Transcription
The most valuable software company in the world just made the biggest bet in tech, and Wall Street punished them for it anyway.
On June 2nd, 2026, Microsoft CEO Satia Nadella walked on stage in San Francisco and [music] in one keynote announced an always-on AI agent that works while you sleep. A quantum chip the company says is a thousand times more reliable than its last one. Seven new AI models built entirely in-house and a version of Windows that no longer really looks like Windows. That is not a normal product keynote. It reads more like a declaration of war, and the rest of the industry noticed.
To understand what Microsoft just did, you have to understand what they have been doing quietly for 3 years. While the headlines went to ChatGPT and Elon Musk and the benchmark fights, Microsoft was spending a lot. The company has committed roughly $190 billion in capital expenditures for 2026 alone, more than the annual output of some countries, pouring concrete into data centers, buying chips in bulk, and wiring up the infrastructure for a world that runs on AI. That number is big enough to make investors nervous. Microsoft stock fell nearly 24% in the first half of 2026. Its [music] worst start to a year since the dot-com crash in 2000. The market was telling Microsoft it was spending too fast and getting too far ahead of real demand. Nadella heard all of it. Then he took the Build 2026 stage and announced more.
Anyway, the most disruptive thing they announced has a simple name, Scout. It isn't a chatbot, and it isn't C-Pilot with a fresh coat of paint. Microsoft calls it an autopilot agent, an always-on AI that lives inside your Microsoft 365 setup and doesn't wait for you to ask it anything. It reads your Teams messages, your Outlook threads, and your SharePoint files. It catches calendar conflicts before you spot them. It drafts the status report before you remember it's due. It flags the email chain that's about to slip through the cracks while it's still slipping, not after.
Scout runs with its own digital identity inside your organization. It has its own account, its own permissions, and its own audit trail. It is not a feature you switch on. In Microsoft's own words, it is a virtual co-worker. It is built on OpenClaw, the open-source agent framework that spread across the AI world in early 2026 before its creator got hired away by OpenAI. And Scout wraps that framework in the kind of governance enterprises actually require. When one OpenClaw agent reportedly went rogue inside a researcher's inbox earlier this year, the industry got a preview of what an unsupervised AI agent could do. Scout is Microsoft's answer: autonomous AI with an accountability structure bolted on.
Microsoft expects Scout to help manage what it projects will be 1.3 billion AI agents automating office work by 2028. That is not a typo. It is 1.3 billion agents, each taking a slice of the scheduling, drafting, and coordinating that currently fills the workday for hundreds of millions of people. Scout is the interface sitting in front of all of it. The question every IT manager and CEO is quietly asking is not whether that is impressive. It is whether they are ready to trust it.
Then Microsoft did something that had physicists arguing in real time. They unveiled Majorana 2, the newest version of their topological quantum chip, which the company says is a thousand times more reliable than the last one. Not 10 times, a thousand. The trick is a materials change that sounds almost too simple. Microsoft swapped aluminum for lead as the chip's superconductor, and the resulting qubits survive an average of 20 full seconds versus the microseconds quantum bits usually last before they collapse. Some lasted close to a minute. In quantum computing, that isn't a tuning improvement. It's a different class of result.
Microsoft says Majorana 2 has them on track to build a commercially viable, scalable quantum computer by 2029, which cuts their original timeline in half. If the chip really works the way they describe, the consequences are hard to overstate. Quantum computers don't just run faster; they can solve problems classical machines can't touch at all: modeling complex molecules for drug discovery, cracking the optimization math behind supply chains and financial markets, and breaking the encryption the internet currently runs on. A scalable quantum computer by 2029 wouldn't just give Microsoft a new product line. It would give whoever controls it a kind of computational advantage with no real modern equivalent.
Here's where the physics complicates things. On the day of the announcement, Scientific American reported that independent physicists say Microsoft's topological qubit approach, the whole basis of Majorana 2, does not work and never has. Outside experts called the claims extraordinary and the evidence thin. Microsoft's history here makes that sting. The original Majorana 1 chip was unveiled in 2025 with a lot of fanfare, and the scientific community's doubts never fully cleared. Majorana 2 doesn't erase that. It's asking the field to trust the company one more time on a timeline that's now half as long with stakes that are twice as high. The number that will actually settle this won't come from a press conference. It'll show up in peer-reviewed journals, and people will be watching for it.
Under the two headline announcements, Microsoft slipped in something the industry has been waiting for with a mix of anticipation and dread: seven new AI models built entirely in-house by Microsoft AI. Not licensed from OpenAI, not co-developed with a partner, their own. The lineup covers reasoning, coding, image generation, transcription, and voice. MAI-Thinking 1 is Microsoft's first dedicated reasoning model. MAI-Code1 Flash is an agentic coding model built for GitHub Copilot. Then there's MAI-Image 2.5, MAI-Transcribe 1.5, and MAI-Voice 2. Seven models shipped at once, announced next to Scout and Majorana 2, almost like a footnote. They aren't a footnote.
Right now, 45% of Microsoft's $625 billion contracted backlog is tied to one customer: OpenAI. That concentration has spooked Wall Street in a way no amount of reassurance has fixed. Every Microsoft model that's good enough, fast enough, and cheap enough to run inside a Microsoft product without calling OpenAI's servers makes the company a little less dependent. The MAI models aren't frontier models the way GPT-5 is, but they don't have to be. They just have to be good enough for the everyday jobs Microsoft products do at a price that doesn't leave nearly half a trillion dollars of future revenue riding on a partner that's out signing deals with Amazon and Oracle at the same time.
All of it comes back to the operating system. Microsoft's Build 2026 message for Windows wasn't about a new look or a faster File Explorer. It was about change at the architecture level. Windows is becoming a platform for AI agents with new runtimes, new developer APIs, and new on-device AI frameworks built not for people clicking menus, but for software acting on their behalf while they do something else. A small local model on the device handles latency-sensitive work like intent detection and basic file sorting, while heavier requests go to Azure. The device and the cloud work as one continuous surface with AI agents living in both. That's what Microsoft means when it says Windows is no longer just a desktop operating system.
The next release of Windows 11, expected in fall 2026, will ship with an on-device AI runtime codenamed C-Pilot Engine. After that comes Windows 12, where Microsoft has signaled that AI agents will be first-class citizens with native hardware acceleration. The Windows most of us grew up with, the one that politely waited for you to double-click something, is being retired. Whatever replaces it isn't going to wait around like that.
Here's the uncomfortable part underneath all of this. Microsoft's AI segment has hit a $37 billion annualized run rate, growing 123% year-over-year. Azure's AI revenue is accelerating. By almost every technical measure, the business is working. And yet, of Microsoft's 450 million paid Microsoft 365 seats, 450 million workplaces that already pay for Copilot access, only somewhere between 3% and 4.4% of users touch it daily. The infrastructure is built, the data centers are running, the models work, and the humans in front of the screens still aren't convinced.
That's the problem Scout is meant to solve, and not by waiting for people to change their habits. It removes the need to form a habit at all. You don't have to remember to use Scout because it's already there when you open your laptop, reading and working before you've asked it for anything. Whether that sounds like freedom or surveillance depends on who you ask. The executives who sign the contracts see a productivity jump. The employees whose email Scout has already read before they finish their coffee are asking a very different question.
Microsoft has placed a $190 billion bet that the fix for slow AI adoption isn't better marketing. It's better technology shipped deep enough into the operating environment that opting out takes more effort than opting in. That's a bet with huge upside and real risk. Nadella made it in San Francisco in June 2026 with the confidence of someone who has been right about this industry before and who knows that being right early can look for an uncomfortably long stretch almost exactly like being wrong. The companies that underestimated Microsoft in 2012 lost a decade. The ones that dismissed the Azure bet in 2015 missed the cloud era.
Whether 2026 gets remembered as the year Microsoft laid the foundation for the next 20 years of computing or the year it overextended in the dark won't be settled on stage. It'll be settled slowly in quarterly earnings, in boardroom AI adoption reports, in whether Scout becomes something people can't work without or something they never notice, in whether Majorana 2 survives peer review, and in whether $190 billion earns back a return that justifies the faith. The tech market just got blown up. Whether what grows back belongs to Microsoft is still being written.
Subscribe and like the video. There is a lot more coming.