A Quick Note on the AI Bubble: Either the Market or Arithmetic Is Wrong

A Quick Note on the AI Bubble: Either the Market or Arithmetic Is Wrong

Center for Economic and Policy Research (CEPR)
Center for Economic and Policy Research (CEPR)Apr 12, 2026

Companies Mentioned

Why It Matters

The steep cost gap threatens the profitability and market caps of U.S. AI leaders, potentially reshaping investment flows toward cheaper Chinese alternatives.

Key Takeaways

  • Chinese AI models cost $2‑3 per million tokens
  • US AI providers charge about $15 per million tokens
  • Token pricing gap could pressure US AI valuations
  • Lower Chinese costs stem from efficient chips and cheap electricity
  • Potential market shift may trigger consolidation in AI industry

Pulse Analysis

Token economics have become a new battlefield in the generative‑AI race. While U.S. firms such as OpenAI price output at roughly $15 per million tokens, Chinese providers are offering the same computational power for $2‑3. This advantage is driven by a combination of highly optimized model architectures, access to low‑cost electricity, and state‑backed subsidies that keep operational expenses minimal. For enterprise customers, the price differential translates into dramatically lower total cost of ownership, especially for large‑scale deployments that process billions of tokens daily.

The pricing gap is more than a cost issue; it strikes at the heart of valuation models for U.S. AI companies. Investors have been rewarding these firms with sky‑high market caps based on projected revenue streams that assume sustained premium pricing. If customers can source comparable performance at a fraction of the cost, revenue forecasts may need to be revised, echoing the over‑optimism that fueled the early‑2000s internet bubble. The article’s comparison highlights the risk that inflated expectations could lead to a correction, forcing American AI players to either slash prices, improve efficiency, or diversify revenue beyond token sales.

Policy and market dynamics will shape the next chapter. While patriotic buying and potential regulatory barriers could provide a temporary moat for U.S. firms, long‑term competitiveness will likely depend on innovation in model efficiency and strategic partnerships. Chinese firms may continue to leverage cost advantages to capture market share, prompting possible consolidation among Western providers. Stakeholders should monitor pricing trends, cost‑structure disclosures, and any government interventions that could tilt the balance in this rapidly evolving sector.

A Quick Note on the AI Bubble: Either the Market or Arithmetic is Wrong

Comments

Want to join the conversation?

Loading comments...