Claude Opus 4.7 + SpaceX Compute — Usage Limits Doubled

1. SpaceX × Anthropic — The Compute Move That Changes Everything

Before May 6, Pro / Max / Team users running Claude Code heavily were hitting "You've reached your usage limit. Please wait ... hours." almost daily — particularly during peak hours (08:00–17:00 UTC, when US markets are active). Compute capacity simply wasn't keeping pace with demand.

The SpaceX deal isn't a small one. It pulls the entire Colossus 1 data center (originally part of xAI's plan to build "the world's largest AI training cluster") into Anthropic's stack:

Spec	Detail
Power	Over 300 MW (roughly the consumption of ~250,000 households)
GPU count	More than 220,000 NVIDIA GPUs (mix expected to include H100 / H200 / Blackwell)
Timeline	"Within the month" — coming online May–June 2026
Future	SpaceX + Anthropic exploring "multiple gigawatts of orbital AI compute capacity" (compute in space)
Existing stack	Amazon (up to 5 GW), Google (5 GW), Microsoft Azure ($30B NVIDIA deal)

The strategic point — Anthropic isn't just "asking for more GPUs." It's diversifying its compute supply chain so it isn't dependent on a single hyperscaler. This mirrors how international firms approach ERP infrastructure (see Anthropic Enterprise AI Firm $1.5B, where diversification was a central theme).

2. Claude Opus 4.7 — Real Improvements, but Where Do They Show Up?

Opus 4.7 went generally available on April 16, 2026, with pricing held flat at $5 / $25 per million tokens (input/output). The improvements land squarely on enterprise-relevant work:

Dimension	Opus 4.7 vs prior version	Where it shows up
Software engineering	+13% on a 93-task coding benchmark vs Opus 4.6 — handles "long-running tasks with rigor and consistency"	Refactoring ERP modules, writing new integrations
Vision	Supports images up to 2,576 px on long edge (~3.75 MP) — 3× higher resolution than prior models	Reading infographics, schema diagrams, UI screenshots
Long-running tasks	Better file system memory across sessions + sustained reasoning	Month-end close, KYC, multi-day audit work (see Claude Finance Agents)
Safeguards	Auto-blocks high-risk cyber requests; Cyber Verification Program for legitimate pen-testing	Reduces security misuse risk inside organizations

Notable — price did not increase, but Anthropic absorbed the additional compute cost from the SpaceX deal to support it. This signals a strategy of "capture market share before raising prices."

3. Usage Limits Doubled — Who Gets What?

The headline change for actual users: five-hour Claude Code rate limits doubled across all paid plans, with peak-hour throttling removed for Pro and Max:

Plan	5-hour Rate Limit	Peak-hour Throttle	Best fit
Pro	×2 (doubled)	Removed	Solo developers, freelancers, technical writers
Max	×2 (doubled)	Removed	Power users, AI researchers, advanced workflow builders
Team	×2 (doubled)	No specific change announced	Teams of 5–50 — see Claude Team Premium
Enterprise (seat-based)	×2 (doubled)	No specific change announced	Larger organizations with SSO, audit log, governance needs
API (Opus)	Considerably raised (specific numbers not detailed)	—	Product teams building integrations into ERP and other systems

Practical impact — teams previously hitting limits daily on Pro should now get through. Max users who used to face heavy throttling during 14:00–22:00 Bangkok time can now run at full capacity all day, eliminating the need to schedule heavy work for late-night sessions.

Importantly, this isn't "raise the price, give more" — it's "same price, give more." Pro at $20/month, Max at $100 or $200/month, Team Premium at $100/seat/month (annual billed) — all see ROI improve immediately without users doing anything.

4. Claude Managed Agents — What Multiagent Orchestration Means

In the same week, Anthropic also released three new features for Claude Managed Agents, opening up a new pattern for enterprise work:

Feature	What it does	Clear use case
Multiagent Orchestration	A lead agent breaks the job into pieces and delegates each to a specialist (with its own model, prompt, and tools), running in parallel on a shared filesystem	Month-end close: 1 lead + (GL, AR, AP, inventory, tax) specialists working simultaneously
Outcomes	Write a rubric describing what success looks like; the agent works toward it, with a separate grader evaluating results	KYC report scoring against a 12-point checklist — pass/fail, not "did the agent feel done"
Dreaming (research preview)	Scheduled process that reviews session and memory stores, extracts patterns, curates memory so agents improve over time	Accounting agent learns each customer's deduction patterns automatically

One detail: multiagent orchestration requires substantially more compute (one orchestrator running 5 specialists in parallel = 5× the token rate at the same time). It is no coincidence that this feature shipped alongside doubled limits and the SpaceX deal — Anthropic needed the capacity in place before opening this pattern up at GA. For more detail, see Claude Managed Agents.

5. Claude Code v2.1.126 + IDE Plugins — Real Desktop Improvements

Beyond the rate limit changes, Claude Code (CLI/desktop) shipped a release in the same window with features developers use daily:

• Plugin URL loading — load plugins directly from a URL, accelerating custom-tool deployment
• Improved MCP server handling — more stable behavior when running multiple MCP servers concurrently (see Claude MCP)
• Strengthened permission controls — IDE integrations and CLI environments are now more secure
• Microsoft 365 integration — Excel/PowerPoint/Word add-ins now GA, Outlook in public beta, with conversation context preserved across apps

6. Implications for Thai SMEs / Enterprises — Three Angles to Assess

Before pivoting your AI strategy on this announcement, run it through three angles:

Angle	Question to answer	Example outcome
1. ROI on existing subscription	How often did our team hit limits per week on Pro/Max? Has it dropped?	If you used to hit 3×/week and now hit 0 — no need to upgrade plans
2. Opportunity to expand scope	Which workloads were "too heavy" for Claude before — can they fit now?	Whole-repo code review, large module refactors, end-of-month report generation
3. Multi-agent in ERP workflows	Which processes naturally fit a "lead + specialist parallel" pattern?	Month-end close, vendor onboarding, internal audit, KPI dashboard generation

At Saeree ERP, our implementation and support teams have been using Claude Code on the Team plan for our daily ERP work. The doubling of usage limits has impacted productivity from day one — particularly for generating test cases and large code reviews that previously had to be split across sessions. Now those finish in a single session.

7. Governance Before Putting Multi-Agent Into Production

Multi-agent + doubled usage limits = increased risk of "using it casually". Before rolling this pattern out into finance/HR/legal workflows, put a governance frame in place:

Summary

Topic	What changed
SpaceX × Anthropic	Colossus 1: 300+ MW, 220,000+ NVIDIA GPUs, online within May 2026
Usage limits (Pro/Max/Team/Enterprise)	5-hour rate ×2 — same price
Peak-hour throttling (Pro/Max)	Removed — full capacity all day
Opus API rate limits	Considerably raised — supporting larger agentic workflows
Claude Opus 4.7 (Apr 16, 2026)	Coding +13%, vision 2,576 px, unchanged $5/$25 per MTok
Multi-agent orchestration	Lead + specialists in parallel — opens new enterprise workflow patterns

"Anthropic doubling usage limits without raising prices isn't just marketing — it's a signal that compute is no longer the bottleneck for organizations using Claude on real work. The opening now is to shift from using Claude as a 'smart autocomplete' to using it as 'an agent that runs real ERP processes.' Teams that start designing multi-agent patterns now will, within 6–12 months, have a meaningful productivity gap over teams still using Claude as a chat assistant."

References

• Anthropic — Higher usage limits for Claude and a compute deal with SpaceX (May 6, 2026)
• Anthropic — Claude Opus 4.7 announcement (Apr 16, 2026)
• 9to5Mac — Anthropic updates Claude Managed Agents with three new features
• Claude Code release notes (GitHub)
• Releasebot — Anthropic updates feed