OpenAI's Money Problem Is WAY Worse Than You Think

OpenAI

Across the shows: 11 speakers in other episodes, 3 bullish and 9 bearish calls, past 30 days

• Jeff Snider · Atlas FinancialBearish

  OpenAI's two competing revenue figures make the AI investment case look unreliable; skepticism is warranted.

  The FT reported OpenAI told investors annualized revenue was approaching $50B at the end of September, about $20B below a widely circulated $70B estimate from its backers. The higher figure was not an official company forecast and included cloud platform sales OpenAI itself excludes. Snider argues that once the AI investment case depends on which accounting method gives the biggest number, skepticism is warranted, and markets are starting to show it.

  “But that distinction is exactly the point. Once the AI investment case depends on which accounting method produces the biggest number, skepticism at the very least is warranted.”at 0:54

Meta Platforms

Across the shows: 9 speakers in other episodes, 13 bullish and 3 bearish calls, past 30 days

• Steve Van Metre · Steven Van Metre FinancialBearish

  If Meta cannot raise AI capital, its profitable core free cash flow will go to debt service.

  Steve argues that if companies like Meta cannot get the capital or generate revenue from mass-market AI, free cash flow from the profitable parts of the business will have to go into debt service. They cannot simply give up on AI after building the data centers and software, so other resources will be diverted, and sacrifices will follow.

  “all of a sudden, free cash flow will take meta from their profitable side of their business, has to go into debt service, but they still can't give up.”at 18:42

Oracle

Across the shows: 6 speakers in other episodes, 0 bullish and 5 bearish calls, past 30 days

• Steve Van Metre · Steven Van Metre FinancialBearish

  Oracle's credit default swaps show the market doubts it can pay its AI-related debt.

  Steve points to the triple C credit space and credit default swaps as signs of stress, singling out Oracle. He says the CDS market is effectively saying we are not sure how you will pay on this debt. He uses this to ask how a bubble bursts and how the credit cycle plays out for AI borrowers.

  “you look at Oracle, I mean, which is just an absolute, like, flaming trash truck of the market saying, look. We're not sure how you're gonna pay on this debt.”at 19:28

AI infrastructure & data centers

• Steve Van Metre · Steven Van Metre Financial

  The AI buildout has too much capacity and too little demand, echoing the .com bubble.

  Steve says companies keep building data centers while claiming infinite demand, but users find they do not need the tools after the first use. He argues the spending is funded by borrowed money and that the situation looks just like the .com bubble. He says there is too much capacity, not enough demand, and a lot of hype.

  “again, too much capacity, not enough demand and a lot of hype, and everybody's just so invested in it.”at 3:12

• Jeff Snider · Atlas Financial

  AI buildout will produce overcapacity, with much of the capacity never used, as in past railroad bubbles.

  Snider argues that companies are building capacity on the assumption they need it, but they may only need part of it. Redundant builds are a feature of past bubbles, and the railroad bubble saw tons of track never used. Because there will be winners and losers, some AI capacity will never be used, so overcapacity is almost by definition inevitable.

  “They build tons of track that never got used. And so there may be a lot of this AI capacity that gets built that will never get used.”at 4:50

AI bubble & tech valuations

• Jeff Snider · Atlas Financial

  The AI boom mirrors the dot-com bubble and is debt-fueled; he hopes it pops sooner to limit damage.

  Snider argues the AI era shares dot-com traits: money-losing companies making enormous bets on speculative ideas, creative accounting to keep Wall Street buying, and now debt funding. He says the stock market prices a distant AI-driven future while the economy people live in today is weak, which adds risk. Because the bubble is piling on astronomical numbers, he thinks a sooner pop would do less damage than a later one.

  “And I've said in a video recently, I kinda hope it pops sooner rather than later. Because if we get this thing pop sooner, then the damage tends to be a lot less.”at 4:03

• Steve Van Metre · Steven Van Metre Financial

  The bubble is unlikely to pop soon because investors have no alternative to long tech and hyperscalers.

  Steve argues the bubble does not look likely to pop soon because investors seeking a large return have only one answer, going long tech stocks and hyperscalers. Past bubbles had an immediate impact only once the broad economy weakened, and this one is staked on AI while the labor market and consumer are fragile. A later pop would therefore hit at a dangerous time.

  “it doesn't look like that's gonna happen this time. ... The really only answer is you've gotta go long tech stocks and hyperscalers because there's really your only hope right now.”at 6:09

Credit markets

• Jeff Snider · Atlas Financial

  AI-linked debt of about $1T could reach $3-4T in a year, multiplying downside if the bubble reverses.

  Snider estimates roughly $1T of debt is riding on AI today and expects that to be about $3-4T a year from now, and astronomical figures two years out if the bubble lasts. Pension funds and institutional investors are building screens to limit AI loans, because nearly every incoming leveraged, syndicated or CLO deal is AI-related. The financial system is so dependent on AI that the downside grows exponentially if the trade reverses.

  “you've got, what, a trillion dollars in debt that's really, if you add everything up, about a trillion dollars in debt that's riding on AI. A year from now, that's not going to be $2,000,000,000,000 it's going to be 3 or 4.”at 9:56

• Jeff Snider · Atlas Financial

  Spreads on AI-linked junk bonds, especially CCC, are rising as lenders become more discerning.

  Snider points to spreads on bonds tied to AI builds creeping up, especially in the junk segment, and to CCC junk bond indexes that keep rising. He reads this as the market not yet scared but increasingly discerning about where it lends. As conditions tighten, the payoff for AI companies that have bet everything gets harder to finance, and the market's pullback makes the payoff less likely.

  “We see some of the spreads on some of these bonds related to AI builds and whatnot creeping up, especially in the junk parts of the debt. The CCCs, CCC junk bond indexes, that continues to rise.”at 17:15

Enterprise AI adoption & AI software

• Steve Van Metre · Steven Van Metre Financial

  AI demand looks weak: once users build a tool, they stop needing the AI, so repeat usage is limited.

  Steve asks who will use all the data centers being built, noting that when people try AI they get an answer and do not need to use it again. From his own use, once a program is built, you do not need AI to change it, so the expected restructuring of businesses and households does not show up. He reads this as too much capacity against weak demand.

  “And the problem is when you do use it and you get the answer, you don't need to use it again.”at 1:48

US jobs & labor market

• Steve Van Metre · Steven Van Metre Financial

  Payrolls have gone nowhere since mid-2024 and could turn negative in the next month or two.

  Steve says the labor market has gone nowhere since mid-2024 and payrolls could be negative in the next month or two. Hourly earnings keep declining, and consumers say conditions are getting worse. With high gas, consumer and food prices, consumers cannot sustain spending, so if the AI bubble pops later, job losses would hit a weak economy.

  “[Since] mid twenty twenty four, it's gone nowhere and looking like payrolls could be negative in the next month or two. Hourly earnings continue to decline.”at 6:09

AI & jobs

• Jeff Snider · Atlas Financial

  AI spending lifts GDP through capex but is not creating hiring or broader economic activity.

  Snider says AI spending looks good on GDP because it counts as capital expenditure, but that spending is not leading to a hiring boom, and labor statistics show the opposite. Chips are being made and data centers built, yet he sees no additional jobs or economic activity. He calls the AI boom hollow, so the real economy is not dependent on AI.

  “Yeah, it looks good on GDP because you get capital expenditures, but it's not like those expenses and that spending is leading to a hiring boom.”at 7:49

US consumer

• Steve Van Metre · Steven Van Metre Financial

  Retiree spending depends on the stock market, so the AI-driven rally is holding up the consumer.

  Steve argues that AI spending itself does not contribute much to the economy, but the stock market does, because the wealth boomers create leads them to spend more. He observes that retirees spend freely when the market is up and stay home during corrections, so boomer spending is a huge part of the economy.

  “the AI spend itself isn't contributing to the economy, but the market is because the more wealth these boomers create, the more they spend.”at 11:12

Banks

• Steve Van Metre · Steven Van Metre Financial

  Bank lending is surging on AI, but the concentration in a few hyperscalers is a growing risk.

  Steve says banks would not be lending at this pace without AI, and that lending is concentrated because banks have few other borrowers. He notes that if the hyperscalers cannot repay, their only real option is issuing more stock, so the banks end up with too much concentration. He also links the risk to the yield curve: if long rates rally and the curve flattens, banks may stop lending to AI companies.

  “If it wasn't for AI, they would not be lending at this pace.”at 12:01

US Treasuries & yields

• Jeff Snider · Atlas Financial

  Falling long-end Treasury yields signal weaker growth and risk reassessment, which is bad for credit.

  Snider argues that lower Treasury yields, especially at the long end, mean lower growth and inflation expectations rather than cheaper credit. The financial system is reassessing risk, so lenders become less willing to lend to borrowers perceived as risky. He concludes that falling yields are not a good sign for the credit cycle or credit markets.

  “when treasury yields go down, especially at the longer end, that means lower growth and inflation expectations. And the flip side of that is the financial system is reassessing risk.”at 16:32

US stock market outlook

• Steve Van Metre · Steven Van Metre Financial

  Market gains depend on 8-10 companies, so money will flow out of them the same way it came in.

  Steve says most inflows are going into eight to ten companies, and market-cap-weighted indices keep that concentration in place. Investors have few options beyond those few names. If something goes wrong, especially a credit cycle break, the money that flowed in will flow out along the same path, leaving the market exposed.

  “If all the money is flowing into 10 companies, it's gonna flow out in the exact same path.”at 19:08

Earnings & corporate profits

• Steve Van Metre · Steven Van Metre Financial

  Reported AI-era profits may overstate reality once off-balance-sheet debt is counted.

  Steve says profits were supposed to be rolling in from AI projects completed this year or early next year, but they are not. He suggests that what looks like profit now may not be real once investors learn about off-balance-sheet debt. As a result, the bubble could burst when panic spreads and everyone sells.

  “what we're seeing as profits now may not really be when you find out about all the off balance sheet debt.”at 19:08