클로드 폭발적 성장, 머스크 슈퍼컴퓨터로 향하다

How Anthropic's 80x growth blew past its own infrastructure and straight into Musk's data center Maximilian Schreiner View the LinkedIn Profile of Maximilian Schreiner May 7, 2026 Midjourney prompted by THE DECODER Key Points Anthropic has struck a deal to use Elon Musk's Colossus 1 supercomputer, gaining access to roughly 220,000 Nvidia GPUs — a remarkable shift after Musk called the company "evil" earlier this year. The partnership is driven by Anthropic's explosive growth: revenue has surged from $9B to over $30B annualized, with demand far outpacing infrastructure and causing rate limits and outages for users. For SpaceXAI, which plans to go public soon, a paying enterprise customer helps justify Colossus 1 economically. Ask about this article… Search Anthropic is set to tap into Elon Musk's Colossus 1 supercomputer. Behind the surprise deal lie a compute crunch, a looming IPO, and a remarkable about-face by Musk. At its developer conference in San Francisco, Anthropic announced an agreement with Musk's SpaceXAI to use the Colossus 1 data center in Memphis. According to Wired , the deal gives Anthropic access to more than 300 megawatts of capacity and roughly 220,000 Nvidia GPUs. The New York Times reports that Anthropic is taking all of Colossus 1's computing capacity. But more interesting than the raw numbers is why this deal is happening at all. Musk went from calling Anthropic "evil" to signing a deal Earlier this year, Musk called Anthropic "misanthropic" and "evil" on X, claiming without evidence that the company's models showed racial and sexual biases. The tone has shifted dramatically since then. Musk now writes on X that he "spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed." Ad The change of heart lines up with a business reality. SpaceXAI, the company formed in early 2026 by the merger of xAI and SpaceX, is planning to go public as soon as next month. At this stage, a paying enterprise customer is worth more than a rhetorical adversary. Colossus 1 was built in record time and is primarily intended for Grok, the company's in-house model. The fact that a direct competitor is now moving in suggests Grok alone can't economically justify the facility. According to Musk, Colossus 2 is currently being used to train the next generation of Grok. Ad DEC_D_Incontent-1 Anthropic's explosive growth is outpacing its own infrastructure Why Anthropic is grabbing every available chip became clear when CEO Dario Amodei spoke at the same conference. Anthropic had planned for roughly tenfold growth this year, only to find itself on a trajectory that could make it 80 times bigger, Amodei said, according to the NYT. "I hope that 80-times growth doesn't continue because that's just crazy and it's too hard to handle." Anthropic's annualized revenue run rate has climbed from $9 billion at the end of 2025 to more than $30 billion. A key driver is Claude Code: the average developer now spends at least 20 hours per week with the tool, according to Anthropic. In recent months, complaints about rate limits and outages have piled up because there simply wasn't enough compute to go around. The SpaceXAI deal is meant to raise usage limits for Claude Pro and Claude Max subscribers. Ad For SpaceXAI, Anthropic also serves as a marketing asset. In a blog post , the company writes that Anthropic has "expressed interest" in partnering on "orbital AI compute capacity" - data centers in space. Ahead of the IPO, that's a useful narrative. Ad DEC_D_Incontent-2 AI News Without the Hype – Curated by Humans Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section. Subscribe now Source: Wired / Anthropic SpaceX Deal | New York Times / Anthropic Growth | Elon Musk / X Post | xAI / Anthropic Compute Partnership