DealBook Summit Anthropic Chief Executive Has A.I. Bubble ‘Concerns’
42 snips
Dec 4, 2025 
Dario Amodei, the chief executive of Anthropic and former OpenAI researcher, shares his insights on the potential risks of an A.I. bubble. He discusses the 'cone of uncertainty' in A.I. spending, emphasizing the delicate balance between investing too much or too little. Amodei contrasts enterprise and consumer strategies in the A.I. landscape, while also addressing the implications of chip depreciation and national security concerns related to A.I. and technology regulation. His perspectives on job disruption highlight the need for proactive societal changes.
AI Snips
Chapters
Books
Transcript
Episode notes
The Cone Of Compute Uncertainty
- The economic timing uncertainty creates a "cone of uncertainty" around how much compute companies should buy.
- Small timing errors in buildouts can lead to severe overextension or missed market share, regardless of the technology's long-term value.
Scaling Laws Unlock Broad Value
- Simple scaling—more compute and data—has produced broad gains across many domains over a decade.
- Those gains translate into value across coding, science, biomedicine, law, finance, materials, and manufacturing.
Plan For The Low‑Side Case
- Plan conservatively within the cone and buy enough compute to be confident in low-percentile scenarios.
- Balance buying capacity to stay competitive while avoiding tail-risk overextension that could lead to bankruptcy.
