Anthropic Inks Massive SpaceX Compute Deal Ahead of $2 Trillion IPO

The AI race continues to heat up, as NVIDIA announces its $81.6 billion revenue report for the first quarter of its fiscal 2027 period, and OpenAI expected to file for its own IPO in the near future. Now, Anthropic and SpaceX have announced the two companies are bolstering their partnership by increasing its GB200 capacity in Colossus 2.

If you're a Claude user who has been hitting rate limits mid-session, here's your explanation, and your fix. Anthropic has struck one of the largest compute procurement deals in the history of the AI industry, signing a contract with SpaceX to take over the full capacity of the Colossus 1 supercomputer cluster in Memphis, Tennessee. Then, just as SpaceX filed its IPO paperwork with the SEC on May 20, Anthropic further expanded the deal. Anthropic co-founder and chief compute officer Tom Brown confirmed on X that the company is "expanding our partnership with SpaceX, and will be scaling up on GB200 capacity in Colossus 2 throughout June," adding that the team at SpaceX has been helping Anthropic "find good homes for the Claudes." That's a stark contrast to how the two AI companies' chief executives have been lobbing public insults at each other for months.

$1.25 Billion a Month

The price tag buried in SpaceX's S-1 filing is perhaps the real headline. Anthropic will pay SpaceX $1.25 billion per month through May 2029, with reduced fees during a capacity ramp-up in May and June 2026, a contract worth nearly $45 billion over its full term. Either party can walk away with 90 days notice, which provides some flexibility but also means SpaceX is carrying a sizable revenue line item that could theoretically evaporate before the IPO roadshow wraps up.

In exchange, Anthropic gets access to more than 300 megawatts of compute capacity and over 220,000 NVIDIA GPUs, with the original deal covering Colossus 1 and the expansion now pulling in Colossus 2's next-generation GB200 hardware.

What Claude Users Actually Get

This isn't just a financial story, it has immediate practical implications for anyone running Claude workloads. Anthropic confirmed the compute will be used to directly improve capacity for Claude Pro and Claude Max subscribers, and the company has already doubled Claude Code's five-hour rate limits across Pro, Max, Team, and Enterprise plans, while also removing peak-hour limit reductions and substantially raising API rate limits for Opus models.

The rate-limit frustration among power users had become a genuine friction point. The Colossus deal appears to be Anthropic's most direct response yet to the infrastructure gap that was holding its products back. This comes at a good time, as some users had been moving their workload over to OpenAI's Codex, which was sweetening the pot with its own deals.

The Sordid History Between The Two

The politics here are almost as interesting as the technology. Elon Musk spent much of the past year publicly characterizing Anthropic in unflattering terms, calling it a threat to Western civilization as recently as February. By the time the deal was announced in early May, Musk had reversed course entirely, posting on X that he had spent significant time with Anthropic's senior leadership and was "impressed," writing that "everyone I met was highly competent and cared a great deal about doing the right thing."  What changed? Likely finances. SpaceX's S-1 filing describes the Anthropic deal as a way to "monetize unused compute capacity in its infrastructure," and the filing notes the company expects to sign additional similar agreements going forward. Usage of Grok, xAI's flagship assistant, has dropped significantly in recent months, freeing up servers that SpaceX is now leasing to one of its closest competitors. That's a brutally practical business decision, and it makes perfect sense viewed through the lens of a company trying to demonstrate revenue diversification before an IPO. And make no mistake, as Musk points out in his post on X (seen above), SpaceX is not looking to have a monogamous relationship with Anthropic.

The Infrastructure At Play

This is where the deal gets genuinely interesting from a tech-industry perspective. SpaceX isn't just a rocket company anymore, it is positioning itself as a neocloud provider, leasing compute to outside AI labs while simultaneously running its own AI models. At the beginning of this month, Elon Musk announced xAI would be dissolved and absorbed into SpaceX, furthering the intent on joining the SpaceX program with the AI side of things. The partnership with Anthropic also includes both companies exploring the development of multiple gigawatts of orbital AI compute capacity, data centers in space, powered by solar panels and cooled by the void. The orbital compute angle suggests SpaceX sees Starlink's low-earth-orbit infrastructure as the foundation for something much larger than internet access.

For investors evaluating the June 12 NASDAQ debut, the Anthropic contract is seen by some as a double-edged data point. On the positive side, the deal accounted for 76% of SpaceX's $10.1 billion in Q1 2026 capital spending, big enough to move the needle on revenue. Anthropic's own run-rate revenue has now surpassed $30 billion, meaning it is both willing and able to sustain a $15 billion annual compute bill. That makes the contract more durable than it might otherwise appear.

On the other hand, that 90-day termination clause is the kind of thing that makes CFOs nervous right before a public offering. Starlink, which was earlier accused of accounting for about 35% of the space sector’s climate impact in 2020 and could rise to 42% by 2029, remains the only SpaceX division generating consistent operating profit, and the AI segment posted a $2.47 billion operating loss in Q1 2026 alone, a deficit the Anthropic payments are now directly subsidizing.

What the deal ultimately signals is that the lines between AI software companies and hard infrastructure providers are blurring faster than anyone predicted. Anthropic needs the GPUs. SpaceX needs the revenue. And both companies need the market to believe their valuations are justified before their respective IPOs land. For now, at least, they need each other if either is to keep up with the Jones in the AI race.