Anthropic doubles Claude rate limits with SpaceX compute deal, adds 220,000 Nvidia GPUs

Anthropic Doubles Claude Rate Limits Through SpaceX Partnership

Anthropic announced Tuesday that Claude rate limits are doubling across multiple subscription tiers, marking a substantial capacity expansion for the AI company. The increases in rate limits for Claude Code now provide doubled five-hour usage windows for Pro, Max, Team, and seat-based Enterprise plans, with peak-hours throttling removed entirely for Pro and Max accounts 1

1

. These higher usage limits take effect immediately, addressing demand from paid subscribers who have experienced capacity constraints during high-traffic periods 2

2

.

The Claude Opus API also received significant attention in the announcement, with rate limits raised considerably across all tiers. Tier 1 users, for instance, are seeing a 1500% increase in maximum input tokens per minute and a 900% increase in maximum output tokens per minute 2

2

. The changes are doubled for most users across paid plans, though free-tier subscribers will not see capacity improvements 2

2

.

New Compute Agreement with SpaceX Powers Expansion

The capacity driving these improvements stems from a new compute agreement with SpaceX that grants Anthropic access to all compute resources at the Colossus 1 data center. This facility provides more than 300 megawatts of new capacity and over 220,000 Nvidia GPUs, which will come online within the month 1

1

. The Colossus 1 data center houses H100 and H200 Nvidia GPUs, along with GB200 accelerators, enhancing Anthropic's ability to manage workload management and AI model training more effectively 3

3

.

Anthropic stated the additional AI compute capacity will directly improve service for Claude Pro and Claude Max subscribers, addressing bottlenecks that have constrained usage during peak demand 1

1

. The company has also expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity, though no formal agreement on orbital AI compute has been signed yet 1

1

2

2

.

Anthropic's Expanding AI Infrastructure Network

The SpaceX deal joins an extensive list of compute commitments Anthropic has secured to support its growing demand. The company maintains an up-to-5-gigawatt agreement with Amazon that includes nearly 1 GW of new capacity online by the end of 2026, alongside a 5 GW agreement with Google and Broadcom set to begin coming online in 2027 1

1

. Additionally, Anthropic has established a strategic partnership with Microsoft and Nvidia that includes $30 billion of Azure capacity, plus a $50 billion American AI infrastructure investment with Fluidstack 1

1

.

Anthropic trains and serves Claude on a mix of hardware, including AWS Trainium, Google TPUs, and Nvidia GPUs, demonstrating a diversified approach to compute capacity 1

1

. The company continues to explore opportunities to bring additional capacity online as demand for Claude services grows across both consumer and enterprise segments 1

1

.

International Expansion Targets Enterprise Compliance Needs

Some of Anthropic's upcoming capacity expansions will focus on international expansion, particularly in Asia and Europe. The company's recent collaboration with Amazon includes additional inference capacity in these regions, designed to address enterprise customers in regulated industries that require in-region infrastructure for compliance and data-residency requirements 1

1

2

2

. Anthropic stated it is being deliberate about where it adds capacity, partnering only with democratic countries whose legal and regulatory frameworks can support investments of the relevant scale 1

1

.

The company also reiterated a commitment made earlier this year to cover any consumer electricity-price increases caused by its US data centers, and indicated it is exploring extending that commitment to new jurisdictions as international expansion proceeds 1

1

. This approach signals Anthropic's focus on sustainable growth that balances capacity needs with community impact, particularly as AI infrastructure demands strain local power grids.