say, twice as good, you can't just double your effort. You have to ramp up computing power by a lot. It's basically a brute force equation. Small gains in intelligence mean massive spikes in capital. It sounds crazy, right? But wait until you see the numbers. Training GPT4, the model that really kicked off the revolution, cost roughly $100 million in computing power. That is for one full training run, which is the process of teaching the model from scratch. For a big tech company, that is expensive but manageable. The next generation, the Frontier models arriving in 2026 and 2027 play by different rules. Each run could cost over $1 billion. We have reached a point where a single training session for one AI model costs more than the GDP of some small island nations. And it gets worse. You can't just train it once and walk away. You have to keep on doing it. Open AI is trapped in a cycle where they must spend these billions of dollars just to stay slightly ahead of their rivals. Rivals who are giving similar tech away for free. This creates a fundamental gap in their business model. Their costs are tied to physical realities, electricity and silicon which are expensive and scarce. But their ability to raise prices is limited because there's so much competition. The math is simple and it is catastrophic. Explosive costs are outpacing revenue and the money is running out and the financial bleed gets even worse.

To do the heavy lifting, OpenAI needs high-end AI chips like Nvidia's Blackwell B200s. These aren't your typical CPUs or GPUs. Each one runs $30,000 to $40,000. And you can't buy just one. To train a Frontier model, you need a cluster. That means tens of thousands of these chips, all wired together with high-speed links and liquid cooling systems. And this is where the costs really start to pile up. But the problem isn't just buying the chips. The problem is that these chips have a limited shelf life. Unlike a machine in a factory or a delivery truck, which might run for 20 years, AI hardware doesn't last. It becomes outdated the moment the next generation of chips hits the market. And then companies are playing catch-up. Open AAI has to replace their entire system of chips roughly every 18 months to 3 years just to stay competitive with Google and Meta. Imagine a trucking company having to buy a brand new fleet every 18 months because the old trucks suddenly can't deliver packages fast enough. That is the economic reality of AI hardware. This means the billions of dollars OpenAI spends on hardware isn't a long-term investment. It's an expense that disappears. The value of that hardware drops fast. But if the cost of the chips wasn't enough, there is another bill that's starting to look even scarier. The electric bill. This is best illustrated by Project Stargate. It's described as just a big new supercomput, but it's actually a $500 billion gamble. 500 billion. Yeah, that's right. To put that into perspective, 10 gawatt could power millions of homes. It's the equivalent of multiple full-scale nuclear reactors just for this one project. Why does this matter? Because the costs aren't going away and the grid can't keep up. The scaling costs aren't going away. They're fixed. You can't build the next generation of AI without this level of power. The bottleneck isn't just the cost of electricity. It is the national grid. Getting enough high voltage transformers and grid capacity is a huge hurdle. The old utility system can't grow fast enough to keep up. So, OpenAI is now in the position of negotiating for direct access to nuclear power and massive solar farms. These utility costs create a high floor for their operating expenses. Every free Chad GPT user is literally costing billions and there is no way around it. It makes it nearly impossible to maintain healthy profits when you are trying to offer a free tier to hundreds of millions of users. Every time someone uses ChatGpt for free, OpenAI has to pay for the electricity and the silicon wear and tear. So, if OpenAI is losing billions of dollars on chips and electricity, how are they still open? How do they pay their employees? And that leads us to one of the most misunderstood pieces of the OpenAI story. Its deal with Microsoft. We often hear that Microsoft has invested billions into OpenAI. And on paper, it looks like billions came in. In reality, it's more like a financial merrygoround that hides how tight the startup's cash really is. When Microsoft invests billions, a lot of that money doesn't actually leave Microsoft. They give OpenAI cloud credits instead, sort of like a gift card. And you might think that that counts as real cash. It doesn't. OpenAI can record it as capital raised. So, it looks like cash, but the credits have to be spent on Azure, Microsoft's cloud service, to run their models. This effectively recycles the investment back into Microsoft's revenue stream. It boosts Microsoft's cloud earnings and stock price. But here's the dangerous part. You cannot pay your employees with cloud credits. When OpenAI hires a top researcher for $2 million a year, they need hard cash. When they have to pay for office space or legal fees, they need money. This creates a financial optical illusion. Microsoft invests 10 billion, but that money doesn't actually land in OpenAI's account. It's basically digital coupons that can only be spent on Microsoft servers. The result is massive pressure. Every fiscal quarter, OpenAI has to raise hard cash from other investors just to pay payroll and cover bills that Microsoft credits can't touch. If the flow of new outside investment slows down, OpenAI faces a cash flow crisis. They might have plenty of computer time, but not enough hard currency to keep their team from leaving for rival companies.

Despite all those costs, investors keep on pouring money in. In March 2025, OpenAI managed to raise $40 billion, the largest private funding round in history, even bigger than the IPO of the oil giant Saudi Aramco. But here is what is really odd about it. Saudi Aramco has hundreds of billions in revenue. And more importantly, it has real tangible assets, oil reserves that you can measure and sell. Open AAI is a startup with no profits burning cash at a rate of billions a year. Its value is mostly intellectual property which anyone can try to copy. So, what does this mean for the long-term survival of Open AAI? The answer will surprise you. Investors are pouring money in based on the promise of a market that doesn't fully exist yet. For Open AI to be worth a trillion dollars, it can't just be impressive. It has to replace dozens of cheaper tools that companies already use. Right now, most businesses spread their AI budgets across multiple smaller providers, not just one giant system. OpenAI is building something massive and expensive, betting that eventually everyone will need it. But right now, there's no guarantee of that demand. And this leads us to the risky business model. In software, companies survive by making it hard for customers to leave. Salesforce does this because moving all your data is a huge pain. Netflix does this because they own shows that you can't watch anywhere else. Open AI is discovering a hard lesson. Users are mercenary. If Google's Gemini or Meta's Llama offers a similar answer for cheaper, they'll leave instantly. About 75% of OpenAI's revenue comes from consumer subscriptions. But the number of cancellations is rising. And once the novelty fades, most users won't pay. Big business is even more skeptical. Only about 20 to 30% are sticking with OpenAI's API long term. Many are choosing open- source models like Llama to keep data private and costs down. With nothing keeping them tied to OpenAI, no built-in network, no way their data is stuck. They could just jump to another provider overnight. And the competition is just as deadly as OpenAI's own cash burn. Meta's decision to release the Llama models for free was not an act of charity. It was a tactical strike. When Mark Zuckerberg gives everyone access to their top-of-the-line AI for free, he effectively sets a ceiling on what OpenAI can charge. Meta can burn cash on open source models because they're using the tech to improve ads on Instagram and Facebook. Their business isn't selling AI, it is selling ads. Open AAI doesn't have that luxury. Their only product is the AI itself. They're fighting to establish themselves while their competitors aggressively undercut the market to keep them from gaining ground. And the clock is ticking. Open AI is squeezed from all sides. On top, giants like Microsoft and Google with practically unlimited cash. On the bottom, lean competitors like Anthropic and Mistral. Anthropic runs a much more efficient operation, focusing on safety and enterprise reliability with a much lower burn rate. Meanwhile, Google's DeepMind keeps stealing talent, forcing Open AI to offer massive stock-based pay packages. Those only work if the company's valuation keeps climbing. If it stalls, the researchers, the company's only real asset, could walk out the door. As if burning billions, fighting competitors, and losing talent weren't enough. Regulators in Washington and Brussels are circling. In early 2026, the FTC and European Union intensified their antitrust probes into the Microsoft OpenAI partnership. Regulators are checking whether Microsoft's investment is actually a deacto acquisition designed to skirt merger laws. If they decide to limit the power Microsoft has over open AI or force a split, it would cut the startup's financial lifeline. And then there's the mounting geopolitical friction. Export controls on AI chips are shrinking the global market, while new AI safety regulations are creating a massive compliance burden. Open AI now needs armies of lawyers and safety researchers. Rules that are costly and generate zero revenue. The danger becomes clear when you look at history. Uber lost billions before its initial public offering or IPO. But it was building a physical network in thousands of cities. Tesla struggled for years, but it was building factories and a global charging network. Something real that competitors couldn't copy overnight. Open AAI, well, it's burning billions with no real network or physical assets to lean on. OpenAI's production is all about raw computing power. The expensive chips that mostly come from Nvidia. Unlike Tesla or Uber, OpenAI's product loses money every time someone asks it a complex question. And there is nothing stopping users from leaving tomorrow. The company is now effectively betting everything on a single desperate timeline. They're racing to build artificial general intelligence or AGI, an AI that can think and learn like a human before the bank account runs out. This isn't a standard software business strategy anymore. If OpenAI can build a model smart enough to do the work of a human expert in any field, their current cash burn wouldn't matter. Revenue could in theory skyrocket. They're picturing a world where their AI doesn't just summarize emails, it replaces entire departments, handling corporate taxes, writing complex code, and planning strategic business moves at superhuman speed.

That milestone and they could charge a premium that covers any debt, no matter how massive. If OpenAI is losing 14 to 17 billion a year, every month of delay costs over a billion. If the breakthrough to AGI takes 5 years instead of two, they'd face a funding gap of nearly $100 billion just to keep the lights on. And no investor can fix that overnight.

So what happens when the money runs out? You might expect a dramatic crash. But the reality is different. The most likely outcome is not a dramatic crash or a bankruptcy filing, but a quiet absorption. By mid 2027, based on current projections, the cash reserves raised in the 2025 rounds will be nearly empty. At that point, OpenAI will face a choice. Raise another massive round at a lower valuation, crushing their employee stock options, or sell.

Microsoft is the natural and maybe the only buyer. They already host OpenAI systems on Azure, and they have deep integration with the software. More importantly, Microsoft has over $80 billion in cash reserves, making them one of the few entities on Earth that could sustain OpenAI's burn rate. For Microsoft, this is the crown jewel, the engine of the next computing era. For investors, it's a fire sale, but one that buys survival.

This is the end of the startup frontier. Open AI proved scaling works, but only if you have a nation-state sized budget. The AI revolution has gone industrial where success is measured in acres of data centers, not lines of code. OpenAI started the trend, but it doesn't have the resources to compete alone. And now the independent pioneer is likely to be absorbed by a larger corporation.

Your boss might be watching you right now. Even if you work from home, the moment your laptop connects to a company network, you have basically invited management to pull up a chair in your home office and spy on you. And in the United States, it is virtually 100% legal. There's even a nickname for it, boss. They know when you slack. They know about every message you send, and they can read them. Their eyes are always on you. It leaves you facing an uncomfortable truth. You're not just working remotely anymore. You're working under observation.

Chapter 1. The green dot illusion. If you're one of the roughly 320 million people who use Microsoft Teams since 2024, you're more than familiar with how it forces you to perform or risk losing your job. If your status light isn't green, your employer starts to get suspicious about how long you've been away from your desk. There are even articles available that give you tips and tricks on avoiding the dreaded idle status. It's all part of something called productivity theater. At this point, it's become a whole industry unto itself. All to the end of keeping that little green Teams light on. It's why if you go to any online retailer, you'll find so many options for mouse jigglers, a desk gadget that keeps your mouse eternally shaking to avoid it ever going idle. But Microsoft has invented the ultimate curtain call for productivity theater, the unified audit log. Those three words might land you on the unemployment line.

For management. One of the biggest advantages of Microsoft Teams and its unparalleled integration into the Microsoft 365 suite is the extremely powerful insights into your overall app interaction telemetry. That word is incredibly important to understand for all of this to make sense. IBM defines it as "the automated collection and transmission of data and measurements from distributed or remote sources to a central system for monitoring, analysis, and resource optimization." And that is exactly what the unified audit log makes possible. Your overall status and whether your mouse is moving is a worthless metric. Bosses will now be able to get a centralized data profile that tells them which apps on the 365 suite you've been interacting with and how often. They'll know if your camera is on in the meetings, what messages you're sending, and how you are spending company time. Thanks to Teams, they can now know at the click of a button, regardless of what color your little status light is.

But that is not even half as sinister as the other thing that the tech allows your bosses to do. Know exactly where you're working at any given time, like they've implanted a tracking chip right there in your laptop or phone.

Chapter 2, the Office Spy. One of the latest features on Microsoft Teams will effectively allow your boss and all of your co-workers sharing the same work Wi-Fi network to stalk you. This isn't just a fear-mongering exaggeration. It's real. Microsoft has declined to comment in interviews, but the application will automatically detect your location whenever you connect to company Wi-Fi. This information will not only be easily accessible to server admins, but also your co-workers. One of the major so-called benefits of this new feature is that it allows you to detect and connect with your co-workers on projects. But anyone who wasn't born yesterday could tell you why that excuse doesn't stand up to scrutiny. If you want to co-work on a project with a fellow co-worker, you can just send them a message. The only people who would benefit from knowing the exact location of all their workers at all times when connected to company Wi-Fi are your managers. They'll be able to see if you're a stack of potential profits or an unacceptable cost. Collecting reams of data on you is how they think they can tell the difference without actually getting involved.

It's important to remember that the system admins at your workplace need to choose to turn this thing on. If they do, depending on your state, they legally only have to give you notice. But bear in mind, this isn't always so straightforward. They might hide this caveat in a larger user agreement, just expecting you to click "I agree" without reading. Or they might try to sell it to you through convenience, saying you'll never need to manually set your work location again. The only time you'll have power over this is before you sign away your right to privacy in the first place.

Chapter 3, the keystroke trap. You might think based on these two cases that if your workplace doesn't use Microsoft Teams, you're in the clear. You'd be wrong. Technical employee surveillance is a thriving industry. With the advent of AI and the never-slowing-down software boom, new methods are being added to the spying toolkit every day. Take for example keystroke logging. This is somehow even more invasive and terrifying than the other options. And major tech companies like Meta are already doing it. As of late April 2026, Meta installed software on all employee devices that will track every key press as they type on their keyboards, every click they make, and even their mouse movements. All supposedly to train their AI models to perform these tasks autonomously. This dystopic technology is called the model capability initiative or MCI and it is awful for employees for two different reasons. First, it is literally watching everything you click and type while at work and storing that information. If you use certain terminology that might be flagged inside the system, like "pay raise" or "union" in chats that you naively think are private, you might suddenly find the boss looking at you an awful lot closer. But even worse, if your company uses a program like this, it means that they're getting you to provide free on-the-job training to your own replacement. To you, it's the destruction of your livelihood and potentially even your industry.

To management. They might reframe it as data-driven cost cutting for the shareholders.

Chapter four, the AI risk score. It seems like nothing is sacred in the office, including your emails and browser history. If you're using a work computer, a work server, or even a work email address, all of these things are not your property. They belong to your employer, and that gives them sweeping rights to run through the content with a fine-tooth comb. Your company IT department knows every embarrassing question you've ever Googled on company time. Your boss can't just see the subject line of your email, its recipient, and the time it was sent, but also the body text, and any attachments you had on it. None of this is off limits. None of it.

Brian Crop, chief of research for Gartner's HR practice, says, "Anything that you write on any company messaging platform, your employer has access to, either through it or HR. Anything you put on these platforms, your employer can look at." Demand for the technology has only improved the technology over time. Services like Sentry PC, Work Examiner, and I MonitorSoft all boast about one particular aspect of their functionality. If their system detects unacceptable employee behavior, and the parameters for this are of course selected by the employer, they'll secretly take a screenshot of the moment they view as a breach for the employer to keep as evidence. Imagine this. You're typing fast and a perfectly innocent word that you were typing now reads as something offensive because you clicked the wrong key. This isn't just a misunderstanding anymore. If your typo is on a secret internal list of unacceptable words or phrases, the screenshot has already been taken and sent directly to your boss to incriminate you. If your boss has already been looking for an excuse to get rid of you for whatever reason, a piece of AI software has just given them the silver bullet and then painted a target on your forehead.

And that is still not nearly as bad as the ways your bosses can invade your privacy if they have bought a subscription to Verato Cerebral Security. It's one of the more high-end employee stalking software services on the market. They have taken the screenshot violation even further with their time capsule feature. It'll capture a continuous video recording of your screen if the software has reasons to believe that you are violating the company code of conduct. Thanks to AI integration, not only do Verato and the hundreds, if not thousands of companies like it use their recordings to gather massive data packets, it also creates profiles on every employee. Their profile will be the product of patterns that the software identifies when it's spying on you. How many files do you download and upload on a daily basis? How much time do you spend interacting with each workplace application? How do you message? How many messages do you send? The purpose of all of this is behavior analytics, allowing them to identify anomalies in your behavior. In other words, if you start acting outside of the pattern that the software has predicted for you, it'll flag you as behaving strangely and let your boss know that something is wrong. Just because a single workday meant that you downloaded more files than usual or sent more emails, suddenly you're under a microscope. We all know just how often AI makes mistakes or suffers from hallucinations. But does your manager know that? If they're buying one of these services, chances are they trust it. And if it randomly declares you the company problem child, it's your word against the service your boss is paying for. They even use all the data they collect to compile a risk score on every employee. It assesses the likelihood that you'll be a security threat to the company. And this is all true. Very own marketing materials brag that it runs in stealth mode by default, meaning that users would have no way of even knowing they're being watched. It's like an invisible person standing behind you right now, staring over your shoulder. Even if it doesn't detect wrongdoing, it'll just take a screen cap of your device every 30 seconds.

Chapter 5. The digital panopticon. Paranoia of being observed breeds obedience. This whole idea was created by English philosopher Jeremy Bentham in 1786, almost 200 years before John Welder Tucky first used the word software in the context of computing. He called it the panopticon, a thought experiment around a perfect prison where prisoners believed they could be watched at all times. And so they always behaved themselves. Now we're seeing Bentham's dream realized across the whole workforce as intricate nets of software webbed together to create a global digital panopticon with every employee at every company trapped inside. And in an effort to stay on top of it in a competitive bossware market, companies are creating increasingly bizarre and invasive services to set themselves apart. Sneak is a piece of software that likes to present itself as a conscientious mental health aid for employees who miss the workplace atmosphere. But is your mental health really improved by your boss having the ability to conference call you at any time without a decline button? Protocore is an AI service that relentlessly tracks your activity and uses it to create a numerical productivity score for your boss. And if you have a position that has a pretty nuanced view of productivity, then there is an excellent chance this one could get you into trouble. And software designed to track your so-called active work can genuinely eat into your paycheck. As workers like finance executive Carol Kramer found out, her $200 an hour salary resulted in a truly pathetic take-home pay because her employers use tracking software to only let her bill for the specific minutes that she was actively working. And this isn't an isolated incident. Eight of the 10 largest private US employers track the productivity metrics of individual workers, many in real time. According to an examination by the New York Times, some services like Flexi Spy, Spy Eera, Clever Control, and I Monitor also provide call tapping functionality. Even more of these services offer webcam and mic surveillance, audio monitoring, and even mobile device access, so they can really get you anywhere. According to Surf Shark, information on these services is even gathered into a centralized war room interface inside the software. The thing they're declaring war against is you ever having a sense of privacy again.

But it's not just Microsoft. You are not safe on some of the most ubiquitous workplace apps on the market. If your workplace is using Slack and your company is buying the enterprise plan, it's easy for administrators to store and view your messages from everyone using the system underneath them. If the company you work at uses Google Workspace and subscribes to plans like legacy G Suite Business, Google Workspace Business Plus, and Enterprise Standard, they can use the vault. This ominously named feature allows them to archive and search specific content from the company's employees across Drive, Gmail, Groups, Chat, Voice, Classic Hangouts, and Meet.

The fact is, all of this is terrifying, and we have the stats to back it up. A study conducted by ExpressVPN found that 56% of employees feel stressed or anxious when they suspect monitoring is happening without their knowledge. Peter Holland, a professor of human resource management from Swinburn University, has gone on record saying that he thinks relentless productivity monitoring actually has the exact opposite of the intended effect. During an interview on the subject of bossware, he said the research indicates that the more you monitor and surveil people, the more likely it is that they are going to be less productive. People have spoken about being bullied. But despite this, market forces seem to be increasingly pressing management to take an aggressively data-driven approach to running their businesses. The result, they can no longer see the forest. In their desire to cut costs, they cut down all the trees.

Chapter 6, the office revolution. The first line of defense will always be your own right to legally consent to this kind of spying. Check any work contracts you signed when you took the job for clauses that say you consent to workplace digital monitoring. It's a tactic employers rely on using vague, carefully worded policies, so you never fully realize how much of your privacy they're actually invading. But you can always check your computer by pressing Control+Alt+Delete or Control+Shift+Escape. You can open up your computer's task manager. This is essentially popping the hood to see what software is currently running. If you spot any of the apps we mentioned or others you don't recognize, you may have just found proof that your boss is quietly looking over your shoulder right now. If you're using a Windows device, you can also use the MS-DOS command line, which means pressing Windows key + R and then typing "cmd" into the search box. From there, you can search "tasklist" and find the same information on all the programs currently running. Another option is analyzing network traffic for sudden spikes during work hours. However, this one is a little more imprecise. There could be multiple explanations for these spikes, and you're not going to have any certainty from just the data alone. If you thought that, you'd be falling for the exact same trick all these expensive software services are pulling on your boss.

In order to prevent yourself from getting caught out by any boss, it's also important to practice data hygiene. This means you should keep any work friends group chats off the official servers and on separate devices. A safe mantra to internalize is: never say or do anything on a work device or on a work network that you wouldn't be fine with your boss immediately being able to see. Sometimes the services your boss uses won't be as stealthy as some of the premium packages. Not everyone can afford to predict all their employees' risk factors. After all, if your browser comes with a message that says "Managed by your organization," then don't make any literally not safe for work searches. And treat it with healthy suspicion if your boss asks you to connect to a specific VPN while working because it might be one that allows traffic inspection from the administrators. And of course, if you don't mind being a little more confrontational, you can always buy and download some reputable anti-spyware software. And even if your boss confronts you over it, at least you'll know that you still have the moral high ground.

The workplace isn't the only place where tensions are rising. Companies might be watching employees more closely than ever. But the tech giants building those systems are starting to turn on each other, too. And the biggest fracture might be happening inside the AI industry itself. Everyone thinks Microsoft won the AI war, but they may have just lost it overnight because Sam Altman just made a $50 billion move with Amazon to stab Microsoft in the back. And almost no one noticed. On the surface, it looks like another massive AI deal, but hidden inside it is a shift that sidelines Microsoft and removes the one clause that actually kept control in check. And that changes everything. Because this isn't just about building powerful AI anymore. It's about who owns it, who controls it, and who's willing to burn billions to get there. By the time most people realize what just happened, the balance of power may already be gone.

Chapter 1, the $50 billion betrayal. To the public, OpenAI looks like one of the biggest success stories of the modern era. A Silicon Valley startup that turned into a household name almost overnight. It built tools used by millions, pushed AI further than anyone expected, and wrapped it all up in a mission to benefit humanity. That was the illusion. In reality, OpenAI is a mess that's bleeding cash. In 2024, its projected revenue was approximately $3.7 billion. Its losses: $5 billion. That's like buying a mid-sized airline and setting the whole thing on fire every 12 months just to keep the servers running. That's not a sustainable or successful business model. That's not a sign of a healthy, stable company. And it wasn't a one-time thing. It was a trend. In the first half of 2025, figures revealed that OpenAI was losing extraordinary amounts of money, generating around $4.3 billion in revenue while recording losses of up to $13 billion. Some estimates suggest the company's total losses could exceed $140 billion between 2024 and 2029 alone. But that is not that surprising. Training frontier AI models isn't cheap. Neither are the salaries of leading researchers and computer scientists or the construction and operating costs of data centers. OpenAI is burning through money at breakneck speed. Its entire business model is founded on the idea of convincing investors that someday, somehow all of this loss will be worth it. Microsoft bought into that idea. It poured billions into OpenAI and secured what seemed like an exclusive hold over the most valuable AI startup on the planet. Microsoft CEO Satcha Nadella even said that it wouldn't matter if OpenAI disappeared tomorrow. "We have the data, IP rights, and the capability." Nadella thought that for all intents and purposes, he owned OpenAI. He was wrong.

In February 2026, Sam Altman orchestrated an enormous $50 billion infrastructure deal with Amazon. In doing so, he effectively ended Microsoft's exclusive cloud rights. That wasn't supposed to happen. Microsoft was supposed to be the only serious player in the game. Microsoft Azure was meant to be the default home for OpenAI's technology. Amazon was the rival, the company that you compete against, not partner with. This wasn't a new vendor agreement or just some sort of multi-partner strategy. It was OpenAI blatantly betraying the tech giant that helped build it. The question is why? Why would Altman risk the wrath of Microsoft? Why jeopardize what seemed to be the most powerful relationship in the industry? Because the Amazon deal wasn't just about getting more servers or resources. It was a weapon built for one mission: to defeat a more powerful enemy, the United States government.

Chapter 2, the FTC's trap. For years, it looked like compute was going to be the biggest challenge OpenAI would ever face. But the funding from Microsoft introduced OpenAI to something else: antitrust. To observers and analysts, including those in government authorities, such as the Federal Trade Commission, this didn't look like one company simply supporting another. It looked like a merger. Naturally, Microsoft and OpenAI didn't label it that way, but the facts were clear to see. Microsoft had poured in billions and secured exclusive rights to its Azure ecosystem. OpenAI technology was also becoming increasingly integrated into Microsoft's most-used systems and applications from Windows to Copilot, Office, GitHub, and beyond. OpenAI, meanwhile, was looking less like an independent organization and more like a subsidiary, just with its own separate branding. The FTC noticed, so did other tech brands. Google even called on the government to investigate and break up the deal. Critics and regulators argued that Microsoft's massive investment and exclusive cloud control over OpenAI's technology gave the company too much influence. It hadn't just invested in an up-and-coming company, it had bought the future. So, the FTC started investigating the two companies. If it could prove that they had effectively entered a de facto merger or that Microsoft had acquired an unfair monopoly over the AI industry, it could force the pair to split. Microsoft would be able to survive that. OpenAI might not. It couldn't afford the risk and it had to find some way to wriggle out of its predicament.

Enter the Amazon deal. By pivoting to Amazon Web Services or AWS, OpenAI gave itself a multi-billion dollar legal shield. Because now, if the FTC's investigators question the company's allegiances or argue it's too friendly with Microsoft, OpenAI's lawyers can simply say, "How can we possibly be a subsidiary of Microsoft if we just signed a $50 billion contract with one of their biggest rivals?" It was the perfect piece of legal theater at just the right time because antitrust cases are built on dependency. Regulators are very wary of any company that appears to be entirely or exclusively dependent on another for its survival. But by inking an agreement with another tech giant, OpenAI proved its independence. Problem solved. Or at least that's how it seemed. In reality, there was much more to this story than meets the eye. Escaping the FTC was only a convenient and timely byproduct of a deeper and darker imagination. The real reason OpenAI needed leverage over Microsoft was hidden inside a bizarre legal contract signed years before. A contract that contained a ticking time bomb.

Chapter 3. The AGI poison pill. For years, OpenAI told the world that it is working toward AGI, artificial general intelligence. Sometimes known as the God model. This is said to be the point at which AI effectively reaches and then surpasses human-level intelligence. According to the experts, AGI will be able to think, reason, and adapt just like a real person. It'll solve problems and switch from task to task rather than being pre-programmed with just one specific function or avenue of activity in mind. In effect, this was OpenAI's justification for everything. All the funding, all the hype, all the resources, it was all said to be in service of the AGI experiment. And for OpenAI, it was the perfect panacea. All they had to do was convince people to trust them, ignore the obvious problems, hand over their money, and then believe that someday they'd build something that would change the world.

There was just one little problem, one buried deep in the contracts tying OpenAI and Microsoft. It was known as the AGI trigger. Basically, Microsoft was granted a seemingly perpetual license to OpenAI's intellectual property. But that perpetual license had a strict limit. As soon as OpenAI achieved artificial general intelligence, the new AGI model would be entirely excluded from the deal and Microsoft would effectively lose its grip on the future of AI. At that stage, all commercial rights to the AGI would remain exclusively with OpenAI. It's a paradox, a snake eating its own tail. Microsoft was pouring billions into a company whose sole stated mission was build AGI. But as soon as that mission was achieved, Microsoft would lose all of its power and benefits. It was funding its own demise.

But Microsoft's executives aren't idiots. They knew the terms of the deal when they signed it. They knew exactly how to work around them. All they had to do was ensure that OpenAI failed at its stated mission. They wanted the AI to be powerful, but never powerful enough to reach AGI. That's where things get complicated. Because AGI isn't a clear finish line. No one can agree what it actually means. So even if OpenAI pushed its systems further and further, Microsoft could always argue it still wasn't AGI. Sam Altman knew this. He knew that as long as the AGI trigger clause existed, Microsoft would never truly be an all-in partner. It would always have leverage, always have limits, always have a way to control OpenAI from the inside. So he had to change the play. With a $50 billion Amazon deal as its loaded gun, OpenAI forced Microsoft back to the table, not just to negotiate, but to completely rewrite the rules of the AI war.

Chapter 4, the April 2026 reset. In April 2026, the balance of power shifted. It wasn't a simple partnership update. OpenAI didn't want to iron out just a few issues or make a couple of amendments to its Microsoft deal. It wanted to demolish it and then rebuild it exactly as it saw fit. On the surface, the two companies saved face, announcing a simplified agreement and next phase for their partnership. But the terms of that agreement painted the real picture. Microsoft was still described as OpenAI's primary cloud partner, but the very next sentence added that OpenAI was now free to serve products across any other cloud provider it wanted. Azure's exclusivity was gone. Microsoft's once perpetual license to OpenAI's intellectual property was also amended and given a fixed end date of 2032. The tech giant no longer had privileged ownership of the future of AI. Revenue sharing was officially given a cap and a deadline of 2030 as well. Most importantly, the AGI trigger was gone. It wasn't redefined, clarified, or amended. It was deleted. That vague, hard-to-define clause that hung over OpenAI like a sword of Damocles for years was gone. It was replaced with something far simpler and far more controlled: a calendar with dates, deadlines, and caps. In other words, a standard corporate agreement. Microsoft's stake was also formalized at approximately 27% of the company. That is still a sizable amount, enough for Microsoft to hold some level of influence over the company's activities, but nowhere near enough for complete control.

It looked like a big victory for OpenAI. The company had won its independence, decoupling its finances from the mythical AGI milestone. It was no longer a research lab trying to trigger a clause in a contract to win its freedom, but a corporation with a clear runway ahead. But there was a catch, a big one. To pull off this extraordinary corporate coup, Altman had to permanently destroy the very foundation that OpenAI was built upon: its nonprofit structure.

Chapter 5. The death of the nonprofit. OpenAI's nonprofit nature was the one thing that separated it from every other Silicon Valley machine. This wasn't just another power to the benefit of humanity. A nonprofit lab working with care and consideration towards something that was supposed to bring great benefits to all. This wasn't Google. It wasn't Meta. It wasn't worried about pleasing shareholders because there were no shareholders. But that idealistic attitude couldn't last. Slowly and surely, the cracks in the mask began to appear. In late 2025, the facade was ripped away entirely. OpenAI shifted from a nonprofit to a public benefit corporation. At a glance, that still sounds like a righteous cause, a compromise between the original mission and a need to make money. PBCs are supposed to strike a balance between pursuing profit while remaining committed to creating a positive impact on society, the community, or the environment. In reality, PBCs still serve their investors almost as much as any other for-profit entity. Just as Altman would go on to smash and then rebuild his deal with Microsoft, he also destroyed what OpenAI once was, reconstructing it as something completely different. The groundwork was laid back in 2023. The organization's original nonprofit board, the one that briefly fired Altman, was removed and replaced by Silicon Valley insiders and former Treasury officials. The safety guardrails that had once been so critical to the organization's overall mission were dismantled. The systems that had kept OpenAI's progress in line with its focus on helping humanity were gone. In their place were product safety teams more concerned with ensuring that their AI doesn't say anything that might offend a big B2B client than actually harm real people. The cogs inside the OpenAI machine were replaced piece by piece until something fundamentally changed. What began as a research-driven system started to look like something else entirely: a profit engine, one that was preparing for a massive IPO, and was increasingly insulated from any meaningful ethical oversight. Behind closed doors, Altman and OpenAI's research leads realized a terrifying technical truth. They weren't moving toward AGI as they originally expected. They were moving towards a brick wall.

Chapter 6. The scaling wall. For years, the AI industry has relied on an unwavering belief in a premise known as scaling loss. If you add more data and more compute, your AI models will become exponentially smarter. It's all a question of resources. Provide more resources and you get a better product. All companies like OpenAI had to do was keep on building bigger and better data centers. They had to invest in more powerful chips and processors. Then they could sit back and watch as their AI followed the linear path to godlike intelligence. Sounds pretty straightforward and the scaling laws worked for a while. Each new generation of AI technology felt like a big leap forward. GPT-2 was impressive. GPT-3 next level. GPT-4 exceeded expectations. So naturally GPT-5 came. Orion was expected to be a game-changer, maybe even the final step toward AGI. Or not. Internal reports suggest that the scaling laws are stalling. Instead of providing some sort of quantum intelligence leap, GPT-5 has hit diminishing returns. It's still getting smarter, but at a slower rate than ever before. All of the sudden, this god model that seemed right around the corner is now a speck on the horizon. This isn't just a hurdle, it's a catastrophe. The scaling wall changes everything. It's no longer a situation where companies can just pour in money and watch their AI become twice as intelligent overnight. Now they're spending billions for only incremental improvements. And that is bad business. Because let's not forget about the burn rate. OpenAI is nowhere close to making a profit. It loses billions each year, but it was always able to justify that with the claim that AGI would eventually arrive and fix everything. The company's entire financial structure and investment incentives were reliant on that premise. That's why removing the AGI trigger clause from the Microsoft contract mattered so much. It wasn't just a legal trick. It was a quiet admission that AGI is a mirage. And if AGI is a mirage, OpenAI is just another software company and one that is failing to provide returns and plateauing fast. So, how does a company like that justify a $5 billion burn rate to its next investors? It stops selling AGI and starts selling AI slop instead.

Chapter 7, the SAS pivot. Selling AI slop. Meet the new OpenAI. It's no longer a valiant nonprofit pursuing civilization-changing superintelligence, but a salesforce for AI business. It's slowly but surely pivoting away from its original mission and towards something much more mundane: enterprise software and corporate workflow automation. And it's happening right before our eyes. Rather than focusing its efforts exclusively on the next evolution of GPT technology, OpenAI is prioritizing alternative projects like agentic middleware and reasoning models like 01 or Strawberry. It's no longer charting a course toward human enlightenment, but making life easier for middle management. The focus has shifted to producing tools built for middle management, routing leads, generating marketing copy, handling support tickets. OpenAI hopes that this shift will bring in the money it needs to satisfy its investors. But there is a massive problem with that plan. The numbers don't add up. Traditional SaaS or software as a service companies like Salesforce and Adobe operate on incredible margins, often exceeding 70%. They make a product and they basically sell it forever, bringing in more and more profit with every new customer. That's why investors love SaaS. It is a gold mine. But OpenAI's attempts to enter this industry are not working because advanced reasoning models cost so much more to run than conventional software. 01, for example, costs around $15 for 1 million input tokens. That might not sound like much at first glance, but in enterprise terms, it is a massive financial burden. Big businesses with hundreds or even thousands of employees can chew through millions upon millions of tokens in a single day. Suddenly, a smart assistant is not a cost-effective component of the tech stack, but a very expensive capital drain. Traditional SaaS doesn't work this way. Microsoft doesn't bill businesses every time they open a new spreadsheet on Excel. CRMs don't suddenly become twice as expensive just because employees clicked a few buttons and generated some reports. AI works differently. The more you use it, the more expensive it gets. OpenAI is desperately trying to force businesses to integrate overpriced automated AI slop software into their corporate workflows to fix its own broken economics. It is trying to sell digital gold for the price of lead, hoping to convince people its money-guzzling AI agent isn't just really expensive software. But to make its margins work, it needs to dramatically decrease its own operating costs. It needs cheaper compute, which brings us back to the $50 billion Amazon Trojan horse.

Chapter 8. Amazon's Trojan horse. The deal with Amazon wasn't about escaping the FTC's investigations or breaking free of Microsoft's shackles. It was about hardware. Part of the $50 billion commitment that Amazon made to OpenAI includes a multi-year agreement for the AI firm to use AWS's Tranium chips. Prior to this, OpenAI was a hostage to Nvidia. It was forced to use Nvidia's H100 GPUs. It's the hardware that everyone in the AI industry wants and needs. The chips that form the beating hearts of AI data centers. These GPUs have proven highly effective in training and improving AI. They are also expensive. Each one can cost tens of thousands of dollars, and that's before the added expense of building the rest of the server around it and actually running the whole thing. Large training clusters come with billion-dollar price tags. OpenAI has paid an H100 tax on every single prompt it processes for years with incalculable amounts of money funneled away into the accounts of Nvidia and Microsoft. The Amazon deal gives OpenAI an off-ramp. Tranium chips aren't necessarily better than Nvidia's H100s, but they do have the potential to be much, much cheaper, up to 50% cheaper according to early estimates. They're also said to consume 40% less energy. By switching to Amazon's own custom silicone, OpenAI hopes it'll be able to make some significant reductions to the cost of its tokens. Cheaper tokens should make OpenAI's AI slop easier to digest for its big business customers. They might even give the company a slim chance of turning a profit before the 2030 revenue cap hits. In the name of saving humanity and building a tech utopia, OpenAI took a very different path: building software on proprietary Amazon chips running inside Amazon data centers, selling AI agents to Fortune 500 companies. This is not what the company's founders envisioned all those years ago. This is not a beacon of open-source enlightenment. It's just a cog in the AWS machine. So, where do we go from here?

Chapter 9, the great AI realignment. The hardware war is over. Unfortunately, humanity didn't win. Instead, the victors are the corporate behemoths that own the silicon. These companies with the money and power to do whatever they want and always get away with it. OpenAI sold the world a dream. A dream of godlike AI that would cure cancer, solve the climate crisis, and bring about a new world where everyone would be happier, freer, and more fulfilled. That utopian dream is dead, replaced by a dystopian corporate reality. Microsoft, Amazon, and OpenAI aren't laying the foundations for a more prosperous and creative age of human advancement. They're building their own locked-down and ludicrously expensive B2B monopoly. The open marriage that now exists between those tech giants all but guarantees that the future of the internet will be flooded with corporate AI slop, automated emails, synthetic reports, and agentic workflows that don't actually provide real benefits to real people. They just streamline the capitalist machine while eroding the value of human thought and creativity. Sam Altman didn't escape from Microsoft to bring about a better world. He broke free so that when the trillion-dollar IPO arrives, he and his shareholders will make as much money as possible. This is the grim reality of AI today. We're not getting AGI. We're not going to see some digital god that solves the world's ills and makes us all happier and healthier. We're getting an inescapable automated corporate bureaucracy instead. And once you follow that logic all the way through, one question becomes unavoidable. What happens when the money stops making sense? Find out in $115 billion burn rate the AI bubble just popped.