Dead Money
We recorded 50 claims from this post. 30 were checked against evidence. 20 can’t be checked. 43 more are held. Read the post at its source.
Claims from this post (50)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: UnresolvedSaid
Causal claim
“In other words, the talking point that NVIDIA’s GPU sales are proof of actual demand for AI services or, indeed, that hyperscaler growth is a result of all those capital expenditures is a complete lie.”
Our reading: The talking point that NVIDIA's GPU sales prove actual demand for AI services, or that hyperscaler growth is a result of their capital expenditures, is a complete lie.
No horizon Evidence as of
Limitation: Management attribution is relevant but is not an independent estimate of causal effects.
Assessment: UnresolvedSaid
Observation
“I estimate that there’s around $22 billion of global, non-Anthropic/OpenAI compute demand, and an indeterminately-large chunk of that is coming from AI startups that can only afford to pay for the compute as long as venture capital continues to fund them…”
Our reading: By the author's estimate there is around $22 billion of global compute demand outside Anthropic and OpenAI, and an indeterminately large chunk of it comes from AI startups that can only pay as long as venture capital keeps funding them.
No horizon Evidence as of
Limitation: Annualization is an interpretation, not an explicit commitment by the speaker.
Assessment: Can’t be checkedSaid
Observation
“In fact, I’d argue that the last year and a half’s worth of AI capex is fundamentally speculative, because there has never been any proof — even with Anthropic and OpenAI’s compute spend — that spending a trillion or more dollars on GPUs and data centers would ever pay off.”
Our reading: The last 18 months of AI capex is fundamentally speculative because there was never proof that trillion-dollar GPU and data center spending would pay off.
No horizon
Qualifier: last year and a half
Assessment: UnresolvedSaid
Conditional forecast
“And now the rest of the world has caught up to what I’ve been saying since October 2025 — that hyperscalers need at least $2 trillion in annual AI revenue by 2030 or they’ve wasted their capex.”
Our reading: Unless hyperscalers reach at least $2 trillion in annual AI revenue by 2030, their AI capex will have been wasted.
Not yet due Evidence as of
Only if: if hyperscalers fail to reach $2 trillion annual AI revenue by 2030
Assessment: UnresolvedSaid
Observation
“As a result, I think it’s fair to say at least 50% of Oracle’s AI revenues are from OpenAI.”
Our reading: The author thinks it is fair to say that at least 50% of Oracle's AI revenues come from OpenAI.
No horizon Evidence as of
Limitation: The SEC quarterly filing could not be read.
Assessment: UnresolvedSaid
Observation
“This puts us at $183 billion in annual AI revenue for Google, Microsoft, Amazon, Oracle and SpaceX, with $118.2 billion, or at least 64.3%, coming from Anthropic and OpenAI.”
Our reading: Google, Microsoft, Amazon, Oracle and SpaceX have about $183 billion in annual AI revenue, at least 64.3% ($118.2 billion) from Anthropic and OpenAI.
No horizon Evidence as of
Limitation: The cited evidence is context only.
Assessment: Can’t be checkedSaid
Observation
“In truth, I think there’s a world where AI revenues are much lower ($150 billion or less) and Anthropic and OpenAI’s share is more like 70% to 80%, especially in the case of Oracle.”
Our reading: Hyperscaler AI revenues may actually be $150 billion or less, with Anthropic and OpenAI's share around 70–80%.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Goldman aggressively clears it throat by adding that “revenues are growing quickly and revenue backlogs are sizable”, but that’s far from a foregone conclusion considering (as I’ve mentioned) the fact that hyperscalers appear to be warehousing hundreds of billions of dollars of GPUs”
Our reading: Goldman's statement that AI revenues are growing quickly and revenue backlogs are sizable is far from a foregone conclusion, given that hyperscalers appear to be warehousing hundreds of billions of dollars of GPUs.
No horizon
Qualifier: hyperscalers appear to be warehousing GPUs
Assessment: Can’t be checkedSaid
Observation
“Considering that The Information had it at around $900 million a month beforehand, I think it’s likely that its revenues sit at around $300 million to $400 million.”
Our reading: Cognition's actual revenues are likely around $300–400 million despite its claimed $1 billion annualized figure.
No horizon
Assessment: UnresolvedSaid
Observation
“We aren’t even in the same universe of scale that would be necessary to justify the capital expenditures made by hyperscalers.”
Our reading: AI industry revenue is nowhere near the scale needed to justify hyperscaler capex.
No horizon Evidence as of
Limitation: The qualitative size threshold is unspecified.
Assessment: Can’t be checkedSaid
Forecast
“And things are only going to get worse from here.”
Our reading: Things are only going to get worse from here.
No horizon
Assessment: UnresolvedSaid
Forecast
“That’s an astonishing number, and one that’s going to run headfirst into the growing price of US Treasuries”
Our reading: The planned AI data center debt will run into rising US Treasury yields, raising borrowing costs.
No deadline stated Evidence as of
Limitation: The forecast is qualitative and has no explicit deadline.
Assessment: UnresolvedSaid
Observation
“On a strictly cash basis, this means that Oracle’s debt would, if issued today, cost it another $6.89 billion in interest.”
Our reading: On a strictly cash basis, Oracle's 2025 bonds discussed in the post would, if issued today, cost it another $6.89 billion in interest.
No horizon Evidence as of
Limitation: Nearby dates and different Oracle bonds cannot settle this date-specific calculation.
Assessment: Can’t be checkedSaid
Conditional forecast
“As Oracle sits exactly one rung above junk — and its debt trades at junk rates (meaning that the markets buy and sell it as if the yields were junk (an average of 7.8%) — it would likely see its debt priced at around 25-50bps more than what we’ve seen here.”
Our reading: Because Oracle sits one rung above junk and its debt trades at junk rates, it would likely see new debt priced around 25–50bps above the flat repricing estimate.
No horizon
Only if: if Oracle issued this debt today
Assessment: Can’t be checkedSaid
Conditional forecast
“And if Oracle raises more debt, it runs the risk that two ratings agencies could downgrade it to “junk,” immediately forcing investment funds and indices that cannot hold junk debt to dump it, turning it into a “fallen angel” (as I covered a few months ago).”
Our reading: If Oracle raises more debt, it could be downgraded to junk by two agencies, forcing funds to dump its bonds.
No horizon
Only if: if Oracle raises more debt
Assessment: UnresolvedSaid
Observation
“As you can see, repricing CoreWeave at today’s rates would increase its costs by 35.8%, adding $1.2 billion to the lifetime cost of the bonds for a company that already pays $640 million a quarter in interest.”
Our reading: Repricing CoreWeave's bonds at today's rates would raise their cost 35.8%, adding $1.2 billion.
No horizon Evidence as of
Limitation: No independent repricing total was computed from incomplete inputs.
Assessment: UnresolvedSaid
Observation
“Make no mistake, CoreWeave needs to raise more debt to build its data centers.”
Our reading: CoreWeave must raise more debt to build its data centers.
No horizon Evidence as of
Limitation: Reading the claim simply as an expectation of more borrowing would produce a different assessment.
Assessment: Can’t be checkedSaid
Causal claim
“And if the price of their debt is crashing — and the expected yield on the future debt is skyrocketing as a result — then their problems are everyone’s problems.”
Our reading: If CoreWeave's and Oracle's debt prices are crashing and expected yields on their future debt are skyrocketing as a result, their problems are everyone's problems.
No horizon
Only if: if their debt prices are crashing and expected future yields skyrocketing
Assessment: SupportedSaid
Causal claim
“So, as more AI data center debt gets issued, said debt becomes more expensive, because the larger the amount of debt any one thing takes up, the more competition it faces, and the more risk an investor carries by holding it.”
Our reading: Issuing more AI data center debt makes that debt more expensive.
No horizon Evidence as of
Limitation: The causal attribution is based on comparative market analysis, not an experiment.
Assessment: UnresolvedSaid
Causal claim
“The other problem is, as I discussed last week, that these things are simply not getting built, either because the power isn’t there or construction is taking longer than expected, which is in turn putting pressure on effectively any data center-related debt”
Our reading: AI data centers are simply not getting built, because the power isn't there or construction is taking longer than expected, and this is putting pressure on effectively any data center-related debt.
No horizon Evidence as of
Limitation: The most relevant delay-to-pricing evidence is a secondary summary.
Assessment: ContradictedSaid
Observation
“I also want to be clear about something: every single AI data center SPV is funded by customer payments which will only arrive if the data center actually gets completed.”
Our reading: Every AI data center SPV is funded by customer payments that only arrive if the data center is completed.
No horizon Evidence as of
Limitation: The evidence consists of rating-agency descriptions; executed leases were not retrieved.
Assessment: UnresolvedSaid
Observation
“As I went into in last week’s premium (and per my Bastard data center model), a 100MW data center costs around $4 billion to $5 billion, with gross margins of 28% at $17.7 million a megawatt…with negative gross margins below $12.10 a megawatt.”
Our reading: Per the author's model, a 100MW data center costs around $4–5 billion, with 28% gross margins at $17.7 million a megawatt and negative gross margins below '$12.10 a megawatt' (as written), but only with a customer the entire time and no debt.
No horizon Evidence as of
Limitation: The complete model and arithmetic are unavailable in the packet.
Assessment: UnresolvedSaid
Observation
“This means that the underlying customer base for effectively every AI data center is either a hyperscaler or somebody that can, by definition, not actually afford to pay for their compute without somebody else giving them the money, with “somebody” often meaning “a hyperscaler.””
Our reading: Nearly every AI data center's customers are hyperscalers or companies that cannot pay for compute without outside money.
No horizon Evidence as of
Limitation: There is no representative facility-level customer census or denominator for April–June 2026.
Assessment: Can’t be checkedSaid
Forecast
“Said customers are brittle, unprofitable and unsustainable, with the very real risk that they simply won’t exist by the time construction is complete.”
Our reading: AI data center customers are brittle and unprofitable, and may not exist by the time construction is complete.
No horizon
Assessment: UnresolvedSaid
Observation
“To be explicit, the B300 is NVIDIA’s latest-generation GPU, and its long-term rental prices are less than half of SiliconData’s hyperscale pricing for the years-old H100.”
Our reading: Long-term B300 rental prices are less than half of SiliconData's hyperscale pricing for the older H100.
No horizon Evidence as of
Limitation: The contract compilation is secondary evidence.
Assessment: Can’t be checkedSaid
Forecast
“Even then, at the end of the five year period, we’ll theoretically have multiple new generations of NVIDIA GPUs (Vera Rubin and Feynman), which will further suppress the ongoing rates that your data center earns”
Our reading: At the end of the five-year period there will theoretically be multiple new NVIDIA GPU generations (Vera Rubin and Feynman), which will further suppress the ongoing rates a data center earns.
No horizon
Qualifier: theoretically
Assessment: SupportedSaid
Observation
“AI data centers are extremely expensive.”
Our reading: AI data centers are extremely expensive.
No horizon Evidence as of
Limitation: The adjective is qualitative; support applies to the stated ordinary reading.
Assessment: UnresolvedSaid
Observation
“AI data center debt is now extremely expensive.”
Our reading: AI data center debt is now extremely expensive.
No horizon Evidence as of
Limitation: Individual borrowing costs cannot establish the sector-wide claim.
Assessment: UnresolvedSaid
Observation
“AI data centers take years to build.”
Our reading: AI data centers take years to build.
No horizon Evidence as of
Limitation: Projected milestones do not establish completed construction durations.
Assessment: UnresolvedSaid
Observation
“AI data center customers are brittle, unprofitable and dependent on near-perpetual funding.”
Our reading: AI data center customers are brittle, unprofitable and dependent on near-perpetual funding.
No horizon Evidence as of
Limitation: No representative customer population or weighting method is established.
Assessment: Can’t be checkedSaid
Rhetoric
“AI data centers — and their associated costs and debt — are priced for a level of perfection that no other industry has ever rivaled.”
Our reading: AI data centers and their associated costs and debt are priced for a level of perfection that no other industry has ever rivaled.
No horizon
Assessment: UnresolvedSaid
Forecast
“Everybody is either egregiously ignorant or hopelessly optimistic in a way that will make this situation so much worse when it collapses.”
Our reading: The AI data center situation will collapse, made worse by widespread ignorance or optimism.
No deadline stated Evidence as of
Limitation: Collapse and the claimed amplification require evaluator interpretation.
Assessment: UnresolvedSaid
Observation
“I estimate that since 2023, there’s been around $800 billion in global venture capital investment in AI companies, with at least $266 billion of that going to Anthropic and OpenAI.”
Our reading: About $800 billion in VC has gone into AI companies since 2023, at least $266 billion to Anthropic and OpenAI.
No horizon Evidence as of
Limitation: Funding-round totals and cash receipts are different measures.
Assessment: Can’t be checkedSaid
Forecast
“This means that before the bubble bursts, these companies will continue to sap the venture capital world of billions more dollars, all with little chance of an acquisition and a near-zero chance of an IPO considering their ugly economics.”
Our reading: Before the AI bubble bursts, AI startups will absorb billions more in VC with little chance of acquisition and near-zero chance of IPO.
No horizon
Qualifier: before the bubble bursts
Assessment: Can’t be checkedSaid
Observation
“Though some of them talk a big game about building or training their own models, doing so is enormously expensive with little chance of a payoff, especially given the massive advantage in compute, capital and talent held by the labs.”
Our reading: AI startups training their own models is enormously expensive with little chance of payoff given the labs' advantages.
No horizon
Assessment: UnresolvedSaid
Forecast
“Nobody wants to answer these questions, because the answer is pretty simple: one day, Cognition, like many AI startups, simply runs out of money and dies, or becomes a much, much smaller company.”
Our reading: Cognition, like many AI startups, will eventually run out of money and die or become much smaller.
No deadline stated Evidence as of
Limitation: Company announcements are not a complete account of financial health.
Assessment: UnresolvedSaid
Observation
“And outside of those two companies, I estimate there’s roughly $22 billion of demand for AI compute.”
Our reading: AI compute demand outside Anthropic and OpenAI is roughly $22 billion.
No horizon Evidence as of
Limitation: The quarter and annual spending pace are explicit evaluator choices, not dates or units supplied by the speaker.
Assessment: Can’t be checkedSaid
Conditional forecast
“Every single day — even on the weekends — someone asks me either how or when all of this breaks, and my answer is simple: when the money runs out.”
Our reading: The AI bubble will break when the money runs out.
No horizon
Only if: when the money runs out
Assessment: UnresolvedSaid
Forecast
“Eventually, AI data center debt is going to become untenable for those raising it, because 11%+ rates on already-meager margins makes the maths a little impossible.”
Our reading: AI data center debt will eventually become untenable because 11%+ rates on thin margins don't work.
No deadline stated Evidence as of
Limitation: Accessible evidence is concentrated on CoreWeave rather than a representative borrower population.
Assessment: Can’t be checkedSaid
Conditional forecast
“Once this happens, there will be a fundamental reevaluation of the value of all AI data center debt, which may lead to a sell-off of the underlying bonds and associated debt, which will make any investor deeply entrenched in the GPU credit business extremely nervous and, in some cases, unable to exit their positions in anything short of an embarrassing fashion.”
Our reading: Once AI data center debt becomes untenable, all such debt will be reevaluated, possibly triggering a bond sell-off that traps GPU credit investors.
No horizon
Only if: once AI data center debt becomes untenable
Assessment: UnresolvedSaid
Forecast
“At some point these projects become too risky for even the most mold-poisoned private credit fund or brainless Japanese bank to stomach, and the timeline will accelerate based on either Treasury rates or further data center developments facing cashflow or construction problems.”
Our reading: AI data center projects will at some point become too risky for lenders to fund, sooner if Treasury rates rise or projects hit cashflow or construction problems.
No deadline stated Evidence as of
Limitation: The breadth required for a financing shutdown is an evaluator interpretation.
Assessment: UnresolvedSaid
Conditional forecast
“If they aren’t going to pay those $1.3 trillion in compute bills, the jig is up for AI data center demand.”
Our reading: If Anthropic and OpenAI don't pay their $1.3 trillion in compute bills, AI data center demand will collapse.
No deadline stated Evidence as of
Only if: if Anthropic and OpenAI don't pay their compute bills
Assessment: UnresolvedSaid
Conditional forecast
“If they fail to do so, everybody suffers. Hyperscalers miss revenue estimates. AI data center debt goes unpaid. Venture capitalists find their holdings washed out.”
Our reading: If Anthropic and OpenAI fail to reach the revenue their valuations rest on (for Anthropic, $190–200 billion a year within two years), everybody suffers: hyperscalers miss revenue estimates, AI data center debt goes unpaid, and venture capitalists find their holdings washed out.
Not yet due Evidence as of
Only if: if Anthropic and OpenAI fail to reach the revenue their valuations are based on
Assessment: SupportedSaid
Observation
“In 2025, Anthropic lost over $8 billion on $4.6 billion in revenue.”
Our reading: Anthropic's revenue for calendar 2025 was about $4.6 billion.
No horizon Evidence as of
Limitation: Zitron's source was a Reuters report on a leaked draft. The record checks his figure against the S-1 as filed.
Assessment: SupportedSaid
Observation
“It spent $12.65 billion in operating expenses to make $4.6 billion of revenue, otherwise known as spending $2.75 to make a dollar.”
Our reading: In 2025 Anthropic spent $12.65 billion in operating expenses to make $4.6 billion in revenue ($2.75 per $1).
No horizon Evidence as of
Limitation: Operating expenses are not identical to cash spending.
Assessment: Can’t be checkedSaid
Observation
“While things could change in 2026, it’s important to note how many people said that Anthropic was “a better business” that would “be profitable faster than OpenAI,” which is, until we are able to see both of their audited 2026 financials, somewhere between a myth and an outright lie.”
Our reading: The claim many people made, that Anthropic was a better business that would be profitable faster than OpenAI, is somewhere between a myth and an outright lie until both companies' audited 2026 financials can be seen.
No horizon
Qualifier: until both audited 2026 financials are available
Assessment: Can’t be checkedSaid
Conditional forecast
“This company will, if allowed to go public, likely lean on the very same junk-grade/high-yield debt that AI data centers and neoclouds like CoreWeave currently need, and it will do so at volumes of somewhere between $50 billion and $100 billion a year for a company with few assets, endless losses and a CEO with the grace of a drunk elephant.”
Our reading: If Anthropic goes public, it will likely rely on junk/high-yield debt of $50–100 billion a year.
No horizon
Only if: if Anthropic goes public
Assessment: Can’t be checkedSaid
Rhetoric
“Anthropic is not the future of technology, nor is it the next Google, nor is it the next Microsoft, nor is it, to quote Reuters, capable of “[transforming] the global economy more profoundly than industrialization, electricity and the internet.””
Our reading: Anthropic is not the future of technology, not the next Google or Microsoft, and not capable of transforming the global economy more profoundly than industrialization, electricity and the internet.
No horizon
Assessment: Can’t be checkedSaid
Rhetoric
“Anthropic is a cloud software company with volatile products and economics, sold in a fundamentally insincere and deceptive manner, pushed upon society with threats of death and destruction by aggressive zealots and members of the media bereft of shame.”
Our reading: Anthropic is a cloud software company with volatile products and economics, sold in a fundamentally insincere and deceptive manner and pushed on society with threats of death and destruction by aggressive zealots and shameless media.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We must treat OpenAI and Anthropic as what they are: economic disasters waiting to happen.”
Our reading: OpenAI and Anthropic should be treated as economic disasters waiting to happen.
No horizon
Held from this post (43)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews reached different assessments. Our reading, not assessed: In calendar 2025, Anthropic lost more than $8 billion.
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: It will be difficult to impossible to raise AI data center debt at rates below 8%.
- The two reviews reached different assessments. Our reading, not assessed: The author has serious questions about the validity of SiliconData's rising GPU price indices, which he believes measure spot rates rather than longer-term rates, creating an illusion of success that doesn't connect to reality.
- The two reviews reached different assessments. Our reading, not assessed: 80% of Anthropic's and OpenAI's enterprise revenues come from 1% of their customers, the vast majority of them unprofitable AI startups, meaning the labs' demand depends on venture capital continuing to fund those startups.
- The two reviews reached different assessments. Our reading, not assessed: Like Cognition, Harvey, Perplexity, Replit and basically every other major AI startup will one day run out of money and die or become a much, much smaller company, though they are all hoping to be acquired by a hyperscaler.
- The two reviews read the claim differently
- The two reviews read the claim differently
- Attribution is uncertain
- The record does not meet the current rules. Our reading, not assessed: Oracle will not repay its debts.
- The criteria are ambiguous. Our reading, not assessed: Anthropic and OpenAI cannot afford a growth slowdown given their compute obligations and valuations that assume, for Anthropic, $190–200 billion annual revenue within two years.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: Assuming it grows faster than OpenAI, Anthropic's 2026 revenue will be around $40 billion.
- The two reviews reached different assessments. Our reading, not assessed: The author does not see how Anthropic becomes able to afford its $518 billion in compute commitments or how it works its way out of its dire economics.
- The record does not meet the current rules. Our reading, not assessed: Without AI startups, Anthropic and OpenAI would lose 60–80% of enterprise revenue, making them even less likely to pay their $1.3 trillion in compute commitments.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: Most AI data center debt deals will not be repaid.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: At least half of NVIDIA and Broadcom AI chip sales are made years before the chips are put to use.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: It appears that $200–300 billion of NVIDIA's sales since 2022 are sitting in warehouses or unplugged in data centers waiting for power.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews reached different assessments. Our reading, not assessed: This kind of SPV is the template for data center debt, its associated investors are a large chunk of the capital funding it, and their ability to keep funding AI data center debt is what is holding up the industry.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: AI startups that are wrappers built on Anthropic's and OpenAI's models are fundamentally built on reselling the labs' tokens at a loss, which is why companies like Harvey and Perplexity must raise hundreds of millions of dollars every few months.
- The criteria are ambiguous. Our reading, not assessed: OpenAI and Anthropic account for 70–80% of all AI revenues and compute demand, funded by over $217 billion largely from Google, Amazon, Microsoft and NVIDIA in the last nine months.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: In the Sharon AI and Catalyst Compute contracts, capex is barely paid off over five years even ignoring opex and debt.
- The record does not meet the current rules. Our reading, not assessed: Anthropic's economics may have worsened in 2026, as shown by its use of adjusted margins to fake a Q3 2026 profit.
- The criteria are ambiguous. Our reading, not assessed: Outside Anthropic and OpenAI there are basically no meaningful revenue-generating AI companies.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: Microsoft, Google, Amazon, Meta, Oracle and far too many other companies have been hoarding hundreds of billions of dollars of AI chips that they either can't plug in or simply want to have in supply.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: Evidence shows Anthropic's economics have decayed over the last two years and its 2025 business was worse than OpenAI's.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: What is happening suggests the AI boom is more like five or six hyperscalers feeding money to NVIDIA and Broadcom so they can feed money to Anthropic and OpenAI, who then feed that money back to them whenever capacity comes online.
- The record does not meet the current rules. Our reading, not assessed: CoreWeave will not repay its debts.