Transcription
Meta just admitted something that every retail investor in big tech needs to hear right now. On July 1st, Meta announced a brand new business unit called MetaMP compute. The pitch: they're going to rent out their excess AI computing power to outside companies. Guess what? Wall Street cheered. The stock jumped nearly 9% in a single day after the news. Here's what Wall Street missed, though. When a company that just guided to $145 billion in capex for one single year suddenly announces it needs to rent out its spare computing capacity, well, that is not a new revenue line. That's a confession. By the end of this video, you're going to understand exactly what is really happening inside the biggest capex spending cycle in the history of corporate America. You're going to see why the 600 billion that big tech is deploying this year is not just a bet on AI. It's a structural play to lock every competitor out of the market by owning the physical infrastructure that runs the digital economy. And you're going to understand what it means for your portfolio because the companies you thought were software businesses are now something else entirely. All right. Scott Galloway, good guy, professor, fellow podcaster, one of the sharpest bare voices on the AI trade right now, has been telling his audience that a productivity reckoning is coming. His argument is very straightforward and it's not wrong. And five of the biggest tech companies in America are spending more on capital infrastructure this year than the entire global oil and gas production sector combined. More than four times over. And unlike oil, AI has yet to produce a clear, measurable economywide productivity dividend that justifies this huge bill.
Let me give you the numbers. Amazon is projecting $200 billion in capex for 2026. Alphabet is guiding to 175 to $185 billion. Microsoft is tracking towards somewhere in $120 billion range or maybe even more. Meta has raised its fullear guidance to as high as $145 billion. And that's up from just 72 billion just last year. Imagine me saying just 72 billion, but 145 billion this year. Okay. Oracle is around 50 billion combined. The big five are in pace for somewhere between 635 billion to 690 billion in total capex for this calendar year alone. You probably have heard that number getting kicked around all over the place. The mainstream narrative calls this a strategic bet on the AI future. Galloway Prof calls it an overestimate o overinvestment cycle for its reckoning. Both framings are about strategy. Neither is telling you what the corporate accounting actually says. And that's where it gets very interesting.
I want to stop here for a second because what I'm about to show you is not in the analyst reports. It's not in the earnings call summaries. It's in the cash flow statements and the balance sheets. And when you see it laid out clearly, the narrative changes completely. Before I get into this stuff further, if you like this type of content, please click like and don't forget to subscribe. It's important to be in the know and this is really how you do it.
Okay, here's the first thing that you need to understand about this spending cycle. Roughly 75% of it, approximately 400, uh, $450 billion out of the 600 billion I talked about, is going directly into physical tangible assets: custom GPU fleets, networking, server blades, data center real estate. This isn't research. This is not software development. This is the acquisition of hard assets at massive scale. Think about what that means for the moment: the companies that you've been sold as software businesses, asset light, high margin, infinitely scalable, are quietly converting themselves into something much more like industrial infrastructure companies. They're buying physical computing infrastructure the way the railroad companies bought tracks in the 1880s. And just like the railroads, the goal is not just productivity. The goal is to build a moat so wide that no startup, no competitor, no sovereign government can close the gap without a multi-hundred-billion-dollar commitment of their own.
And here's the shadow data that you know I love to talk about. These companies are not covering this spending entirely out of organic free cash flow while simultaneously maintaining their historic share buyback programs. To bridge the gap, the biggest names in tech have quietly abandoned their cash-rich, zero-debt identity and launched one of the largest corporate bond issuance waves in market history. Morgan Stanley estimates that Hyperscalers will issue a collective 400 billion in bonds in 26 alone. That's up from 165 billion last year. Meta issued 25 billion in corporate bonds. Oracle issued 25 billion and Alphabet raised $20 billion worth, including the first Century Bond ever issued by a tech company. Remember, Century Bond matures in 100 years. Wow. I hope I'm around for that maturity. They are leveraging themselves at current 4% macro yields to ensure that they don't lose one inch of ground in the infrastructure race. The balance sheet that made these names feel safe in your portfolio is not what it was.
And then came July 1st. MetaMP compute. Bloomberg broke the story. Meta is building a merchant cloud unit led by infrastructure chief Santosh Jonard to rent out its excess AI computing capacity and spare GPU racks to outside enterprise clients and third party developers. Wow, that was a mouthful. The market reacted by sending Meta shares up 8.8%, 8% to 612 and change that day on nearly triple the daily average volume. Obviously, the market was paying attention. Wall Street read this as a new revenue line, a smart pivot, a sign that Meta is thinking like an infrastructure business. And they're not wrong. That's exactly what Meta is becoming. But read the actual acknowledgement from Zuckerberg carefully. He explicitly confirmed that selling raw GPU cycles was a strategic contingency, a necessity on the table of Meta over built. Let me say that again. The CEO of a company guiding to $145 billion in capex for a single year, a company that just issued $25 billion in corporate bonds to fund that spending, told you he doesn't know if he overbuilt. And the mechanism he's deploying to hedge that risk is to dump spare hardware capacity on the market before the depreciation clock eats his advertising margins. That's not a new business. That's a balance sheet defense strategy. When you see it that way, the 9% pop in Meta Stock on that announcement is not a celebration. It's the market pricing the wrong thing.
Here's how I think about this from an institutional perspective. I've watched capital spending cycles for a long time now. The pattern is pretty much always the same. The leaders overbuilt, then they find a way to monetize the overbuilt. The monetization gets celebrated as innovation. And somewhere in the middle of that celebration, nobody notices that the original spending thesis never fully delivered.
Here's what this means for your portfolio. And I want to be precise about this because it's not simple. The first implication is the K-shaped economy inside tech itself. If owning the physical computing infrastructure becomes the structural moat, then the entire software as a service model that defined the last 15 years of equity outperformance is now in structural question. Small and midcap software companies that depend on renting compute from these same hyperscalers are now renting from their competitors. The cost of compute is being deliberately set by entities with every incentive to keep it high enough to price out challengers. Traditional software subscription growth stories are being repriced by the market in real time. In case you haven't noticed, the repricing is unfortunately probably not finished.
The second implication is the depreciation wave. Custom GPU hardware has a useful life of roughly 3 to 5 years, and that's being fair. The massive spending that began in earnest in 2024 is going to start generating serious depreciation charges across 26 and 27. Meta is already trying to get ahead of it with MetaMP compute. But $145 billion in a single year of hardware spending does not quietly depreciate away. It shows up in the income statement. And when it does, the earnings multiples that justify today's valuations get tested in a way the market isn't fully prepared for at the moment.
Now, the third implication is this. For anyone holding traditional software names alongside these hyperscalers, and we all probably have some of these, the valuation premium that SaaS, that's the software as a service companies commanded built on asset light models and high recurring revenue margins, is being arbitraged away by companies with $600 billion in annual infrastructure spend that are now entering the cloud services markets as vendors. MetaMP compute is going to compete with niche cloud providers. Amazon AWS already competes with everyone. The playing field is not level. It's not designed to be.
What should you actually watch? Well, let's talk about three things to watch. First, the depreciation line item in big tech's quarterly earnings starting next quarter. If it accelerates faster than revenue growth, the free cash flow story gets complicated real fast. Second, MetaMP compute's pricing when it officially launches. The rate at which they price excess GPU capacity will tell you exactly how desperate the overbuild situation might actually be. And third, watch the corporate bond market. $400 billion in tech bond issuance in a single year is an experiment. If long-term yields rise on the back of fiscal expansion and tech companies need to refinance at higher rates, well, the leverage math changes entirely. You might have seen in the press just this morning, Amazon announced a huge bond offering. Where do you think that's going? Well, you now know.
So, your truth bomb for today is this. Big tech is not overspending on AI because they believe in the future. They're overspending because whoever owns the physical computing infrastructure owns the toll road. And when Meta launches MetaMP compute to rent out its spare GPU racks, it's not innovating, it's hedging an overbuild that they're not allowed to even admit. Join me every day for Wall Street Truth Bombs, where I drop them right here before the market figures them out.