This was a defining week in Anthropic’s trajectory. The company secured 300+ MW of GPU capacity from SpaceX’s Colossus 1 data center โ enabling doubled Claude Code rate limits and the removal of peak-hour throttling effective immediately. That compute breakthrough arrived alongside 10 financial services agent templates with Microsoft 365 add-ins, a new enterprise AI services company backed by Blackstone and Goldman Sachs, and creative tool connectors for eight professional platforms. Here is the Claude Weekly for May 7, 2026.
SpaceX Colossus 1 Deal: More Compute, Higher Limits
Anthropic signed an agreement with SpaceX to use all compute capacity at Colossus 1 โ over 300 MW with 220,000+ NVIDIA GPUs, coming online within the month (Source). This sits on top of existing deals with Amazon (up to 5 GW), Google and Broadcom (5 GW), and Microsoft and NVIDIA ($30B Azure). Anthropic also expressed interest in partnering with SpaceX on orbital AI compute capacity.
The immediate impact: Claude Code five-hour rate limits are doubled for Pro, Max, Team, and seat-based Enterprise plans. The peak-hours limit reduction that throttled Pro and Max accounts is removed entirely. API rate limits for Claude Opus models were raised “considerably.” International expansion via Amazon adds inference capacity in Asia and Europe for data residency.
For enterprise teams, compute has been the binding constraint on Claude adoption. This is the single largest capacity injection in Anthropic’s history โ the constraint is materially looser today than it was last week.
Agents for Financial Services: 10 Templates with M365
Anthropic shipped 10 ready-to-run agent templates for financial services, available as plugins in Claude Cowork and Claude Code, plus cookbooks for Claude Managed Agents (Source). The templates cover research and client coverage (Pitch Builder, Meeting Preparer, Earnings Reviewer, Model Builder, Market Researcher) and finance and operations (Valuation Reviewer, General Ledger Reconciler, Month-End Closer, Statement Auditor, KYC Screener).
Each template packages skills, governed data connectors, and subagents into a deployable unit with SOC compliance configurations, audit logging, and PII masking โ the specific guardrails regulated institutions have been waiting for.
Microsoft 365 integration is the strategic headline. Claude now works inside Excel (financial models, formula audits), PowerPoint (decks with live-updating numbers), and Word (credit memos against firm templates), with context carrying across applications. Outlook is coming soon. For enterprise teams on M365, Claude now lives inside the tools their people already use.
New data connectors include Dun & Bradstreet, Fiscal AI, Financial Modeling Prep, Guidepoint, IBISWorld, SS&C IntraLinks, Third Bridge, and Verisk. Moody’s launched an MCP app for credit ratings on 600M+ companies. Claude Opus 4.7 leads the Vals AI Finance Agent benchmark at 64.37%.
New Enterprise AI Services Company
Anthropic joined Blackstone, Hellman & Friedman, and Goldman Sachs to form a new company targeting mid-sized enterprises that lack in-house resources for AI deployment (Source). Applied AI engineers from Anthropic will work alongside the firm’s engineering team to build custom Claude solutions. Additional backers include General Atlantic, Apollo, GIC, and Sequoia Capital.
The entity joins the Claude Partner Network alongside Accenture, Deloitte, and PwC. Per CFO Krishna Rao: “Enterprise demand for Claude is significantly outpacing any single delivery model.”
The target audience โ community banks, mid-sized manufacturers, regional health systems โ are organizations that need Claude but cannot afford Big Four rates. For mid-market consultancies like Big Hat Group, this signals both competition and validation: the mid-market enterprise AI category is real enough to attract the largest private equity firms on the planet.
Claude for Creative Work
Eight new creative tool connectors position Claude beyond coding and productivity (Source): Ableton (documentation-grounded for Live and Push), Adobe (50+ Creative Cloud apps), Affinity by Canva, Autodesk Fusion (conversational 3D modeling), Blender (official MCP connector, open to all LLMs), Resolume Arena/Wire (VJ and live visual control), SketchUp, and Splice (royalty-free samples). Anthropic made a donation to the Blender Project. The Blender connector works with other LLMs โ signaling commitment to the MCP standard over platform lock-in.
Product Ecosystem: Managed Agents, Dreaming, and MCP
Claude Managed Agents entered public beta with Dreaming โ a scheduled process that reviews past sessions and memory, curating patterns, mistakes, and preferences across agents over time (Ars Technica, Docs). This solves context window limits for long-running multi-agent projects โ similar to how BHGBrain provides persistent agent memory with tiered retention and vector search. Outcomes (success criteria) and multi-agent orchestration graduated from research preview to wider availability.
Enterprise features include per-tool permissions, managed credential vaults, full audit logs in the Claude Console, and user review workflows before client-facing output.
MCP: The spec remains at 2025-11-25. GitHub added an MCP Registry entry to its navigation โ a small signal the standard is becoming infrastructure. For teams building custom MCP connectors or agent skills, Big Hat Group’s application development practice designs and deploys governed solutions aligned with SOC compliance requirements.
Company Expansion: AU/NZ, Japan, and Beyond
Theo Hourmouzis (ex-Snowflake SVP APAC) was named GM of Australia and New Zealand, with a Sydney office now open (Source). Partnerships include Commonwealth Bank, Quantium, Canva, and Xero (multi-year deal). Seoul is next.
In Japan, NEC Corporation will use Claude to build a ~30,000-person AI engineering organization โ Anthropic’s first Japan-based global partner (Source). Joint development targets finance, manufacturing, and local government.
Competitive Landscape
Apple is preparing iOS 27 “Extensions” โ letting users choose which third-party LLM powers Siri, Writing Tools, and Image Playground โ already tested with Anthropic models (TechCrunch). This could be Anthropic’s largest distribution opportunity.
Google shut down Project Mariner, folding it into Gemini Agent (The Verge). Separately, Gemma 4 with Multi-Token Prediction achieves ~2x speed on consumer GPUs โ pressuring Claude on local inference costs (Ars Technica). Google I/O is May 19.
SAP invested โฌ1B in European AI and restricted its agent ecosystem to approved platforms including Nvidia’s NemoClaw โ early signs of enterprise platform lock-in (TechCrunch).
DeepSeek eyes a $45B first venture round, continuing to pressure frontier pricing (TechCrunch).
What to Watch
- Colossus 1 going live within the month โ expect further limit increases and Asia/Europe inference regions.
- Apple WWDC (June). If “Extensions” formalizes with Claude as an option, it reshapes the mobile AI landscape.
- Google I/O (May 19). Agent ecosystem announcements and potential MCP adoption signals.
- Enterprise platform wars. SAP’s NemoClaw and the new Anthropic-backed services company point toward fragmentation in enterprise AI delivery models.
That is this week’s Claude Weekly. The SpaceX compute deal removes the biggest friction point for Claude power users, the financial services templates open a regulated industry that has been waiting for these guardrails, and the creative connectors prove Anthropic is building beyond coding. Infrastructure, product, and distribution are accelerating simultaneously.
Keeping up with the Claude ecosystem is essential for any AI strategy. Big Hat Group’s AI and Automation consulting helps enterprise teams deploy and optimize Claude solutions โ from financial agent integrations to custom MCP connectors. Contact us for a free 30-minute strategy session.
Check back next week for the latest.