Hyperscale Normalization
We recorded 26 claims from this post. 13 were checked against evidence. 13 can’t be checked. 16 more are held. Read the post at its source.
Claims from this post (26)
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: Can’t be checkedSaid
Observation
“Sidenote: NVIDIA deploying capital at random to help counterparties raise debt, despite being legal, should immediately trigger memories of Enron, if only because their actions have the same intention of artificially inflating revenues. Real businesses do not need financial wizardry.”
Our reading: NVIDIA's investments helping counterparties raise debt are intended to artificially inflate its revenues, comparable to Enron.
No horizon
Assessment: Can’t be checkedSaid
Observation
“This is not the same government that could once afford to pump trillions of dollars into any kind of bailout without running the very real risk of destabilizing the US dollar.”
Our reading: The US government, with national debt now at $40 trillion, can no longer pump trillions of dollars into a bailout without running a very real risk of destabilizing the US dollar.
No horizon
Qualifier: with US national debt grown to $40 trillion
Assessment: Can’t be checkedSaid
Rhetoric
“The entire AI bubble — every bit of hype — is an attempt to rebrand old things as new.”
Our reading: All AI bubble hype is an attempt to rebrand old things as new.
No horizon
Assessment: UnresolvedSaid
Observation
“Similarly, when Anthropic hit $65 billion in “annualized run rate,” the number was printed without a second’s hesitation despite it being completely-undefined and indicative of nothing other than a snapshot of an indeterminately-long period multiplied by a number that we do not know.”
Our reading: Anthropic's reported $65 billion annualized run rate is an undefined figure that was reported uncritically.
No horizon Evidence as of
Limitation: Several relevant reports could not be read.
Assessment: UnresolvedSaid
Observation
“AI companies never share their actual revenues — $11.6 billion in Q2 2026 — or their underlying economics, and reporters have been so systemically-sedated that they believe that using a marketing number is reporting.”
Our reading: Anthropic's actual revenue in Q2 2026 was $11.6 billion, and AI companies do not disclose actual revenues or underlying economics.
No horizon Evidence as of
Limitation: The quarterly filing data are unaudited.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“There is nothing rational about valuing Cognition at $26 billion, let alone the $40 billion one it’s allegedly raising at right now.”
Our reading: There is nothing rational about valuing Cognition at $26 billion, let alone at the $40 billion it is allegedly raising at now.
No horizon
Qualifier: $40 billion round is 'allegedly' being raised
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“I have no specific beef with Cognition, because the same can be said of Perplexity, Higgsfield, Harvey, or any number of other AI companies with triple-digit “annualized run rates” with double-digit billion valuations for businesses with questionable business models.”
Our reading: The same criticism made of Cognition applies to Perplexity, Higgsfield, Harvey and any number of other AI companies with triple-digit 'annualized run rates' and double-digit-billion valuations for businesses with questionable business models.
No horizon
Scope: AI companies with triple-digit annualized run rates and double-digit-billion valuations
Assessment: UnresolvedSaid
Observation
“When finished, a data center creates roughly 100 to 200 jobs, or less than a large Walmart for something that brings little or no economic value.”
Our reading: A finished AI data center creates roughly 100–200 jobs and brings little or no economic value.
No horizon Evidence as of
Limitation: Earlier general data-center evidence cannot directly establish current AI-facility outcomes.
Assessment: UnresolvedSaid
Observation
“As for the industries it’s created, it’s unclear. Even if we were being generous to describe the companies that exist to throw a layer over an existing AI model, those “industries” employ a negligible amount of people.”
Our reading: Industries created by AI, such as companies wrapping existing models, employ a negligible number of people.
No horizon Evidence as of
Limitation: The employment estimate covers 2024 and includes model developers and diversified businesses [e7].
Assessment: UnresolvedSaid
Observation
“None of this matters, because basically every AI data center I’ve seen uses behind-the-meter gas turbines that are an environmental disaster.”
Our reading: Basically every AI data center Zitron has seen uses environmentally disastrous behind-the-meter gas turbines.
No horizon Evidence as of
Limitation: The environmental description is qualitative.
Assessment: UnresolvedSaid
Observation
“Every article about building them automatically assumes this is a necessary buildout because lots of money has gone into them, but nobody can seem to explain why other than “there’s so much demand for AI services (which is not actually true when you remove OpenAI and Anthropic).””
Our reading: Excluding OpenAI and Anthropic, there is not much demand for AI services.
No horizon Evidence as of
Limitation: A broader reading that counts free usage would make Google's figures more directly adverse to the claim.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“There is nothing ‘anti-progress’ about opposing AI data centers, and nobody has a compelling explanation as to why we need more of them.”
Our reading: Nobody has a compelling explanation for why more AI data centers are needed, and there is nothing 'anti-progress' about opposing them.
No horizon
Assessment: UnresolvedSaid
Causal claim
“While the money is absolutely real, it’s dependent on both the continued value of investing in AI data centers — which is an open question — and the ability for these three to five customers to be able to keep raising tens of billions of dollars whenever they need to.”
Our reading: NVIDIA's revenue, though absolutely real, depends both on the continued value of investing in AI data centers, which is an open question, and on its three to five main customers being able to keep raising tens of billions of dollars whenever they need to.
No horizon Evidence as of
Limitation: Financing activity demonstrates financial links, not necessarily dependence.
Assessment: UnresolvedSaid
Forecast
“Google, Amazon and Microsoft are currently spending hundreds of billions of dollars on capex to, for the most part, pull in around $440 billion in revenue in the next three years from Anthropic and OpenAI, as they represent more than 70% of their AI revenues in the next three years, and more than 48% of Google Cloud’s 2027 revenue.”
Our reading: Google, Amazon and Microsoft are spending hundreds of billions on capex mostly to bring in around $440 billion in revenue from Anthropic and OpenAI over the next three years, and those two will make up more than 70% of their AI revenues in that period and more than 48% of Google Cloud's 2027 revenue.
Not yet due Evidence as of
Limitation: One metric-scope object cannot separately encode the capital-spending component and all three revenue measures; their distinct requirements are stated in the contract.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“CoreWeave is a bad company that only exists because of the simple, systemic belief that If The Right Numbers Are Going Up, Everything Is Fine.”
Our reading: CoreWeave is a bad company that only exists because of the simple, systemic belief that if the right numbers are going up, everything is fine.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Everything, even often in critical pieces, comes back to statements that reinforce the status quo, like “AI is, of course, transformative,” or “bubbles always form and leave value afterward,” even though that’s not really true at all when you look at history, and certainly isn’t true of this particular era.”
Our reading: The common claim that bubbles always form and leave value afterward is not really true historically, and certainly is not true of the current AI era.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“I’m sure there are automations that people have made with this stuff that help them — good for you! — but it will take a lot more than that to justify a trillion-plus dollars in investment or endless fucking prattling about how impressive and world-changing AI is.”
Our reading: Automations people have built with LLMs that help them are not enough; it will take a lot more to justify a trillion-plus dollars of investment or the endless talk about how impressive and world-changing AI is.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The thing is, LLMs have improved in the last year, just not in a way that’s tremendously impressive to me given the amount of capital invested, or in a way that has manifested in a tangible product that can be described in a sentence.”
Our reading: LLMs have improved in the last year, but not in a way that is tremendously impressive to the author given the capital invested, nor in a way that has produced a tangible product describable in a sentence.
No horizon
Qualifier: given the amount of capital invested
Assessment: Can’t be checkedSaid
Causal claim
“Sidenote: While I don’t dispute that the AI industry has done something, the scale of its importance is entirely a result of the media’s puffery, subsidized subscriptions, venture capital and hyperscaler funding and endless debt.”
Our reading: The AI industry's apparent importance is entirely due to media hype, subsidized subscriptions, VC and hyperscaler funding and debt.
No horizon
Assessment: UnresolvedSaid
Observation
“Anthropic signing a $35 billion cloud compute deal with an NVIDIA-backed neocloud Lambda will immediately be used as proof that the company is “here to stay,” and that Lambda “has huge customers and a giant backlog,” even though in reality it shows that there wasn’t anyone else willing or capable of signing a deal that large, nor was there enough diverse demand.”
Our reading: Anthropic's $35 billion compute deal with Lambda shows that nobody else was willing or capable of signing a deal that large, and that there was not enough diverse demand.
No horizon Evidence as of
Limitation: Public announcements are not a complete record of demand.
Assessment: UnresolvedSaid
Forecast
“Neither of these companies can actually afford this compute, but that doesn’t matter, because this data center won’t actually exist for years, if it ever does.”
Our reading: The data center under the Anthropic-Lambda deal will not exist for years, if ever.
Not yet due Evidence as of
Limitation: The two-year boundary and operational meaning of 'exist' are evaluator interpretations.
Assessment: Can’t be checkedSaid
Forecast
“In fact, every time you read about some multi-billion dollar data center deal, know that it won’t be built for years, but NVIDIA will likely book the revenue immediately, because all it has to do is help raise the debt for the banks to send the money and the GPUs to be sold.”
Our reading: Multi-billion-dollar data center deals will not be built for years, but NVIDIA will likely book the associated revenue immediately.
No horizon
Qualifier: multi-billion dollar data center deals
Assessment: UnresolvedSaid
Causal claim
“NVIDIA’s future growth — which is one of the load-bearing perceptual elements of the AI bubble — is entirely dependent on whether large companies both want to and are able to spend hundreds of billions of dollars, and their intent to do so is largely based on whether two companies (Anthropic and OpenAI) can spend more than $400 billion on compute in the next few years.”
Our reading: NVIDIA's future growth is entirely dependent on whether large companies want to and can spend hundreds of billions of dollars, and their intent to do so is largely based on whether Anthropic and OpenAI can spend more than $400 billion on compute in the next few years.
No horizon Evidence as of
Limitation: Company announcements describe intentions and may emphasize favorable demand drivers.
Assessment: UnresolvedSaid
Observation
“AI GPUs are, outside of rentals to Anthropic and OpenAI, generating very little economic value, and there are few signs that they will do so in the future outside of anecdotes and copium.”
Our reading: Outside rentals to Anthropic and OpenAI, AI GPUs generate very little economic value, with few signs that will change.
No horizon Evidence as of
Limitation: The commercial and technical evidence is company-reported, without matched independent measurements.
Assessment: UnresolvedSaid
Conditional forecast
“When this ends, so many people will ask how it happened, why nobody stopped it, and try and rationalize it within the systems they know. They’ll crave a bailout, even if they hate the companies, because the destruction of allowing things to wilt and die is so unusual within our society, and it’s scary to imagine them doing so.”
Our reading: When the AI bubble ends, many people will ask how it happened and will want a bailout of AI companies.
No deadline stated Evidence as of
Only if: when the AI bubble ends
Assessment: Can’t be checkedSaid
Observation
“LLMs create outputs based on a summary of past data, passing that off as “intelligence” in a way that coddles the user.”
Our reading: LLMs create outputs based on a summary of past data and pass that off as 'intelligence' in a way that coddles the user.
No horizon Evidence as of
Limitation: This does not reject the documented concern about sycophancy.
Held from this post (16)
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: Meta does not really have an AI business despite its AI capex.
- The two reviews reached different assessments. Our reading, not assessed: NVIDIA's customer base is contracting, with receivables increasingly concentrated in a few companies.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: Neither OpenAI nor Anthropic can actually afford the compute they are contracting for.
- The record does not meet the current rules. Our reading, not assessed: AI does not appear to have produced any national-scale productivity gains.
- The two reviews reached different assessments. Our reading, not assessed: Once the venture-capital- and debt-subsidized spending of the big AI players is removed, there is very little real money in the system, because nobody other than Jensen Huang and the data center developers is making a profit from AI.
- The criteria are ambiguous. Our reading, not assessed: Cognition's business loses hundreds of millions of dollars and requires burning billions more for only a chance to make less revenue than Duolingo.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: It is not clearly known whether running AI services or renting AI GPUs is profitable.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently