We’ve agreed to a partnership with SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API. Effective today, we are: 1) Doubling Claude Code’s 5-hour rate limits for Pro, Max, and Team plans; 2) Removing the peak hours limit reduction on Claude Code for Pro and Max plans; and 3) Substantially raising our API rate limits for Opus models. Our agreement with SpaceX means we will use all the compute capacity at their Colossus 1 data center. This will give us over 300 megawatts of additional capacity to deploy within the month. Read more: https://lnkd.in/eSWRY3dH
Increasing limits is the headline but the real story is the infrastructure shift. Securing a site like Colossus 1 is a massive move to de-risk a supply chain that most of the industry is still struggling to manage. There is always a significant operational risk when scaling this fast, especially regarding the energy demands and deployment timelines, but this is exactly the kind of move required for institutional grade reliability. This type of high stakes execution is right in my wheelhouse and I am genuinely intrigued to see how they handle the deployment because if they can stick the landing it sets a new standard for how we should be thinking about AI stability and long term growth.
This is a major convergence signal. Frontier AI is scaling through compute access, energy capacity, data center infrastructure and strategic infrastructure partnerships. That changes the governance question. AI capability now depends on concentrated physical infrastructure, which means resilience, access, dependency and accountability become part of the same architecture. The future of AI will be shaped not only by who builds the models, but by who controls the infrastructure through which they scale.
Great news! But can you please make Opus smarter again?
RSE Labs
2w
Given Musk’s control of xAI & SpaceX, there’s a real question to be asked. Will Anthropic’s bet on SpaceX come back to haunt them? It certainly enhances their capabilities and places them a step ahead of OpenAI particularly at a time when Musk is suing OpenAI, Altman and Brockman. Seems like a massive win for both Anthropic and SpaceX but is it the right choice looking 1-5 years down the line for Anthropic? In Feb SpaceX and xAI merged. On one hand, the play to monetise data centre infrastructure is smart on SpaceX’s side. If they keep expanding this, they could attempt to position themselves as one of if not the main infrastructure provider for almost all AI Frontier models. On the other hand, if we employ a more cautiously skeptic pov we might raise the question of why now? Why is SpaceX xAI conceding such a competitive advantage to Anthropic in detriment to their own model? A large part of the power is certainly in who controls the infrastructure for the AI Frontier. With SpaceX’s orbital AI data centre plans they seem to be making the ultimate bet on infrastructure. Bringing in Anthropic as the most advanced frontier AI safety research company lends credibility even at the detriment of xAI’s Grok development.
Anthropic Frustrating experience with Anthropic and Claude. We’ve encountered multiple unexpected charges $210, $190, and now over $900 total against claude pro subscription where gift credits appear to be automatically sent and then auto-redeemed, all without clear consent or transparency. What’s even more concerning is the complete lack of support. There has been no response from the official support email, and the in-app agent Fin has been ineffective in resolving the issue. At this point, we’re simply trying to connect with a human, understand what’s going on, and request a refund but we’re getting nowhere.
The strategic value here is compute optionality: no single provider dependency, no single hardware architecture. When compute is the binding constraint on frontier AI, supply chain diversity becomes a competitive advantage in itself. That's what makes this more than a SpaceX headline. Anthropic just assembled one of the most diversified compute portfolios in AI - SpaceX Colossus 1 adding 300 MW within weeks, Amazon delivering up to 5 GW with nearly 1 GW by year-end, Google and Broadcom bringing 5 GW from 2027, $30B in Azure capacity with Microsoft and NVIDIA and a $50B infrastructure investment with Fluidstack. All running across AWS Trainium, Google TPUs and NVIDIA GPUs simultaneously. The immediate payoff is already visible, doubled Claude Code rate limits and Opus API input jumping from 2M to 10M tokens per minute at the highest tier. Does multi-provider compute strategy become table stakes for frontier AI companies?
Kevin Foo
1w
Anthropic is outlining a much broader philosophy for scaling frontier AI — and the central challenge that comes with it. Scaling is increasingly being treated as a governance problem as much as a technical one: geopolitics, regulation, public trust, energy policy, and national infrastructure strategy wrapped into one. Like much of what Anthropic does, the sequencing feels deliberate: compute partnerships, jurisdictions, regulated-market expansion, supply-chain security, energy commitments, and now even orbital compute exploration. The model is only one part of the system now.
The AI race is no longer only about models anymore — it’s now about compute, infrastructure, energy, and scalability. Partnerships like this between Anthropic and Anthropic clearly show where the future is heading. When AI companies begin leveraging space-grade infrastructure and massive data center capacity, it signals a shift from experimental AI to industrial-scale intelligence deployment. The next generation of engineers won’t survive by learning only one tool or one framework. The future belongs to people who understand AI ecosystems end-to-end — cloud, automation, infrastructure, APIs, agents, and scalable systems together. Huge move by Anthropic and SpaceX. The AI infrastructure war has officially entered a new era. ⚡ Himanshu Agarwal #AI #Anthropic #SpaceX #Claude #GenerativeAI
Compute is now the product. Not the model, not the API, raw throughput is what differentiates in 2026. I keep thinking about what "all the compute capacity at Colossus 1" actually means operationally. SpaceX is not a hyperscaler. Colossus was built for Grok. The cooling, networking, and reliability assumptions baked into that facility were not made with Anthropic's workloads in mind. When the first major outage hits, who owns the SLA and what does remediation actually look like for an enterprise customer mid-deployment?