Anthropic Taps SpaceX to Boost Claude Compute and Usage Limits

Anthropic has partnered with SpaceX to utilize the Colossus 1 data center, adding 220,000 GPUs to its compute infrastructure. This move allows the company to double Claude Code usage limits and significantly raise API limits for Claude Opus models. The expansion also includes a focus on international data residency and the exploration of space-based AI computing.

Key Points

• Anthropic is partnering with SpaceX to access 300MW of compute capacity and 220,000 GPUs at the Colossus 1 data center.
• Usage limits for Claude Code have been doubled, and peak-hour restrictions have been eliminated for premium subscribers.
• API rate limits for Claude Opus models have been significantly increased to accommodate higher demand.
• The company is expanding its infrastructure internationally to Asia and Europe to address regulatory and data residency needs.
• Anthropic and SpaceX are exploring the future development of orbital AI compute capacity.

Sentiment

The community is largely skeptical and critical. While some acknowledge the practical business logic of the deal for both parties, the dominant sentiment focuses on the contradiction between Anthropic's safety-first positioning and partnering with xAI's environmentally controversial infrastructure. Many commenters view the broader AI industry spending as a financially engineered bubble. The orbital compute aspect was widely mocked as either a grift or PR concession to SpaceX.

In Agreement

• Anthropic desperately needs compute capacity and this deal makes practical sense for both parties, with SpaceX monetizing idle infrastructure and Anthropic getting immediate GPU access
• The Colossus facility is located in an existing heavy industrial area next to steel mills, refineries, and power plants, and xAI pays substantial local taxes with no tax breaks
• SpaceX turning an underutilized asset into revenue is smart business ahead of their IPO
• Orbital compute may eventually become viable as terrestrial power and permitting bottlenecks worsen, and expressing interest costs Anthropic nothing

Opposed

• Anthropic's safety-focused branding is hypocritical when they partner with a facility that operated unpermitted gas turbines, faced Clean Air Act lawsuits, and posed health risks to low-income communities near Memphis
• The deal reveals xAI/Grok's failure — if Grok had sufficient demand, the datacenter wouldn't be available to lease to a competitor
• AI industry deals are increasingly circular and financially engineered, resembling subprime-era schemes where companies juice each other's revenue metrics ahead of IPOs
• Orbital AI compute is impractical due to heat dissipation challenges in vacuum, radiation-induced bit flips, GPU failure rates, massive launch costs, and the fundamental economics not making sense
• Anthropic claims to care about democracy while dealing with entities funded by non-democratic governments and using infrastructure built by someone who supported anti-democratic political movements