Descriptions:
Matthew Berman’s livestream reacts to the announcement that Anthropic has entered a partnership with SpaceX, granting full access to the Colossus 1 supercomputer — one of the world’s largest AI compute facilities with over 300 megawatts of additional capacity coming online within the month. The deal immediately translated into doubled Claude Code rate limits for Pro, Max, and Team subscribers, the removal of peak-hour throttling, and substantially higher Opus model API rate limits, marking a significant reversal of the compute constraints that had frustrated developers for months.
Berman traces the strategic context behind the deal: Anthropic CEO Dario Amodei’s early decision to be conservative on GPU acquisition — justified as risk management if AI demand failed to materialize — ultimately left the company compute-starved as demand surged beyond projections. He maps Anthropic’s subsequent infrastructure acquisition spree: an Amazon collaboration for up to 5 gigawatts of new capacity, partnerships with Google, Nvidia, and Fluid Stack involving $30 billion in Azure capacity, and now the XAI/SpaceX agreement despite Elon Musk’s history as a vocal Anthropic critic.
The video also covers Anthropic’s multi-hardware inference strategy across AWS Trainium, Google TPUs, and Nvidia GPUs, and touches on the emerging orbital compute discussion — with Jensen Huang expressing openness while Sam Altman dismisses it as impractical. Berman offers pointed commentary on Anthropic’s developer relations challenges, including opaque quota management and the broader competitive dynamics between Anthropic, OpenAI, and XAI that make the SpaceX partnership so structurally unusual.
📺 Source: Matthew Berman · Published May 07, 2026
🏷️ Format: Reaction
