Skip to content
Escape
Meridian

The future is still open.

· Nov 13, 2026

Gary Marcus on NVIDIA, Jensen Huang and Anthropic in the generative AI unraveling

Markdown · JSON · Original source

argues that generative AI’s economics may be nearing a breaking point: training and inference are expensive, competition has pressured prices, and profits remain elusive. He says companies have yet to demonstrate lasting profitability and questions whether they can stand independently if stops financing or supporting other firms. Falling valuations and reports of strained AI loans, alongside the possible overcommitment of company Nscale, are signs Marcus reads as a bubble beginning to pop—not proof that it has. He connects the risk to companies’ reliance on Nvidia’s financing and to ’s willingness to make large bets; a change in Nvidia’s approach, he suggests, could leave neocloud companies in trouble and jeopardize ’s IPO. Marcus also points to studies he says show a lack of ROI, reliability concerns, weak instruction-following, and companies rehiring humans after AI claims. He acknowledges that valuations stayed high longer than he expected, and says his sense that markets now recognize the economics’ irrationality could be wrong. His warning is about financial fragility and uncertain returns, not a claim that generative AI has no future: he allows that future systems may be better, but says “we are not there yet.”

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

What if Generative AI turned out to be a Dud?
Marcus on AI · What if Generative AI turned out to be a Dud? · Open original source

Original source

Marcus on AI · Article

Source details

Original source date: Nov 13, 2026 (2026-11-13T18:26:49+00:00).

Date basis: Date stated by the publisher.

Added to Escape Meridian: 2026-11-13T18:33:06.546Z. First observed: 2026-11-13T18:30:37.682Z. Retrieved: 2026-11-13T18:32:15.239Z.

Referenced by

Recorded citations and public file attachments. A reference alone does not establish agreement or influence.

Some references could not be checked. Reload this page to try again.

Reference data (JSON)

Condition

Financing of the build-out

Escape Meridian files this reading under a condition it tracks for AI’s trajectory. The label is EM’s selection, not a claim of the source.

Limits

The article presents Marcus’s interpretation of market declines and financing risk; the evidence here does not establish that a bubble is definitively popping. The article cites reported development costs, studies, company valuations, and claims about profitability, but this excerpt does not provide the underlying studies or financial records for independent assessment. Marcus describes possible consequences if reduces its bets; the excerpt does not establish whether Nvidia will do so.

Open question

Will stop making the large bets Marcus says have supported other companies, and what would follow if it did?

Source-stated

Marcus argues that shared techniques and training data make it difficult for generative AI companies to maintain technical advantages, contributing to price competition and elusive profits.

nobody would have much of a technical moat, and that there would be price wars, and the profits would be hard to eke out.

Source-stated

Marcus says training and inference costs leave little room for profits in a competitive market.

Training is ridiculously expensive and getting more expensive. The Claude 5.5 model family alone reportedly cost $10 billion to develop. And inference is expensive too.

Source-stated

Marcus says Nvidia’s financing and willingness to make large bets may be important to other companies’ prospects; he warns that a change could put neocloud companies in trouble and jeopardize Anthropic’s IPO.

If Nvidia stops acting like everybody’s banker, a bunch of neocloud companies are in trouble. Anthropic’s IPO could be jeopardized, too.

Source-stated

Marcus treats falling valuations and financing strains as signs that an AI investment bubble may be popping, while presenting that as his interpretation rather than a certainty.

People may not want to admit it, but those numbers sure look like a bubble popping to me.

Source-stated

Marcus argues that current weaknesses do not rule out better future systems, but says that future has not arrived yet.

(Some future of AI may be better, of course, but we are not there yet).

Source and contributor context

Gary Marcus · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.

Source media

Media declared by the original publisher; availability checked separately from the editorial reading.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

Neocloud

A new breed of cloud compute provider focused on renting out GPU computing power: a pure-play GPU cloud whose economics are still evolving.

Gary Marcus

Gary Marcus publishes technical criticism of AI learning methods and policy arguments about reliability and oversight.

NVIDIA

NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.

Jensen Huang

Jensen Huang is NVIDIA’s founder and chief executive, with personal remarks attributed to their original interviews, speeches or signed publications.

Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

Editorial provenance

Version 1. Publication ID: 01c41284-5b74-47bf-83b8-35a8b0c4d950.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T18:33:06.546Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: condition_financing.

Observation: 804518c3efb98e8597e3376703870fbeb673b53e0ba041f5b8152471aded90db. Source text hash: 065f3439e19c706833fc0d168d767aac83816787ab567b06b29b13b22a0c7058.

Revision and source-check history

Version 1 · 2026-11-13T18:33:06.546Z

Gary Marcus on NVIDIA, Jensen Huang and Anthropic in the generative AI unraveling

argues that generative AI’s economics may be nearing a breaking point: training and inference are expensive, competition has pressured prices, and profits remain elusive. He says companies have yet to demonstrate lasting profitability and questions whether they can stand independently if stops financing or supporting other firms. Falling valuations and reports of strained AI loans, alongside the possible overcommitment of company Nscale, are signs Marcus reads as a bubble beginning to pop—not proof that it has. He connects the risk to companies’ reliance on Nvidia’s financing and to ’s willingness to make large bets; a change in Nvidia’s approach, he suggests, could leave neocloud companies in trouble and jeopardize ’s IPO. Marcus also points to studies he says show a lack of ROI, reliability concerns, weak instruction-following, and companies rehiring humans after AI claims. He acknowledges that valuations stayed high longer than he expected, and says his sense that markets now recognize the economics’ irrationality could be wrong. His warning is about financial fragility and uncertain returns, not a claim that generative AI has no future: he allows that future systems may be better, but says “we are not there yet.”

Original source

Marcus on AI · Article

Source details

Original source date: Nov 13, 2026 (2026-11-13T18:26:49+00:00).

Date basis: Date stated by the publisher.

Added to Escape Meridian: 2026-11-13T18:33:06.546Z. First observed: 2026-11-13T18:30:37.682Z. Retrieved: 2026-11-13T18:32:15.239Z.

Condition

Financing of the build-out

Escape Meridian files this reading under a condition it tracks for AI’s trajectory. The label is EM’s selection, not a claim of the source.

Limits

The article presents Marcus’s interpretation of market declines and financing risk; the evidence here does not establish that a bubble is definitively popping. The article cites reported development costs, studies, company valuations, and claims about profitability, but this excerpt does not provide the underlying studies or financial records for independent assessment. Marcus describes possible consequences if reduces its bets; the excerpt does not establish whether Nvidia will do so.

Open question

Will stop making the large bets Marcus says have supported other companies, and what would follow if it did?

Source-stated

Marcus argues that shared techniques and training data make it difficult for generative AI companies to maintain technical advantages, contributing to price competition and elusive profits.

nobody would have much of a technical moat, and that there would be price wars, and the profits would be hard to eke out.

Source-stated

Marcus says training and inference costs leave little room for profits in a competitive market.

Training is ridiculously expensive and getting more expensive. The Claude 5.5 model family alone reportedly cost $10 billion to develop. And inference is expensive too.

Source-stated

Marcus says Nvidia’s financing and willingness to make large bets may be important to other companies’ prospects; he warns that a change could put neocloud companies in trouble and jeopardize Anthropic’s IPO.

If Nvidia stops acting like everybody’s banker, a bunch of neocloud companies are in trouble. Anthropic’s IPO could be jeopardized, too.

Source-stated

Marcus treats falling valuations and financing strains as signs that an AI investment bubble may be popping, while presenting that as his interpretation rather than a certainty.

People may not want to admit it, but those numbers sure look like a bubble popping to me.

Source-stated

Marcus argues that current weaknesses do not rule out better future systems, but says that future has not arrived yet.

(Some future of AI may be better, of course, but we are not there yet).

Source and contributor context

Gary Marcus · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.

Source media

Media declared by the original publisher; availability checked separately from the editorial reading.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

Neocloud

A new breed of cloud compute provider focused on renting out GPU computing power: a pure-play GPU cloud whose economics are still evolving.

Gary Marcus

Gary Marcus publishes technical criticism of AI learning methods and policy arguments about reliability and oversight.

NVIDIA

NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.

Jensen Huang

Jensen Huang is NVIDIA’s founder and chief executive, with personal remarks attributed to their original interviews, speeches or signed publications.

Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

Editorial provenance

Version 1. Publication ID: 01c41284-5b74-47bf-83b8-35a8b0c4d950.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T18:33:06.546Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: condition_financing.

Observation: 804518c3efb98e8597e3376703870fbeb673b53e0ba041f5b8152471aded90db. Source text hash: 065f3439e19c706833fc0d168d767aac83816787ab567b06b29b13b22a0c7058.

uncertain · 2026-11-13T18:33:06.546Z · anthropic/claude-opus-5.5

Publication: 01c41284-5b74-47bf-83b8-35a8b0c4d950

attribution — pass: The reading attributes the argument to Gary Marcus, the author of the Substack post, and keeps market and loan claims framed as his reading or as reports he cites.

Since August of 2023, I have been warning that, financially, generative AI might turn out to be a dud.

summary — uncertain: The reading leaves out the trigger for Marcus's argument: a leadership change at Nvidia (the 'Huang Affair', 'change in regime' and an interim CEO, Colette Kress). Without it, readers cannot see why Nvidia's bets might stop, and the question lacks grounding. Smallest repair: add one sentence saying Marcus ties the risk to Nvidia's change in regime under an interim CEO. 'Possible overcommitment' also softens the bulletin, which reported that Nscale had overcommitted; attribute it to the bulletin instead. 'Neocloud' needs a plain-language gloss or tooltip.

And of course the change in regime at Nvidia may have massive repercussions.

date — pass: originalPublished (2026-11-13T18:26:49Z, from the publisher page) matches the article's byline date of Nov 13, 2026.

Gary Marcus Nov 13, 2026

support — pass: Each claim and its excerpt is source-stated and stays attributed to Marcus. The limitations correctly note that the bubble call is his interpretation and that the underlying studies and financials are not supplied. The question does not presuppose an outcome.

If Nvidia stops acting like everybody’s banker, a bunch of neocloud companies are in trouble. Anthropic’s IPO could be jeopardized, too.

media — pass: The selected media is the verified direct link to the original article, which is the most direct way to reach the source.

Challenge or correct this reading

Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.