Claude Code Limits Doubled - Anthropic SpaceX 220K GPU Deal (2026.05)

Claude limits doubled overnight after SpaceX deal
▲ Claude limits doubled overnight after SpaceX deal

Claude usage limits are the 5-hour rolling caps Anthropic applies to its paid plans (Pro, Max, Team, Enterprise) for Claude Code and the Claude API. As of May 6, 2026, those caps just doubled. Anthropic leased SpaceX's Colossus 1 data center in Memphis - roughly 220,000 NVIDIA GPUs and over 300 megawatts of capacity - in a single deal.

What exactly changed in Claude limits

Anthropic's announcement covers three changes that took effect the same day. Claude Code's 5-hour usage caps doubled across all paid tiers - Pro, Max, Team, and seat-based Enterprise. The peak-hours throttle that quietly cut Pro and Max usage during daytime is now removed. Claude Opus API rate limits were also raised significantly, which directly benefits enterprise developers running heavy workloads.




Three changes: 5-hour 2x, peak removed, Opus API up
▲ Three changes: 5-hour 2x, peak removed, Opus API up

The SpaceX-Anthropic compute deal behind it

Looser limits required raw compute. Anthropic leased SpaceX's entire Colossus 1 data center in Memphis: roughly 220,000 NVIDIA GPUs (H100, H200, GB200 mix) and over 300 MW of capacity, online within a month. The twist: two months earlier, Elon Musk publicly attacked Anthropic on X, claiming the company "hates Western civilization." Now Musk is renting them his crown-jewel site. The two firms are also exploring multi-gigawatt orbital AI compute together.




Workflow headroom doubles for Claude Code users
▲ Workflow headroom doubles for Claude Code users

What it means for daily Claude users

Claude Code developers will stop hitting the cap mid-afternoon. Pro and Max subscribers get full daytime speed instead of silent throttling. Enterprise teams on Opus API gain meaningful throughput for long-context, document-heavy tasks. CEO Dario Amodei framed the deal as a way to "increase usage limits for Pro and Max subscribers."

The fine print: weekly cap stays the same

One detail is easy to miss. The 5-hour rolling window doubled, but the weekly cap was not increased. Anthropic's announcement makes no mention of weekly cap changes. Heavy users who burn through their weekly budget will hit the same wall - they just get more headroom inside any single session. Read carefully if you're a power user.

Key Takeaways

① 5-hour caps doubled - Claude Code's rolling window cap is 2x across Pro, Max, Team, Enterprise, effective immediately.

② SpaceX nets 220K GPUs - Colossus 1 in Memphis: 300MW and 220,000 NVIDIA GPUs online within a month.

③ Weekly cap unchanged - Only the 5-hour window grew; weekly budgets stayed flat, heavy users still need to manage spend.

The bigger story is the AI compute arms race. Anthropic also locked in up to 5GW with AWS, 5GW with Google/Broadcom, $30B Azure via Microsoft/NVIDIA, and $50B with Fluidstack. Whoever controls the most electricity and GPUs gets to write the next chapter of AI.

👉 Apple iOS 27 Extensions: Siri Will Run Claude or Gemini This Fall (2026) - also worth a read.

📌 Sources: Anthropic Official, Bloomberg, CNBC, TechCrunch (2026)

위치: 미국 뉴욕

댓글

댓글 쓰기

이 블로그의 인기 게시물

SK Hynix US ADR Listing 2026 — $10.5B SEC Filing to Close Micron Valuation Gap

이미지
▲ SK Hynix US ADR listing announced March 2026 An ADR (American Depositary Receipt) is a US-traded certificate representing shares in a foreign company, allowing American investors to buy international stocks directly on US exchanges. On March 25, 2026 , SK Hynix — the world's leading HBM (High Bandwidth Memory) chipmaker — confidentially submitted Form F-1 to the U.S. Securities and Exchange Commission (SEC), formally initiating its US ADR listing process. This marks a historic step for Korea's semiconductor industry and signals SK Hynix's ambition to compete for global capital on equal footing with US chip rivals. Why Is SK Hynix Pursuing a US ADR Listing Now? The answer lies in a stark valuation gap. SK Hynix currently trades at a P/E (price-to-earnings) ratio of approximately 5.9× on the Korea Stock Exchange (KOSPI). Micron Technology, its closest US competitor, trades at over 18× — despite SK Hynix holding an acknowledged lead in HBM3E and HBM4 technology for...
자세한 내용 보기

Nvidia $3.2B Corning Bet: AI Data Center Optical Fiber Megadeal (2026)

이미지
Nvidia bets $3.2 billion on Corning to rewire AI data centers The Nvidia Corning deal is a $3.2 billion investment in which Nvidia gains warrants to buy up to 15 million shares of Corning, in exchange for three new US factories that will replace copper cabling inside AI servers with optical fiber. The agreement is the clearest signal yet that the next phase of the AI buildout will be won by interconnect, not just by GPUs. For anyone using ChatGPT, Gemini or Claude, this deal points to faster responses and lower power-bill pressure over the next few years. Why Nvidia Is Betting on a 175-Year-Old Glass Maker A single Nvidia AI rack such as the upcoming Vera Rubin system contains roughly 5,000 copper cables wiring GPUs together. Copper has hit a wall: signal loss climbs with distance, and the heat and power draw grow non-linearly as model sizes explode. Light is the answer. Optical fiber moves data faster and uses far less power. The technology Nvidia is now hardwiring into its...
자세한 내용 보기

Alibaba Zhenwu M890 - 3x Nvidia H20 Performance, China AI Chip 2026

이미지
▲ Alibaba Zhenwu M890 targets Nvidia H20 Alibaba Zhenwu M890 is a domestic AI accelerator unveiled by Alibaba's chip subsidiary T-Head. Designed for both training and inference workloads, the chip is the clearest evidence yet that China's semiconductor independence push is moving from slideware to silicon. For US readers, the consequences land in two places that matter: cloud AI pricing and the global memory and semiconductor-equipment trade. Why China's Own AI Chip Matters Now Since 2022, US export controls have blocked Nvidia's top-tier H100-class GPUs (graphics processing units, the high-end chips that power AI training) from being sold into China. Even H20 - the dialed-down chip Nvidia built specifically for the China market - was eventually cut off. Blocked imports pushed Beijing's biggest tech companies straight into mass-producing domestic alternatives. Alibaba, Huawei, and Cambricon have all moved into volume production, and the Zhenwu M890 is the la...
자세한 내용 보기