Pay the Piper
Risk is the sort of word that is easy to discuss upfront but tough to handle when it comes time to pay the piper – Nathan Myhrvold
Data center growth and artificial intelligence have combined to make Microsoft, Google, Nvidia, Amazon, Meta, ASML, Tesla, TSM, Micron, and Apple insanely over-valued. They were given a ten-year run where building “the Cloud” was extraordinarily profitable for this group. It was Magnificent, so much so that it seems almost divinely inspired.
For the past year, pundits have been calling AI a bubble. Detractors have been saying that the “Large Language Models” that are the “brains” of AI are inherently flawed for years. The detractors have been proven correct but not the stock market pundits.
Bond investors have figured out that lending to projects such as Stargate are a recipe for capital losses. The cold feet seems to have started in August 2025. OpenAI, Softbank, and Oracle are going to have difficulty obtaining the $500 billion needed to finish the massive data center in Abilene, Texas. Future data centers in Wisconsin and Michigan look less likely to be completed.
Early returns on AI data centers have been poor. Blue Owl has experienced problems with loans to build a data center in Louisiana with Meta. Their closed-end fund (OBDC) is trading at a 16% discount to net asset values. Plans to combine a retail fund with OBDC fell through.
The funding environment is only getting worse. BlackRock’s TCP Capital private-credit fund disclosed a 19% markdown in net asset values. While not invested in data centers, TCP’s markdown will only make bond investors increasingly wary.
Bubble’s end when the marginal investor is sucked in to mark the top. Bond investors are at that point and I suspect further capital deployed to data centers will need to come from either the major tech companies deploying operating cash flow or by “forcing” this debt into passive retirement accounts at higher yields.
OpenAI
The edge of a black hole, the event horizon, is a boundary that marks the point of no return. Once an object crosses the event horizon, it cannot escape and will be ripped to pieces, atom by atom – Kevin McCarthy
I view OpenAI like I view a black hole; it’s the place where capital disappears. It represents an unlimited opportunity to lose money. OpenAI lost a mere $5 billion in 2024 and is expected to lose $8 billion in 2025. We get a glimpse of the rapid depreciation on Microsoft’s books because Microsoft owns 49% of OpenAI. Tellingly, cap exp has risen 300% since 2020. Capital lease obligations are up 130%, net tangible assets are up over 100% ($200 billion), and depreciation is up over 200%.
OpenAI’s voracious capital needs have turned MSFT into a capital-intensive investment, not the cash-flow machine of yesteryear. It also makes Microsoft riskier. Much riskier when you consider that only 5% of OpenAI’s users pay a subscription for the service. It’s why OpenAI is moving to run ads on the free version. Good luck with that.
AI is already a commodity service. I can easily shift from ChatGPT to Grok to Gemini to Deepseek, etc. There is nothing offered that justifies premium pricing and that means there will never be a payback of capital.
OpenAI has been hinting about a government investment for months. When high-level AI salesmen do the media rounds, they emphasize the need to develop AI faster than the Chinese. That’s a big change in narrative away from the untold riches AI offers. Unless the Chinese have learned to make LLM’s work, I’m fine with letting them win.
Politics
Never believe anything in politics until it has been officially denied – Otto von Bismark
The Trump Administration has tied itself to AI with the announcements about foreign investments from Softbank and the like when taking office. They need AI to remain viable through the November elections. But will they bailout OpenAI?
I don’t believe a populist movement can bailout an industry that promises to replace human capital with robots. In addition, there is a growing backlash against data centers – a return of NIMBY, or “not in my backyard.” Neighbors of data centers tend to pay more for electricity and water when capacity is built close by.
Besides, a bailout of OpenAI et al., is a bailout of the most valuable companies on Wall Street. That dog won’t hunt – especially in an election year.
Instead, I expect the Mag 7 and their satellites to be pushed aggressively to keep the liquidity taps open to complete projects already started. Companies like Softbank will be pushed to write checks on capital already promised, despite it being a clear loser. Announcements like Meta’s $6 billion deal to purchase Corning photonic cables maintain the narrative that data centers are still viable. Wash, rinse, and repeat through the year.
Is it a coincidence that Amazon and Microsoft have been cutting staff? Amazon is expected to announce 30,000 more job eliminations this week. The cashflow needs of AI are a significant drain on all companies involved. Personally, I view this as the Swan Song of the Magnificent Seven stock and other high-flyers.
Index Funds
If you got the devil to pay, he’ll want extra red-eye on his biscuit. He always do – Randy Thornhorn
The last 12 years have been extraordinarily profitable for technology companies. First, they benefited from buybacks, then using debt to fuel buybacks, then the buildout of the “Cloud” which paved the way for AI.
Success begets success and index funds rapidly purchased shares as these companies became bigger pieces of stock indexes. Today, the total value of the US stock market is $69 trillion. The combination of Google (Alphabet), Microsoft, Nvidia, Meta, Amazon, Tesla, Apple, Broadcom, and Oracle equals 34% of the total value of US stocks or $23.5 trillion.
Given that US Index funds equal around $12.5 trillion and ETF equity around $5.5 trillion, it indicates that index products represent around 26% of the US stock market. Applied to the above names, it suggests around $6 trillion of passive funds are invested in these tech names.
Furthermore, given an expected 2026 GDP of $32 trillion, the above names represent 73% of US GDP. History has never been kind to such concentrations of wealth and in this case, I expect the realization that AI is a dud to blow an enormous hole in the wealth of the US.
The stock market is trapped in these 9 names. If earnings don’t materialize, there is no getting out. The concentration in indexes suggests that the failure of AI would decimate US household wealth. AI won’t be allowed to fail, at least not yet. As investors, we can’t live with it and we can’t live without it.
Conclusion
The empty vessel makes the loudest sound – Plato
We’re in limbo and it will probably remain this way for much of the year, barring an out-of-the-blue event such as Chinese civil war – material for my next piece. Informed investors know AI is complete nonsense but that doesn’t matter.
If you try to bet against it, you’re going to get destroyed until we reach the event horizon. I have no idea when it will happen; nobody does. Something is going to happen that breaks this capital destruction narrative and when it happens, the Magnificent 7 plus 2 will leave retirement funds empty.
AI is like a high school party that gets so rowdy, it will ultimately result in the police arresting everyone in attendance. Sitting at home, doing your homework while the party happens, seems soul-crushing until Monday morning. I prefer to spend my time understanding real financial opportunity instead of chasing a trend off a cliff.
If you’re interested in learning more, visit us at https://geovestadvisors.com/ and contact Paul Hurley.
Philip M. Byrne, CFA