A Warning from an Industry Insider

Anthropic CEO Dario Amodei delivered a blunt message to the AI industry at the 2025 DealBook Summit. While he expressed confidence in the technical advances of AI, he cautioned against reckless financial commitment to infrastructure and over-optimistic revenue projections.

Amodei highlighted a phenomenon he called “YOLOing,” where companies make aggressive, high-risk bets on AI compute infrastructure, often backed by circular deals with chip manufacturers. He warned that even small timing or demand miscalculations could lead to serious financial trouble.

The Economics Behind the Concern

Building large-scale data centres for AI is expensive; Amodei cited the cost of a single “gigawatt” data-centre buildout as around US$10 billion over five years. A poor timing decision now could leave companies with excess capacity and insufficient demand.

He pointed out that while his company has seen rapid revenue growth, from near zero in 2022 to several billion dollars in 2025, the path ahead is uncertain. He described this uncertainty using what he calls a “cone of uncertainty,” indicating a wide range of possible financial outcomes.

Why This Warning Resonates Right Now

Across the tech industry, there is mounting evidence of an “AI bubble.” Some analysts note that many generative-AI (gen-AI) pilots fail to deliver real business value, a “genAI paradox,” where the hype significantly outpaces actual returns.

The valuations of AI companies have soared, sometimes without solid long-term business models or predictable revenue streams. For investors and stakeholders, the worry is whether these valuations can be justified, or whether we’ll see a crash similar to the dot-com bubble.

What This Means for AI Stakeholders

The warning from a leading insider at Anthropic matters for more than just finance executives. For AI companies, it underlines the importance of cautious scale-up, investing in infrastructure only when demand and revenue visibility justify it.

For investors, it suggests the need to scrutinize AI businesses not just on ambition or hype but on realistic adoption rates, profitability, and long-term sustainability. For policymakers and public observers, it is a signal that regulators and watchdogs might soon need to step in — to manage risks that go beyond technology, including economic instability.

Broader Implications for Society and Tech Strategy

If the industry is overinvesting in compute capacity now, a sharp correction later could affect jobs, R&D budgets, and pricing in sectors reliant on AI. This could slow down innovation and shake confidence in generative-AI models.

It may also shift AI development strategies: from rapid expansion to consolidation, prioritizing enterprise-grade use cases with stable revenue rather than consumer-facing hype-driven products.