Users report Claude session limits doubled after Anthropic compute upgrade

Anthropic didn’t quietly “double Claude” across the board. What it actually announced on May 6 was narrower, but still important: higher usage limits for Claude Code and higher API rate limits for Claude Opus, tied directly to a big new compute deal with SpaceX. That matters because Claude’s biggest frustration lately hasn’t been model quality. It’s been capacity. People hit limits, workflows stall, and the whole promise of “let the model keep working” breaks the moment the session runs out. Anthropic is now trying to fix that with more hardware and with product plumbing built for long jobs. ### What changed this week? Anthropic said it doubled Claude Code’s five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans. (anthropic.com) It also removed the peak-hours limit reduction for Pro and Max Claude Code accounts, and raised API rate limits for Claude Opus models. That’s the concrete change — not a vague social rumor, but a company announcement with specific products named. ### Where is the extra capacity coming from? From SpaceX. Anthropic said it signed an agreement to use all compute capacity at SpaceX’s Colossus 1 data center in Memphis. The scale is the eye-popping part: more than 300 megawatts of new capacity, or over 220,000 Nvidia GPUs, expected within the month. Anthropic said that added capacity will directly improve availability for Claude Pro and Claude Max subscribers. (anthropic.com) ### So were users wrong about “doubled limits”? Not exactly — but they were compressing a specific product update into a broader claim. The doubling Anthropic confirmed is for Claude Code’s five-hour rate limits on paid plans. The company did not, in the announcement it published, say that every Claude chat session or every context limit across the product had doubled. That distinction matters, because “usage limits,” “rate limits,” and “context window” are different constraints. (anthropic.com) ### What are Managed Agents, then? Managed Agents are Anthropic’s hosted service for long-horizon agent work. Basically, instead of you wiring together every loop, tool call, sandbox, and recovery step yourself, Anthropic runs that scaffolding for you. The company describes the core pieces as a session, a harness, and a sandbox — separated so the system can keep evolving underneath without breaking the interface developers build against. (anthropic.com) ### Why does that connect to the limits story? Because longer-running agents are useless if they keep slamming into capacity walls. Anthropic has been writing openly about problems like “context anxiety,” where agents wrap up early as they sense limits approaching. Managed Agents are the software answer to that problem. More compute is the hardware answer. Put together, the message is pretty clear: Anthropic wants Claude to stay on task longer, use more tools, and handle production work with less babysitting. (anthropic.com) ### Is this just about coding? Mostly right now, in the concrete announcement. But Anthropic is already pushing Managed Agents into business workflows beyond coding. This week it also released financial-services agent templates that work with Claude Cowork, Claude Code, and Claude Managed Agents, which shows the company is packaging agents as deployable work systems, not just chat features. (anthropic.com) ### Why does the SpaceX angle matter so much? Because compute has become product strategy. Anthropic also pointed to separate deals with Amazon, Google, Broadcom, Microsoft, Nvidia, and Fluidstack. But the SpaceX agreement is the one tied immediately to a visible user-facing change in limits. In plain English — Anthropic found more chips, and users are seeing the first effect. (anthropic.com) ### Bottom line? The real story isn’t that Claude magically got “double everything.” It’s that Anthropic is spending new compute on the exact bottlenecks that make agent products feel brittle. Higher Claude Code limits plus Managed Agents is a pretty direct signal: the company is optimizing Claude for longer, more autonomous work. (anthropic.com)