OpenAI goes on offensive against Anthropic in internal memo

Summary

• OpenAI sent an internal memo to shareholders accusing rival Anthropic of a "lack of compute" (hardware, data centers, energy).
• Compute scale is framed as the key strategic differentiator between the two leading AI labs.
• OpenAI outlines plans to reach 30GW of compute by 2030, claiming its ramp is "materially ahead and widening."
• Anthropic is expected to have roughly 7 to 8GW of compute by the end of next year.
• The strategic disagreement is pronounced: OpenAI plans to spend about $600 billion on compute over the next five years.
• Anthropic's CEO, Dario Amodei, criticizes this as a "YOLO approach" and frames Anthropic's more conservative buildout as responsible capital allocation.
• OpenAI cites commentator Ben Thompson, who argues Anthropic's compute constraint could put a "ceiling on growth."
• OpenAI dismisses Anthropic's slow, limited rollout of its new model, suggesting it's due to insufficient compute, not security concerns.