Claude Opus 4.7 and the SpaceX deal — orbital compute, doubled Code limits, same prices

Two announcements from Anthropic on May 4 that ought to be read together: a new flagship model, and a compute deal that materially changes their capacity ceiling.

Claude Opus 4.7

The model itself is positioned as an incremental upgrade over Opus 4.6:

Notable gains on advanced software engineering, particularly on the hardest tasks. Anthropic’s framing — and matching independent reports — is that Opus 4.7 is the first Claude where you can confidently hand off your hardest coding work without babysitting.
Substantially better vision — higher resolution image inputs, improved fine-detail extraction.
Same pricing as 4.6 — $5/M input tokens, $25/M output tokens.

Pricing-flat-with-quality-up is the move I’ve been expecting from all the frontier labs as competition tightens. Worth noting that the headline numbers don’t move; the unit economics do.

The SpaceX partnership

The structurally interesting announcement. Anthropic signed a deal with SpaceX for the entire compute capacity at Colossus 1:

300+ megawatts of new capacity.
220,000+ NVIDIA GPUs coming online within the month.
The two companies are exploring orbital data centres — Anthropic mentioned interest in multiple gigawatts of orbital AI compute capacity.

Yes, orbital. The pitch for space-based compute is solar (constant illumination), thermal (radiative cooling against deep space), and political (jurisdiction, supply-chain redundancy). Whether it’s actually economical at scale is a different question, but the fact that it’s being seriously discussed at the GW scale by an organisation that ships product is worth filing away.

The downstream effects are immediate:

Claude Code’s 5-hour rate limits doubled for Pro, Max, Team, and Enterprise plans, effective immediately.
API rate limits raised across the board for Opus.
This is the second compute capacity announcement from Anthropic in two months — they’ve been visibly compute-constrained, and this addresses it.

What I take from this

The compute war is here. The frontier labs are now openly trading capital for capacity in ways that show up as customer-facing rate limits. Operating at the frontier is a capex question first, R&D question second.

The duopoly dynamic is real. Anthropic at $30B annualised, OpenAI at $24B, both racing to lock in compute. The mid-2025 view that the field would have 4–6 frontier labs is now harder to defend. Most of the rest are de-facto open-source partners or cloud-vendor offerings.

Vendor risk is back on the agenda. When two companies are responsible for most of the inference for the GenAI economy, single-vendor dependency starts costing real procurement points. Build for vendor-neutrality (MCP, abstracted model calls) accordingly.

We doubled our Claude Code allocation today and used it. The new Opus is genuinely better. If the orbital data-centre piece comes together, this gets even more interesting.

Sources: