• inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.

Puts numbers on the physical limit behind future model scale: single training runs projected at 2GW to 5GW by 2028 and roughly 50GW of US capacity needed, which bounds how fast frontier capability and inference supply can grow.

• articleSecuring America S Compute Advantage Anthropic S Position On The Diffusion Rule
• articleOur Framework For Developing Safe And Trustworthy Agents
• articleClaude Gov Models For U S National Security Customers
• article3 5 Models And Computer Use

• articleSpaceX 10GW in 2027 – Why It’s Real, Will Drive $300B ARR for SpaceX, and Why Microsoft Will Be the Largest OfftakerJeremie Eliahou Ontiveros
• articleAnthropic Invests 50 Billion In American Ai InfrastructureAnthropic News
• articlexAI’s Colossus 2 – First Gigawatt Datacenter In The World, Unique RL Methodology, Capital RaiseJeremie Eliahou Ontiveros

Checking sign-in…

Loading comments…