From 'Evil' to Essential: Elon Musk Hands Anthropic the Keys to His Supercomputer
Three months ago, Elon Musk called Anthropic 'evil.' Yesterday, he handed them his supercomputer. In what might be the most dramatic pivot in the AI industry this year, Anthropic and SpaceX announced a partnership that gives Anthropic full access to Colossus 1, SpaceX's massive data center in Memphis, Tennessee. The deal delivers over 220,000 NVIDIA GPUs and more than 300 megawatts of compute capacity, effective within the month. And that is just the terrestrial part. The two companies are also exploring something far more ambitious: multiple gigawatts of AI compute capacity in orbit.
By Ivaylo Tsvetkov, Co-Founder
What Exactly Happened
On May 6, 2026, both Anthropic and SpaceXAI (the entity formerly known as xAI, now folded into SpaceX) published simultaneous announcements confirming a compute infrastructure deal. Anthropic's blog post laid it out plainly: they have signed an agreement to use all of the compute capacity at Colossus 1. Not a slice. Not a partition. The entire facility. Colossus 1 is one of the world's largest AI supercomputers. Built from the ground up in record time, it houses dense deployments of NVIDIA H100, H200, and next-generation GB200 accelerators. It was designed for frontier-scale AI training and inference, exactly the kind of workload Anthropic needs to keep Claude running and improving.
Why Anthropic Needed This Badly
The context behind this deal is a capacity crisis. Anthropic CEO Dario Amodei revealed that the company saw 80x year-over-year growth in revenue and usage during Q1 2026, against a plan that had only accounted for 10x. That eight-fold underestimate created real pain: usage limits frustrated Claude Pro and Max subscribers, API rate limits constrained enterprise customers, and Claude Code users hit hourly caps during peak hours. This was not a strategic expansion. It was an emergency compute acquisition to keep pace with runaway demand. The SpaceX deal is part of a broader compute buildout. Anthropic has also secured capacity from Amazon (up to 5 GW total, with nearly 1 GW coming online by end of 2026), Google and Broadcom (5 GW, arriving in 2027), Microsoft and NVIDIA (30 billion dollars of Azure capacity), and Fluidstack (50 billion dollars in American AI infrastructure). Anthropic trains and runs Claude across AWS Trainium, Google TPUs, and NVIDIA GPUs, a deliberately multi-cloud, multi-silicon strategy.
What Changes for Users Immediately
The deal's impact is already hitting production. Anthropic announced several changes effective immediately. Claude Code rate limits have been doubled for Pro, Max, Team, and seat-based Enterprise plans. The peak hours limit reduction has been removed for Claude Code on Pro and Max accounts. API rate limits have been raised considerably for Claude Opus models. These are not future promises. The Colossus 1 capacity is being deployed now.
Why Musk Did It
With xAI dissolved as a separate entity and folded into SpaceX under the SpaceXAI brand, Musk had a world-class AI data center with capacity to monetize. By leasing Colossus 1 to Anthropic, SpaceX converts an expensive, power-hungry asset into a high-margin revenue stream. The timing is not accidental. SpaceX filed confidentially with the SEC on April 1, targeting a valuation between 1.75 trillion and 2 trillion dollars. The public S-1 is expected by late May, with the roadshow set for the week of June 8. Having Anthropic as a marquee compute customer, paying real money for real capacity, materially strengthens SpaceX's pitch as not just a launch company, but an AI infrastructure provider. Musk said he made the decision after meeting with senior Anthropic leaders, noting that their work on safety and alignment impressed him. The tone shift from adversary to business partner happened fast, but the logic is straightforward: Musk gets revenue and legitimacy for SpaceX's data center division, Anthropic gets the GPUs it desperately needs.
The Orbital Play: Multiple Gigawatts in Space
The most forward-looking element of the announcement is also the most audacious. As part of the agreement, Anthropic expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity. SpaceXAI's announcement put it in stark terms: the compute required to train and operate the next generation of AI systems is outpacing what terrestrial power, land, and cooling can deliver on the timelines that matter. Their argument is that SpaceX is the only organization with the launch cadence, mass-to-orbit economics, and constellation operations experience to make orbital compute a near-term engineering program rather than a research concept. Space-based compute offers near-limitless sustainable power with less environmental impact on Earth. This is still in the expressed-interest phase with no timelines or pricing shared. But the fact that it is in the official announcement signals serious intent from both sides.
What This Means for the AI Industry
This deal reshapes several narratives at once. For Anthropic, it demonstrates the ability to move aggressively on infrastructure without being locked into a single cloud provider. Their compute portfolio now spans Amazon, Google, Microsoft, NVIDIA, Fluidstack, and SpaceX, a level of supplier diversification that no other frontier AI lab has achieved. It also positions Anthropic well for its own potential IPO, reportedly being discussed for as early as October 2026. For SpaceX, the deal validates the data center business as a serious revenue line ahead of the IPO. Colossus 1 goes from a cost center built for the now-dissolved xAI to a contracted, revenue-generating asset. For the broader industry, it signals that compute scarcity is the real bottleneck, not algorithms, not data, not talent. When a company's CEO goes from public hostility to leasing his supercomputer in 90 days, the demand-supply imbalance in AI compute is doing the talking.
The Irony Nobody Is Missing
Musk is actively suing OpenAI, the company he co-founded, while simultaneously powering Anthropic, the company founded by people who left OpenAI over safety concerns. The enemy of my enemy, it turns out, is his customer. That irony aside, the underlying logic is clean. Anthropic has demand it cannot serve. SpaceX has supply it cannot use internally. Strategic feuds bend when the economics are strong enough.
What We Are Watching Next
Several threads to follow from here. The speed of Colossus 1 integration, since Anthropic said within the month, so we should see the capacity reflected in user-facing limits and API performance by early June. SpaceX's S-1 filing, where the Anthropic deal will almost certainly be highlighted as proof of the data center division's commercial viability. Anthropic's own IPO timeline, which with this level of compute secured gets significantly stronger for an October listing. And the orbital compute program, where any concrete engineering milestones or prototype announcements would move this from vision to active program. This is one of those deals where the headline sounds like satire but the underlying logic is ironclad. The math does not care about social media feuds.
Want to discuss how this applies to your business? Book a free call.
Ready to add AI to your business?
We help businesses identify, design, and deploy AI systems that actually work. Book a free discovery call and see what's possible.
Book a free call →Related Posts
View all posts →AI Chatbots for E-commerce Support: What to Automate First
A practical framework for e-commerce teams deciding what an AI chatbot should handle first: repetitive questions, product discovery, policies, and human handoff.
AI Security Is Now a Buying Issue
Anthropic's Mythos preview and an OpenAI tooling incident say the same thing: buying AI means vetting permissions, connectors, and blast radius, not just the model.
Why a Local Dev Studio Beats a Freelancer for Custom Software
Freelancers are cheap. Agencies are expensive. Boutique dev studios offer something better: senior expertise, direct accountability, and no outsourcing.