Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Ed Zitron Where's Your Ed At Published

Four Horsemen of the AIpocalypse

We recorded 31 claims from this post. 19 were checked against evidence. 12 can’t be checked. 27 more are held. Read the post at its source.

Claims from this post (31)

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.

  1. Assessment: UnresolvedSaid

    “Pale Horse: Any further price increases or service degradations from Anthropic and OpenAI are a sign that they’re running low on cash.”

    What we tested: Further price increases or service degradations from Anthropic and OpenAI would indicate they are running low on cash.

  2. Assessment: Can’t be checkedSaid

    “Let’s start with a fairly direct statement: Anthropic should stop taking on new customers until it works out its capacity issues.”

    What we tested: Anthropic should stop accepting new customers until it resolves its capacity issues.

  3. Assessment: UnresolvedSaid

    “It is blatantly, alarmingly obvious that Anthropic cannot afford to provide a stable and reliable service to its customers”

    What we tested: Anthropic cannot afford to provide a stable and reliable service to its customers.

  4. Assessment: Can’t be checkedSaid

    “We’re likely hitting the absolute physical limits of available AI compute capacity, if we haven’t already done so”

    What we tested: The AI industry is likely hitting, or has already hit, the physical limits of available AI compute capacity.

  5. Assessment: UnresolvedSaid

    “Neither company is facing a financial shortfall as a result of being unable to provide their services (rather, they’re facing financial shortfalls because they’re providing their services to customers), and the only ones paying that price because of these “capacity limits” are the customers.”

    What we tested: Neither Anthropic nor OpenAI faces a financial shortfall from being unable to provide their services; their shortfalls come from providing their services to customers, and the only ones paying the price of the capacity limits are the customers.

  6. Assessment: UnresolvedSaid

    “As a result, adding more capacity simply makes the product work better for a much higher cost.”

    What we tested: Because Anthropic's capacity issues are not affecting its growth, adding more capacity simply makes the product work better for a much higher cost.

  7. Assessment: UnresolvedSaid

    “Venture capital funding allows it to massively subsidize its services.”

    What we tested: Venture capital funding allows Anthropic to massively subsidize its services.

  8. Assessment: Can’t be checkedSaid

    “Anthropic is a con, and said con is only made possible through endless, specious hype.”

    What we tested: Anthropic is a con sustained by hype.

  9. Assessment: Can’t be checkedSaid

    “These deals shouldn’t be legal.”

    What we tested: Deals like Anthropic's agreement to spend $100 billion on AWS as part of Amazon's investment in it should not be legal.

  10. Assessment: UnresolvedSaid

    “So, to summarize, at least 4.6GW of the 15.2GW of data center capacity under construction is for OpenAI, with at least another 4GW of that reserved for Anthropic through partners like Microsoft, Google and Amazon.”

    What we tested: At least 4.6GW of the 15.2GW of AI data center capacity under construction is for OpenAI and at least 4GW for Anthropic.

  11. Assessment: UnresolvedSaid

    “NVIDIA Is Selling Years’ Worth of GPUs In Advance And Warehousing Them”

    What we tested: NVIDIA is selling years' worth of GPUs in advance of installable capacity, and those GPUs are being warehoused.

  12. Assessment: UnresolvedSaid

    “If we then cut that up into GB200 or GB300 NVL72 racks with a power draw of around 140KW, that’s around 71,429 racks’ worth of hardware at an average of $4 million each, which gives us around $285.7 billion in revenue for NVIDIA.”

    What we tested: By his estimate, the 15.2GW of data center capacity under construction through 2028 works out to around 71,429 NVL72 racks, or about $285.7 billion in NVIDIA revenue.

  13. Assessment: Can’t be checkedSaid

    “A flat 10% bump on the cost of software engineering is the direct opposite of what AI was meant to do, and in the event that costs continue to rise, I’m not sure how anybody justifies the expense much further.”

    What we tested: A flat 10% increase in software engineering costs is the direct opposite of what AI was meant to do, and if costs continue to rise he is not sure how anybody justifies the expense much further.

  14. Assessment: UnresolvedSaid

    “And we’re going to find out fairly quickly, because the world of token subsidies is going away.”

    What we tested: Subsidized AI token pricing is ending soon, exposing customers to the true costs.

  15. Assessment: UnresolvedSaid

    “This is a significant move, driven by (per my own reporting) Microsoft’s week-over-week costs of running GitHub Copilot nearly doubling since January.”

    What we tested: Microsoft's GitHub Copilot changes are driven by its weekly running costs nearly doubling since January 2026.

  16. Assessment: UnresolvedSaid

    “Nevertheless, these services still burn tokens at whatever rate that it costs to pay for them — for example, $5 per million input and $25 per million output for Opus 4.7, as I mentioned previously — meaning that the company almost always loses money unless a person doesn’t use the subscription very much.”

    What we tested: AI subscription services almost always lose money on users unless they use the subscription lightly.

  17. Assessment: UnresolvedSaid

    “An aside/explainer: if you’re confused as to what “token-based billing” means, know that the vast majority of AI services currently subsidize their subscriptions, using another measure (such as “requests” or “rate limits”) to meter out how much a user can use the service.”

    What we tested: The vast majority of AI services currently subsidize their subscriptions.

  18. Assessment: Can’t be checkedSaid

    “Companies did this to grow their subscriber numbers, and I think they assumed things would get cheaper somehow.”

    What we tested: AI companies subsidized subscriptions to grow their subscriber numbers, and he thinks they assumed things would somehow get cheaper.

  19. Assessment: Can’t be checkedSaid

    “I’ve literally never seen a company do this outside of products it intended to kill entirely, and that’s likely because — per my source — it intends to move paid customers over to token-based-billing”

    What we tested: Microsoft's halt to new paid GitHub Copilot signups is likely because, per his source, it intends to move paid customers to token-based billing.

  20. Assessment: Can’t be checkedSaid

    “Though Microsoft is yet to formally announce the move to Token-Based Billing, I imagine it’ll be sometime this week that it rips off the bandage.”

    What we tested: Microsoft will probably formally announce token-based billing for GitHub Copilot this week.

  21. Assessment: UnresolvedSaid

    “I’m making the call that by the end of 2026, a majority of AI services will move some or all of their customers to token-based billing as they reckon with the true costs of running AI models.”

    What we tested: By the end of 2026, most AI services will move some or all customers to token-based billing.

  22. Assessment: Can’t be checkedSaid

    “Though my arguments might be verbose, they’re ultimately pretty simple: AI does not provide even an iota of the benefits — economic or otherwise — to justify its ruinous costs.”

    What we tested: AI does not provide even an iota of the benefits, economic or otherwise, that would justify its ruinous costs.

  23. Assessment: UnresolvedSaid

    “The entire industry — including OpenAI and Anthropic’s theoretical revenues of $13.1 billion and $4.5 billion — hit around $65 billion last year, and that includes the revenues from providing compute generated by neoclouds like CoreWeave and hyperscalers like Microsoft.”

    What we tested: The entire AI industry, including OpenAI's and Anthropic's theoretical revenues of $13.1 billion and $4.5 billion, hit around $65 billion in revenue last year.

  24. Assessment: Can’t be checkedSaid

    “Cursor is also reportedly at $6 billion in annualized revenue (or around $500 million a month) and “gross margin positive” — which I also doubt given that it had to raise over $3 billion last year and is apparently raising another $2 billion this year.”

    What we tested: Cursor is probably not gross margin positive as reported.

  25. Assessment: UnresolvedSaid

    “Things have gotten so dire that even Deidre Bosa of CNBC agrees with me that AI demand is inflated by token-maxxing and subsidized services.”

    What we tested: AI demand is inflated by token-maxxing and subsidized services.

  26. Assessment: UnresolvedSaid

    “Even if said numbers were real, the majority of OpenAI, Cursor and Anthropic’s revenues come from subsidized software subscriptions.”

    What we tested: Most of OpenAI's, Cursor's and Anthropic's revenue comes from subsidized software subscriptions.

  27. Assessment: UnresolvedSaid

    “Otherwise, everybody else is making single or double-digit millions of dollars and losing hundreds of millions of dollars to get there.”

    What we tested: Every AI startup other than OpenAI, Cursor and Anthropic is making single- or double-digit millions of dollars in revenue and losing hundreds of millions of dollars to get there.

  28. Assessment: UnresolvedSaid

    “And at the top of the pile sits NVIDIA, the largest company on the stock market, which is selling more GPUs than can be possibly installed, and very few people seem to notice or care.”

    What we tested: NVIDIA is selling more GPUs than can possibly be installed.

  29. Assessment: UnresolvedSaid

    “I’m talking about hundreds of billions of dollars of GPUs sitting in warehouses that aren’t being installed, with it taking six months to install a single quarter’s worth of GPU sales.”

    What we tested: Hundreds of billions of dollars of GPUs are sitting uninstalled in warehouses, with a quarter's GPU sales taking six months to install.

  30. Assessment: Can’t be checkedSaid

    “If, in fact, NVIDIA is actually selling as many GPUs as it says, it’s likely taking liberties with “transfers of ownership” where NVIDIA marks a product as “sold” to somebody that has yet to actually take it on.”

    What we tested: If NVIDIA is selling as many GPUs as it claims, it is likely booking sales before customers actually take delivery.

  31. Assessment: Can’t be checkedSaid

    “But every sign points to me being right, and if I’m right at the scale I think I’m right, I think we’re about to have a legitimacy crisis in investing and mainstream media, because regular people are keenly aware that something isn’t right, in many cases, it’s because they’re able to count.”

    What we tested: If he is right at the scale he thinks he is, he thinks investing and mainstream media are about to have a legitimacy crisis, because regular people are keenly aware that something isn't right.

Held from this post (27)

These records did not pass every check, so nothing about them is published. Held is not a finding.