Anthropic Taps SpaceX’s 222,000-GPU Colossus Cluster to Boost Claude Limits
Anthropic is getting a serious compute boost, and yes, it involves Elon Musk’s SpaceX.
According to Tom’s Hardware, Anthropic has signed a deal to use SpaceX’s Colossus 1 data center, a massive AI supercomputer cluster with more than 222,000 Nvidia GPUs. That includes H100 and H200 chips, plus newer GB200 accelerator systems. The setup reportedly delivers over 300 megawatts of AI compute power.
For normal users, the big question is simple: does Claude become less annoying to use when the limits hit? Anthropic says this extra capacity is aimed at improving the experience for paid Claude customers, especially Pro, Max, Team, and Enterprise users.
The immediate changes are pretty practical. Claude Code’s five-hour rate limits have doubled across paid tiers. Anthropic has also removed the peak-hours limit reduction for Pro and Max users, meaning the service should be less punishing during busy periods. Developers using Claude Opus models through the API are also getting significantly higher request limits.
That matters for SEA users more than it might sound. In Malaysia, Singapore, Indonesia, and the Philippines, a lot of small teams are already using AI tools for coding, content, support, translation, and internal automation. When cloud AI platforms throttle too hard, work gets disrupted. If Claude becomes more stable and less rate-limit-heavy, it could make paid plans easier to justify for startups, agencies, creators, and dev teams here.
The funny part is the politics of this deal. Colossus 1 was originally built for xAI’s Grok models, which puts Anthropic in direct rival territory. Musk said on X that he approved the lease after meeting senior Anthropic people and trying to understand how they keep Claude aligned with human benefit. His line was that nobody triggered his “evil detector.” That is quite a shift, considering he previously described Claude in much harsher terms earlier this year.
Anthropic is not only relying on SpaceX either. The company has also pointed to compute partnerships with Amazon, Google, and Microsoft as part of a bigger push to secure gigawatts of future AI capacity. Basically, frontier AI is no longer just a software race. It is a power, chips, cooling, land, and data center race.
The wildest bit? Anthropic and SpaceX are also talking about orbital AI compute — data centers in space. The argument is that future AI systems may need more power, cooling, and physical infrastructure than Earth-based facilities can scale fast enough to provide. Sounds sci-fi, but with SpaceX building space infrastructure and AI companies hungry for compute, this is the kind of idea that no longer feels completely unserious.
Anthropic also said future infrastructure will focus more on international regions, including Europe and Asia. The reason is enterprise demand for local data residency and compliance, especially in sensitive industries like healthcare, finance, and government. For Malaysia and SEA, this is the part to watch. If AI providers eventually expand serious capacity closer to the region, latency, compliance, and enterprise adoption could all improve.
Alongside the compute news, Anthropic also introduced a Claude feature called “dreaming.” It lets Claude-powered agents review past work, spot repeated mistakes, and reorganise memory files containing user preferences and context. In plain terms, Anthropic wants AI agents to improve between sessions instead of resetting like blur sotong every time.
For now, the headline is clear: Claude is getting access to one of the biggest Nvidia GPU clusters around, and paid users should see fewer bottlenecks. For Malaysian devs and AI-heavy teams, this could mean smoother coding sessions, better API throughput, and less waiting around when work is actually urgent.
Source: Tom's Hardware
