Infrastructure · Models

Anthropic partners with SpaceX, doubles Claude Code limits, and puts orbital AI compute on the roadmap

· 7 min read

Anthropic announced three immediate changes to Claude's usage limits, a new compute partnership with SpaceX, and — buried in the announcement — an expression of interest in building "multiple gigawatts of orbital AI compute capacity." The rate limit improvements are the headline for users today. The SpaceX deal is the headline for infrastructure watchers. The orbital compute line is the one that changes the conversation entirely.

What changed for Claude users today

Three changes, all effective immediately, apply to paid plans:

1. Claude Code rate limits doubled. The five-hour rate limits on Claude Code are doubling for Pro, Max, Team, and seat-based Enterprise plans. For heavy users of Claude Code — and Anthropic disclosed yesterday that 100% of Walleye Capital's 400-person hedge fund uses it — this meaningfully reduces the friction of hitting a wall mid-session.

2. Peak hours penalty removed. Anthropic previously reduced Claude Code limits during peak usage hours. That reduction is gone entirely. No more watching the clock to decide when to start a coding session.

3. Higher API rate limits for Opus models. The Claude Opus tier — Anthropic's flagship for complex reasoning — is getting "considerably" higher API rate limits. The company published an updated rate limit table alongside the announcement, though it did not specify exact numbers in the blog post.

These changes are the direct result of a compute capacity expansion that Anthropic has been assembling over the past month through a series of deals. The latest — and most attention-grabbing — is SpaceX.

The SpaceX partnership

Anthropic has agreed to a partnership with SpaceX that will "substantially increase our compute capacity." The announcement did not include dollar figures or gigawatt commitments, but described it as part of a series of deals that together unlock the headroom to raise limits across the board.

Then came the line that made infrastructure analysts sit up:

"As part of this agreement, we have also expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity."

Orbital AI compute. Data centers in space. This is not a near-term deployment — it is an intention, a vector — but it signals that Anthropic is thinking about compute constraints on a timeline that extends well beyond terrestrial power grids. SpaceX brings launch capacity and, via Starlink, the satellite infrastructure that could eventually connect orbital nodes to ground stations. The partnership is compute-focused today; the orbital piece is a long bet on where the bottleneck moves next.

The wider compute buildup

The SpaceX deal is the latest in a rapid-fire series of infrastructure announcements from Anthropic over the past month:

Anthropic trains and runs Claude on a mix of hardware — AWS Trainium, Google TPUs, and NVIDIA GPUs — and the company says it continues to "explore opportunities to bring additional capacity online."

The combined picture is striking: across Amazon, Broadcom, Microsoft/NVIDIA, Fluidstack, and now SpaceX, Anthropic has assembled a multi-vendor, multi-architecture compute pipeline that hedges against any single supplier bottleneck. No other frontier AI lab has publicly disclosed a comparably diversified hardware strategy.

International expansion — with conditions

Some of the new capacity will be deployed internationally. Anthropic's enterprise customers in regulated industries — financial services, healthcare, government — increasingly require in-region infrastructure for compliance and data residency. The company's recently announced international expansion includes additional inference capacity in Asia and Europe.

But Anthropic drew a line: it will only partner with "democratic countries whose legal and regulatory frameworks support investments of this scale, and where the supply chain on which our compute depends — hardware, networking, and facilities — will be secure." This is an explicit political commitment. It rules out certain jurisdictions and signals to enterprise customers that their data will not transit through regimes with conflicting legal standards.

The electricity commitment

Anthropic also reiterated a commitment to cover any consumer electricity price increases caused by its data centers in the United States. As part of international expansion, the company is exploring ways to extend that commitment to new jurisdictions and to "partner with local leaders to invest back into the communities that host our facilities."

This matters. Data center power consumption is becoming a political flashpoint. Anthropic getting ahead of it with a direct consumer-cost guarantee is both good PR and a genuine structural commitment that other labs have not matched.

Why this matters

Three things make this announcement significant beyond the rate limit changes:

1. The compute land grab is accelerating. Anthropic has now announced deals with Amazon, Broadcom, Microsoft, NVIDIA, Fluidstack, and SpaceX in the span of roughly one month. The total committed capacity — if all deals reach their stated targets — would place Anthropic among the largest compute buyers in the world. This is not a research lab scaling up. This is an infrastructure company being built inside an AI company.

2. SpaceX is now in the AI supply chain. SpaceX's entry as an AI infrastructure partner is new. Elon Musk's xAI competes directly with Anthropic, making this a notable cross-competitive arrangement. SpaceX provides launch and satellite infrastructure; the orbital compute interest suggests a deeper, longer-term collaboration that could reshape where training and inference clusters physically live.

3. The rate limits are a leading indicator of capacity. When Anthropic raises limits, it means capacity came online. When it removes peak-hour penalties, it means capacity is exceeding peak demand. These are the closest thing to real-time infrastructure health signals the public can observe from an AI lab. Today's changes suggest Anthropic's compute pipeline is delivering ahead of schedule.

All changes are effective immediately. Updated rate limits are available in Anthropic's documentation. The SpaceX partnership and orbital compute interest were disclosed in Anthropic's May 6, 2026 announcement.