It was one of the busiest weeks the AI industry has seen this year. Anthropic announced a landmark compute partnership with SpaceX, shipped a meaningful upgrade to Claude, and quietly doubled rate limits for many of its paying users overnight. Here's the short version of what changed and why it matters for anyone running AI inside a business.
1. Anthropic × SpaceX — the compute deal nobody saw coming
On 6 May, at Anthropic's "Code with Claude" developer event in San Francisco, Anthropic and SpaceX announced a major compute partnership. Under the agreement, Anthropic takes all of the compute capacity at SpaceX's Colossus 1 data centre in Memphis, Tennessee — one of the largest AI data centres in the world, with more than 220,000 Nvidia processors including H100, H200, and next-generation GB200 accelerators. The facility brings around 300 megawatts of new capacity online for Anthropic within a month — enough electricity to power more than 300,000 homes.
The deal is striking for two reasons. First, scale: it materially eases the capacity crunch Anthropic has openly admitted has been hurting reliability and peak-hour performance for Claude users. Second, politics: Elon Musk has publicly criticised Anthropic in the past — yet at the announcement he said he had spent significant time with senior Anthropic leaders over the prior week and was "impressed" by their work to make Claude "good for humanity". Both companies also signalled longer-term ambitions including data centres in space.
2. Higher Claude limits — effective immediately
The day of the announcement, Anthropic also pushed through usage upgrades that real teams felt the same evening:
- Claude Code's five-hour rate limit was doubled for Pro, Max, Team, and seat-based Enterprise plans.
- Peak-hour limit reductions for Pro and Max Claude Code accounts were removed.
- Claude Opus API rate limits were significantly increased.
The practical effect: long coding sessions, repository analysis, refactors, debugging runs, and multi-step agent workflows are far less likely to stop mid-task. If you've been frustrated by Claude Code halting on big jobs, that pain just eased.
3. Claude Opus 4.7 is generally available
Anthropic also released Claude Opus 4.7 across Claude apps, the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. It is a focused upgrade rather than a flashy new generation, but the gains are real:
- 13% lift on a 93-task coding benchmark over Opus 4.6, including four tasks neither Opus 4.6 nor Sonnet 4.6 could solve.
- Better attention to instructions and self-verification on long-running tasks.
- Substantially improved vision — it sees images at higher effective resolution.
- More tasteful and creative output for interfaces, slides, and documents.
- Roughly the same quality at low effort as Opus 4.6 at medium effort — meaning lower cost per task for most workloads.
Anthropic shipped Opus 4.7 with new safeguards that automatically detect and block requests indicating high-risk cybersecurity uses. A separate, invitation-only research preview called Claude Mythos Preview — built for defensive cybersecurity work as part of Project Glasswing — autonomously identified and exploited a 17-year-old remote code execution vulnerability in FreeBSD with no human involvement after the initial prompt. A clear signal of where agentic security tooling is heading.
4. Claude Design
Alongside Opus 4.7, Anthropic Labs launched Claude Design — a new product for collaborating with Claude to create visual work directly. It pushes Claude further into the territory currently occupied by tools like Figma and Gamma.
5. The wider landscape — briefly
The rest of the field did not stand still. OpenAI made GPT-5.5 Instant the default model in ChatGPT, claiming roughly 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts in medicine, law, and finance. It also previewed GPT-5.5-Cyber for vetted security teams, released three new realtime audio models (GPT-Realtime-2, GPT-Realtime-Translate, GPT-Realtime-Whisper), and rolled out ChatGPT for Excel and Google Sheets to Business customers worldwide. OpenAI separately disclosed it expects to spend around $50 billion on compute this year — underlining how compute access is fast becoming the primary constraint on the entire industry.
What it means for your business
Three takeaways for South African operators and leaders watching from outside Silicon Valley:
- Reliability is improving fast. If Claude Code dropped tasks for your team in March or April, retry the same workflows — many of those failures should be gone.
- Compute is the new oil. The Anthropic–SpaceX deal and OpenAI's $50bn spend are both about the same scarcity. Expect pricing power to follow capacity ownership.
- Coding and visual work are converging on agents. Opus 4.7's gains, Claude Design, and Mythos Preview's autonomous exploit all point to AI moving from assistant to operator. Plan your workflows accordingly.
The pace is unlikely to slow. We will keep summarising the bits that actually matter for getting work done.