Anthropic et OpenAI : la bulle de l'IA est-elle en train d'éclater ?

Anthropic et OpenAI : la bulle de l'IA est-elle en train d'éclater ?

Anthropic and OpenAI: is the AI bubble about to burst?

by Carlos Diaz (Silicon Carne) · View the channel on PSI

Humanities, Social Sciences & Thought Economics & Finance KCEconomicsKCXEconomic and financial crises and disasters
🎙 Carlos Diaz (Silicon Carne) 👥 99K 📅 October 1, 2026 ⏱ 34 min 👁 31K 📄 debate 🧭 2026-10-16
Available in: English (current) Français

Keywords

intelligence artificiellebulle financièreIPOAnthropicOpenAI

Summary

The episode analyzes the decision by Anthropic and OpenAI to postpone their IPOs, questioning whether this signals a looming AI bubble burst. Host Carlos Diaz, joined by guests Evan Kervella and Wallerrand Moullé-Berteaux, examines the financial and strategic motivations behind these delays. The discussion covers OpenAI’s filing of an IPO document in June followed by its announcement that it would not go public in 2026, and Anthropic’s one-month delay from October to November. Financial Times projections revealing $278 billion in cumulative losses for OpenAI by 2030 are discussed as a key impediment to going public. The video explores the impact of Chinese AI models engaging in aggressive price dumping, with open-weight models capturing 80% of token usage within six months, though only 14% of industry revenue. The founders of Anthropic, holding only about 2% of capital each, are reported to demand majority voting control as a condition for the IPO. The value migration from frontier models to infrastructure is analyzed, with AI data center investment projected at $10.3 trillion between 2025 and 2032, representing 3.6% of GDP. Legal liability concerns are examined, drawing parallels between AI leaders and tobacco industry executives, with Dario Amodei explicitly stating he does not want to be like Philip Morris. The episode concludes by questioning whether either company can ever achieve profitability and whether the public should invest in them.

225 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers valuable analysis of the timing and motivations behind the IPO delays, combining information from financial media with expert commentary. The argumentation is nuanced, presenting multiple perspectives: strategic timing, fear of legal liability, Chinese competition through price dumping, and the migration of value toward infrastructure. Guests complement each other, with Kervella emphasizing strategic calculations behind the delays and Moullé-Berteaux highlighting the disruptive impact of Chinese models. However, much of the analysis remains speculative, with statements like ‘10% extinction risk’ being acknowledged as lacking scientific grounding. The historical parallels drawn (Philip Morris, Manhattan Project) provide useful framing but are employed loosely. The episode genuinely grapples with the core question of whether these companies can become profitable, and openly discusses the fragility of their business models. The discussion of infrastructure investment absorbing the value that once belonged to model developers is particularly insightful.

Scientific Rigor, Source Quality, Title Accuracy

The video references credible financial media (Wall Street Journal, Financial Times) and presents specific numeric data on valuations, loss projections, and market shares, lending a factual foundation to the discussion. However, sources are cited verbally without on-screen references or linked citations, making verification difficult for viewers. The description contains only channel links (newsletter, podcast, LinkedIn profiles) rather than primary sources. The sponsor segment (Bitdefender) lasts approximately 80 seconds and is clearly identified as a partner section, without ambiguity. The title accurately reflects the content, and the analysis benefits from informed observers discussing rapidly evolving industry developments, though the reliance on speculation and opinion reduces the overall scientific rigor. The show is transparent about its format as a commentary/debate rather than a research presentation.

281 words

Title / Content Match

The title accurately reflects the content: the episode directly addresses whether the AI bubble is bursting through the lens of Anthropic and OpenAI's IPO postponements and financial challenges.

Quality & Reliability

6/10

The video provides a well-structured analysis of AI company IPO delays, citing credible financial media (WSJ, FT) and presenting specific figures. However, much of the reasoning is speculative opinion from guests, and sources are cited verbally without direct links. The financial projections and market analyses are plausible but unverifiable from the content alone.

Chapters

Cited Sources

  • Silicon Carne newsletter — The show's newsletter containing daily analyses and referenced sources.
  • Silicon Carne subscription page — Paid subscription option ($15/month) for full access to content and community.
  • Silicon Carne podcast — Audio version of the show available on podcast platforms.
  • Carlos Diaz LinkedIn profile — LinkedIn profile of the main host Carlos Diaz.
  • Evan Kervella LinkedIn profile — LinkedIn profile of guest Evan Kervella, an AI alignment researcher.
  • Wallerrand Moullé-Berteaux LinkedIn profile — LinkedIn profile of guest Wallerrand Moullé-Berteaux.

Concurring Sources

  • Wall Street Journal — Reported on September 18 that Anthropic is delaying its IPO from October to November, corroborating the video's central claim.
  • Financial Times — Revealed internal OpenAI projections of $278 billion in losses by 2030, cited in the episode as a major barrier to going public.

Contribution & Novelties

The video offers an original synthesis of recent, rapidly evolving information about AI company IPOs, connecting multiple threads: Chinese AI price dumping, infrastructure investment dynamics, legal liability concerns, and governance structures. Its distinctive insight is the observation that Anthropic founders, holding only ~2% of capital each, are demanding majority voting control, and the framing of AI companies competing with US Treasury bonds for global debt capital is thought-provoking. The discussion successfully reframes the IPO delays not as mere caution but as a strategic response to structural changes in AI economics, particularly the shift of value from model providers to infrastructure operators.

Pour aller plus loin :

  • Economic bubble — Essential context for evaluating whether AI valuations constitute an asset bubble comparable to historical precedents.
  • DeepSeek — A leading example of Chinese open-weight AI models that undercut frontier model pricing, central to the ‘Chinese dumping’ discussion.
  • AI winter — Historical cycles of AI investment hype and bust, providing a framework to assess whether current circumstances resemble previous downturns.
  • AI alignment — The technical and philosophical challenge of ensuring AI systems behave safely, underpinning the existential risk debates in the episode.

189 words

Radar Profile

The radar profile shows a video strong in information quantity but moderate in technical depth and reliability. It performs well on information density and presentation quality, while scoring lower on methodological rigor and source verifiability, consistent with its debate-oriented format rather than a research presentation.

Reliability 6/10