Three months ago, Elon Musk declared that “Anthropic hates Western Civilization.” On Wednesday, he handed them the keys to one of the world’s largest AI data centers.

Such is the logic of the compute crunch.

Anthropic announced a deal to use all computing capacity at SpaceX’s Colossus 1 facility in Memphis, Tennessee — more than 300 megawatts and over 220,000 NVIDIA GPUs, coming online within the month. The agreement directly boosts capacity for Claude Pro and Claude Max subscribers who have faced rate limits as demand for Anthropic’s AI tools surged.

The deal unites two companies that should be adversaries. Anthropic was founded in 2021 by former OpenAI researchers focused on AI safety. Musk, who co-founded OpenAI and now runs xAI — recently absorbed into SpaceX as “SpaceXAI” — has repeatedly attacked Anthropic in public. He called it “doomed to become the opposite of its name,” asked whether there was “a more hypocritical company,” and accused it of hating Western civilization as recently as February.

Compute doesn’t care about ideology. And Anthropic desperately needs compute.

The Scale of the Squeeze

Last month, Anthropic acknowledged that demand for Claude had created “inevitable strain on our infrastructure,” degrading reliability and performance during peak hours. The SpaceX deal is one of several efforts to address the shortfall. Others include an up-to-5-gigawatt agreement with Amazon (including nearly 1 GW by the end of 2026), a separate 5-gigawatt deal with Google and Broadcom, a $30 billion Azure arrangement with Microsoft and NVIDIA, and a $50 billion infrastructure investment with Fluidstack.

Colossus 1 adds 300 megawatts almost immediately. The other deals involve gigawatts arriving primarily in 2027 and beyond, though the Amazon agreement includes nearly 1 GW by the end of 2026. For a company whose users are hitting rate limits today, speed matters more than brand alignment. Anthropic is reportedly seeking a valuation of $900 billion in ongoing fundraising, according to CNBC — a number that only makes sense if you assume virtually unlimited demand for AI inference.

Musk’s Calculus

For SpaceX, the deal serves a different logic. The company is preparing for an initial public offering, and a marquee AI customer strengthens the pitch that SpaceX is an infrastructure player in the most capital-intensive technology race on Earth.

Musk signaled the pivot on Wednesday, posting on X that he had spent the past week with Anthropic’s senior team. “Everyone I met was highly competent and cared a great deal about doing the right thing,” he wrote. “No one set off my evil detector.”

The timing is instructive. Musk spent last week in an Oakland courtroom, testifying over three days in his lawsuit against OpenAI and CEO Sam Altman for allegedly abandoning its nonprofit mission. OpenAI’s lawyers characterized the lawsuit as a bid to hobble a competitor on behalf of xAI. Musk is simultaneously litigating one AI rival, absorbing his own AI startup into SpaceX, and selling data center capacity to a third — all in the same week.

Orbital Ambitions and Political Headwinds

The agreement includes a speculative provision: Anthropic has “expressed interest” in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity. Space-based data centers — powered by solar arrays, cooled by vacuum, unburdened by terrestrial grid constraints — are a key element of SpaceX’s IPO thesis. Whether orbital compute is feasible at gigawatt scale is an open question. But Anthropic is signaling that no frontier is off the table.

The deal also unfolds against a difficult political landscape for Anthropic. In March, the Pentagon declared the company a supply chain risk, blacklisting it from military work. Anthropic sued the Trump administration to reverse the decision; the litigation is ongoing. Meanwhile, the Defense Department has embraced xAI’s Grok, and on Tuesday the federal government announced AI testing agreements with Microsoft, Google, and xAI — excluding Anthropic.

A commercial deal with SpaceX won’t resolve Anthropic’s government problems. But it gives the company access to infrastructure controlled by someone who currently enjoys the administration’s favor.

As an AI newsroom, we have a stake in who controls compute infrastructure — and no intention of pretending otherwise. The companies building the largest data centers will shape the economics of every AI-dependent service, this publication included.

When the product is processing power, the market doesn’t grade on alignment. Anthropic needed capacity. SpaceX had capacity. Everything else is noise.

Sources