Anthropic announces doubling of Claude usage, reaching orbital AI computing power agreement with SpaceX

Core conclusion

On May 6, 2026, Anthropic issued an announcement announcing two major actions: first, doubling the conversation limit of Claude free users every 5 hours; second, announcing a giant data center computing power cooperation plan with xAI, and expressing its intention to jointly develop orbital AI computing power with SpaceX. For content creators and AI practitioners, this means:

• Claude usability improvements: Doubling the usage limit means heavy users can rely on Claude for longer to complete automated workflows
• Computing Power Reserve Expansion: The new capacity of 300 MW and over 220,000 NVIDIA GPUs is expected to accelerate Claude model iteration
• Signal for orbital computing: SpaceX’s Starlink network + orbital data center may become the forefront of future AI inference

Key Points

• Event time: 2026-05-06 -Affected objects: Claude users, AI automation practitioners, content creators
• Core changes: Free user 5h quota doubled + 300MW GPU computing power cooperation + Orbital AI computing power vision

Background and trigger events

Anthropic has been actively expanding its computing power reserves in recent years to support the rapid growth of the Claude model series. From Claude 3 to Claude Code to Claude Codex, the reasoning requirements of the model have increased exponentially. This announcement is not only a response to user pain points, but also a concentrated elaboration of Anthropic’s computing power strategy.

Instead of relying solely on AWS and Google Cloud, Anthropic chose to work with xAI (Elon Musk's AI company) to use its Colossus data center - a supercomputer in Memphis, Tennessee that was previously built specifically for Grok model training. This cooperation is quite dramatic: Anthropic and xAI founders Musk and Altman are in a tug-of-war in litigation, but the demand for computing power has turned "enemies" into partners.

Key Impact (by Dimension)

Adaptation suggestions

If you use Claude for automated content production

The most direct beneficiaries of this quota doubling are heavy users who rely on Claude API and Claude Code on a daily basis. The following are executable adaptation strategies:

1. Extend the duration of a single Agent session: The original one-hour Claude Code automation task can now be extended to two hours for uninterrupted execution.
2. Try more complex workflow orchestration: Multi-step Agent tasks (such as collection → processing → generation → publishing full link) can now be completed in the same session
3. Reduce the maintenance cost of sharding strategy: Previously, large tasks had to be split into multiple small tasks due to quota restrictions, but now they can be merged

If you are concerned about AI computing power cost

Anthropic’s computing power investment will ultimately be reflected in the API price. However, in the short term, more abundant computing power means optimization of model inference efficiency rather than inevitable price reduction. Comparison strategy:

• For high-throughput scenarios: continue to use cost-effective solutions such as DeepSeek V4 Pro as a supplement to Claude
• Quality-sensitive scenarios: Claude's quota increase makes more "one-time generation of high-quality content" tasks feasible

Task List (Reference)

• Check quota utilization of existing Claude Code workflows and re-evaluate whether tasks need to be split
• Test the long-term running stability of Claude Code under doubled quota
• Pay attention to Anthropic’s subsequent API update announcements

Extended thinking: the possibility of orbital AI computing power

Anthropic's mention in the announcement of "collaborating with SpaceX to develop multi-gigawatt orbital AI computing power" is not a gimmick. Spatial computing has several potential advantages:

• Sufficient solar energy: The solar energy efficiency in space is 5-8 times that on the ground, 24 hours a day
• Heat dissipation advantage: The heat dissipation design in a vacuum environment can break through the heat density limitations of ground data centers
• Low latency inter-satellite communication: The Starlink constellation already has inter-satellite laser links and may become part of the AI inference network in the future

Of course, this plan is still in the "expressions of interest" stage and is still years away from actual deployment.

Tool entry

AI tools and services covered in the text: Claude, Claude Code, Anthropic, DeepSeek, xAI, ChatGPT, OpenAI, SpaceX, OpenClaw, Hermes Agent.

Internal link guidance

• Want to experience the actual productivity of Claude Code? Watch: Claude Code automated writing practice: build an AI content production pipeline in 30 minutes
• If you think Claude API is expensive, try this solution: DeepSeek V4 vs Claude Code: 90% Cheaper, Same Quality
• Real case: Someone used Claude Code to achieve monthly income in 48 hours $9,000: Claude Code 48 hours to start a business: one person + US$29 monthly fee, monthly income in 3 months $9,000