Elon Musk just made the most unexpected AI deal of the year. SpaceX, which now owns xAI, signed an agreement today giving Anthropic full access to Colossus 1, the largest AI supercomputer the company operates. Three hundred megawatts of compute. Two hundred and twenty thousand NVIDIA GPUs. Online within the month. The same Elon Musk who wrote on X in February that Anthropic "hates Western civilization." The same Elon Musk who spent three days last week testifying in federal court trying to dismantle Anthropic's biggest rival OpenAI. The same Elon Musk who runs a company that competes directly with Claude. Today he handed Anthropic the keys to one of the most powerful AI data centers on earth. And the agreement does not stop on the ground. Anthropic and SpaceX are also exploring building gigawatts of AI compute capacity in actual orbit. AI data centers in space. The most bitter rivalry in AI just turned into the wildest partnership in tech.

Here is what Anthropic and SpaceX announced today and why every detail matters.

Three hundred megawatts of compute capacity. That is enough electricity to power more than 300,000 homes. Anthropic gets access to the entire Colossus 1 facility in Memphis, Tennessee. The full computing power. Not a slice. Not a partition. The whole data center.

Two hundred and twenty thousand NVIDIA GPUs. Colossus 1 was built from the ground up at record speed and packs dense deployments of H100, H200, and next generation GB200 accelerators. For context, this is one of the largest concentrations of frontier AI compute in the world. Most labs would kill for a fraction of this capacity. Anthropic just got all of it.

Online within one month. This is the part that should make every Anthropic competitor pay attention. Not next quarter. Not next year. Within thirty days. Anthropic is about to have a massive infrastructure boost while OpenAI is still negotiating with Microsoft and Oracle for additional capacity.

Higher Claude limits starting now. Anthropic is using the additional compute to immediately raise rate caps. Claude Code is doubling its rate limits for paid plans. Pro and Max subscribers are getting peak-hour usage caps removed entirely. Opus request volumes are increasing sharply. Anthropic also announced inference capacity expanding into Europe and Asia for customers in regulated industries who need data stored in specific regions.

Space data centers. Read that again. Anthropic and SpaceX are exploring developing multiple gigawatts of orbital AI compute capacity. Actual data centers in actual space. SpaceX has been working on this concept for months as part of its IPO pitch. Anthropic just became a marquee customer for the orbital compute vision. If this happens, Anthropic will be running Claude on hardware orbiting the planet within a few years.

The "dreaming" announcement. Anthropic also unveiled a new Claude feature today called dreaming. Available as a research preview, it lets Claude review work between sessions, spot patterns, and update files that store user preferences and context. The feature ships alongside software for managing AI agents that perform tasks with minimal human oversight. Anthropic is investing the new compute capacity directly into making Claude smarter at autonomous work.

Because the math finally beat the rivalry.

Elon Musk has been one of Anthropic's most public critics for years. He has compared their stance on AI safety to fragility. He has accused them of being insufficiently committed to American AI leadership. He has championed xAI as the alternative. As recently as February of this year he wrote that Anthropic "hates Western civilization."

Then last week he spent significant time with senior members of the Anthropic team. After those meetings he posted on X that he was impressed. "Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector. So long as they engage in critical self-examination, Claude will probably be good." That post went up before today's announcement and turned out to be the early signal.

The business logic is straightforward. SpaceX is preparing for an IPO. Colossus 1 is a massive capital investment that needs a marquee customer to justify the spending and signal demand to investors. Anthropic is the most credible enterprise AI customer on earth right now with $30 billion in annualized revenue and growing fast. SpaceX gets the IPO story they need. Anthropic gets the compute they need. Both companies win even though their leaders disagree on basically everything.

There is also the rivalry layer. Musk is suing OpenAI for $130 billion. He testified for three days last week trying to revert OpenAI back to a nonprofit. He wants Sam Altman and Greg Brockman removed from the board. The cleanest way to hurt OpenAI is not to compete with them through xAI. The cleanest way to hurt OpenAI is to give their biggest rival the compute they need to overtake them. Today Musk did exactly that.

Some Cool Stuff Worth Noting

While Anthropic was making history with SpaceX, OpenAI was quietly working on something that could be even bigger long-term. According to Ming-Chi Kuo, the most reliable hardware analyst in tech, OpenAI is fast-tracking its first AI agent phone with mass production targeted for early 2027.

Here is what Kuo reported today. The device is being called an AI agent phone, designed specifically to run autonomous AI tasks on-device rather than as a smartphone with AI features bolted on. MediaTek will be the sole processor supplier with a customized Dimensity 9600 chipset built on TSMC's N2P process. The phone features an upgraded image signal processor with enhanced HDR pipeline, specifically optimized for AI applications that rely on continuous environmental awareness.

The strategic context matters. Anthropic just locked in massive ground-based and orbital compute through SpaceX. OpenAI is going the opposite direction. They are building hardware. The early 2027 timeline also lines up neatly with OpenAI's planned IPO narrative. You cannot just be a software company forever when your competitors are vertically integrating into compute, design, and now consumer hardware.

Sam Altman has been hinting at this hardware play for over a year. The Jony Ive collaboration. The acquisition rumors. Today the timeline got real.

If you use Claude, your experience just got significantly better. Claude Pro and Max subscribers are getting peak-hour usage caps removed immediately. Claude Code rate limits are doubling. Opus request volumes are increasing. The compute crunch that has been making Claude slower during peak hours is about to ease substantially.

If you build with the Claude API, the practical implications are large. Higher rate limits mean you can deploy Claude in workflows that previously required expensive workarounds. Inference capacity is expanding into Europe and Asia which matters for regulated industries that need data residency in specific regions.

If you watch the AI race, today is a structural moment. Compute has been the single biggest constraint on Anthropic's growth for the last six months. Demand for Claude has consistently outstripped capacity. Today that constraint just got significantly looser. Anthropic now has compute commitments from Amazon at 5 gigawatts, Google and Broadcom at 5 gigawatts, Microsoft and Nvidia at 30 billion dollars of Azure capacity, Fluidstack at 50 billion dollars, and now SpaceX at 300 megawatts immediately. The total adds up to something nobody else in the AI industry can match.

If you are watching Elon Musk, this is the most strategically interesting move he has made in years. He cannot beat OpenAI through xAI alone. So he is helping their biggest rival beat them while simultaneously suing OpenAI in federal court. The compounding pressure on OpenAI from both sides is the real story of the next year.

Elon Musk and Anthropic announced a partnership today that nobody saw coming. SpaceX is giving Anthropic full access to Colossus 1, 300 megawatts of compute, 220,000 NVIDIA GPUs, online within thirty days. The same Musk who said Anthropic hates Western civilization three months ago. The same Musk who is suing OpenAI for 130 billion dollars in federal court. The deal also includes plans to develop multiple gigawatts of AI data centers in actual space orbit. Anthropic is using the new compute to remove peak-hour caps for Pro and Max subscribers, double Claude Code rate limits, and expand into Europe and Asia. The rivalry between Musk and Anthropic is not over. They both just decided that beating OpenAI matters more. The most unexpected partnership in AI just gave the most demanded AI lab on earth the compute it needed to keep winning.

Anthropic x SpaceX Announcement👉 Here

Claude Managed Agents 👉 Here

Stay building. 🤖