AI financing risks are already visible, yet 40% of investors still refuse to see them
A Barclays survey shows only 60% of investors think AI financing will become a problem, but the real troubles of companies like CoreWeave and Cipher in recent weeks already prove the answer should be 100%—the market is still in denial.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Markets are in denial about when—and if—AI becomes profitable
Altman says if you put a gun to any investor's head and asked whether they'd still bet on AI, the answer would almost universally be yes—the train is still moving forward and nobody wants it to stop. But he also notes this is precisely the first stage after a bubble pops: denial. The evidence: some Korean memory chip stocks trade at a 3x P/E ratio. By cyclical stock logic, the market is actually pricing AI supply chains as if earnings have peaked and won't command premium multiples again. Yet publicly everyone speaks in unison about staying bullish.
— Alex AltmanNeoclouds, not cloud incumbents, will crack under AI financing pressure first
Altman compares neoclouds to the subprime mortgage instruments of the 2008 financial crisis—trouble always starts at the periphery, never at the core. These companies have almost no operating cash flow yet are expected to build compute capacity ahead of revenue. CoreWeave's bond selloff wasn't triggered by CoreWeave's own news; it was Cipher, another neocloud, failing to issue convertible debt and getting its stock "nuked" that made investors suddenly connect the two.
— Alex AltmanMarkets mispricing AI financing risk by a factor of nearly two
A Barclays survey of investors shows only about 60% believe AI financing will become a problem in the foreseeable future. But by Altman's logic, what CoreWeave and Cipher have already endured in recent weeks proves the answer should be 100%. That gap itself is the signal: the market is in denial—things are already breaking, but most people haven't admitted it yet. It's not panic, and it's definitely not priced in.
— Alex Altman2026 debt market breaks the interest-rate pattern held for three decades
Barclays has tracked since 1995 the ratio of net issuance to market cap across corporate bonds, government bonds, and equities, finding a stable pattern: rates down, issuance up; rates up, issuance down. In 2026, that relationship has completely broken. Rates remain elevated, yet issuance has hit record highs. The reason: AI is competing against every other industry for the same finite pool of capital. Projected 2026 net issuance in investment-grade and high-yield bonds is $1.1 trillion. That figure is not infinite. AI isn't issuing into a vacuum.
— Alex AltmanAI's promised TAM exceeds total S&P 500 corporate profits by 5x
Altman mentions the "SpaceX TAM" that was widely discussed at the time—$30 trillion, nearly the size of U.S. GDP. By comparison, Microsoft and Apple, the two most successful tech companies in modern history, have generated a combined inflation-adjusted cumulative TAM since 1990 of under $10 trillion. Zoom out further: all S&P 500 companies combined generate about $17 trillion in annual revenue, $11 to $12 trillion in expenses, and $5 to $5.5 trillion in net profit. That's the crux: the money enterprises can actually spend on AI is far smaller than these macro TAM narratives suggest.
— Alex AltmanOpen-weight models have made centralized data centers uneconomical for routine tasks
Data shows open-weight models have risen from about 20% to a peak of 80% market share, while the cost per token has collapsed to one percent of its prior level—one of the fastest price deflations on record. Altman poses a sharper question: if a phone can run billion-parameter models locally to handle everyday tasks like email, why do we still need flagship chips in data centers? He frames this as "going long and short on human creativity"—what might actually be scarce isn't compute, but the ability to create something new with it.
— Alex AltmanPod shop hedging hides AI concentration behind index-level calm
Altman explains why CoreWeave and similar stocks are wildly volatile while the VIX and realized volatility stay calm: pod-shop hedge funds managing assets that have ballooned from $300 billion in 2020 to over $1 trillion today are hedging away factor risk, offsetting it at the index level while concentrating it at the factor level. The deeper problem: nobody can actually model the "AI factor." Early in the year, everyone was shorting software stocks as part of their AI hedge. Now software is near 12-month highs. That factor's beta keeps shifting, and funds can't hedge it.
— Alex AltmanS&P 500 earnings growth will derive 75% from AI ecosystem next year
Altman offers a figure: approximately two-thirds of S&P 500 earnings growth next year will come from semiconductors; when you include the entire AI ecosystem, that proportion rises to about 75%. In other words, the S&P 500 is no longer a diversified index but increasingly a concentrated bet on AI—which explains why on earnings day Google can drop 7.9% and Microsoft can surge 6.4% while the index barely moves.
— Alex AltmanIn their own words · checked verbatim
The consensus AI outlook is that AI is still going to, this train is in motion. We're not stopping this train.
Alex Altman2:08
we're in the denial phase right now because everyone is still extremely bullish.
Alex Altman3:09
around about 60% of investors, we asked them a question, do you think financing is going to be a problem? AI financing is going to be a problem for the foreseeable future. Only 60% said yes.
Alex Altman7:13
2026 has completely shattered what you'd seen in the norms.
Alex Altman10:16
open weight has kind of gone from a 20% market share to anything up to about an 80% market share.
Alex Altman17:27
I call it like a knife fight in a phone booth, you know, for all these funds that have to run long, short portfolios.
Alex Altman27:46
the bottom line is, is you're either long AI or you're short AI, right?
Alex Altman31:52
Figures
| Some Korean memory chip stocks P/E | 3x | 4:10 |
| Investors surveyed who expect AI financing to become a problem | 60% | 7:13 |
| 2026 IG + HY bond market net issuance forecast | $1.1 trillion | 11:19 |
| SpaceX TAM discussed at the time | $30 trillion | 13:22 |
| Microsoft + Apple cumulative TAM since 1990 (inflation-adjusted) | under $10 trillion | 14:24 |
| S&P 500 companies combined annual revenue | $17 trillion | 14:24 |
| Open-weight model market share range | from ~20% to peak 80% | 17:27 |
| Microsoft Azure cloud YoY growth vs. expected 44.8% | 48% | 32:53 |
Glossary
- neoclouds
- Cloud service providers that build and lease AI compute capacity with minimal profit margins.
- pod shop
- Hedge fund structure where independent trading teams within the same fund hedge against each other's factor exposures at the portfolio level.
- frontier sink
- Industries requiring PhD-level expertise that can continuously absorb improvements in model capability.
- AI factor
- A quantitative risk category used by hedge funds to tag "AI-exposed" positions, though its definition and beta keep shifting.
- open weight model
- Model whose weights are publicly downloadable and can run locally, contrasting with closed-source frontier models.
How to listen
Investors and analysts tracking AI infrastructure financing, hedge fund positioning structure, and S&P 500 concentration risk.
Opening pleasantries and mid-episode scheduling asides can be skipped without affecting the core argument.