🧭 Anthropic Closes $65B Series H at $965B Valuation — Near-Trillion-Dollar AI Lab With $47B Run-Rate Revenue

Anthropic has closed a $65 billion Series H, bringing its post-money valuation to $965 billion — the largest private technology financing ever completed. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with numerous other significant investors participating. CFO Krishna Rao confirmed the company's annualised revenue crossed $47 billion earlier in May, making Anthropic the fastest AI company — and one of the fastest companies in any sector — to reach that revenue milestone.

How the capital will be deployed

• Safety and interpretability research — expanded mechanistic interpretability, model welfare research, and red-teaming programmes
• Compute infrastructure — capacity to meet escalating Claude demand; infrastructure partners named in the announcement include Amazon (5 GW commitment), Google/Broadcom (5 GW next-generation TPU capacity), and SpaceX (GPU access via Colossus 1 and Colossus 2)
• Product and partnerships — scaling customer-facing tooling, new Claude platform integrations, and enterprise support headcount
• Strategic supply chain — semiconductor partners Micron, Samsung, and SK hynix join as infrastructure partners to reduce supply-chain bottlenecks

Context: a credible path to $1T

The $965B valuation is not a post-money number padded by future commitments — it reflects capital deployed at that price. The compute partnerships disclosed alongside the raise are notable: the 5 GW Amazon and Google/Broadcom agreements are long-term capacity contracts, not spot commitments. Combined with the SpaceX Colossus deal already in place (an estimated $15B/year), Anthropic is locking in preferential compute access at scale ahead of anticipated capacity constraints in 2027.

🧭 Anthropic Opens Milan Office — EMEA Revenue Up 9× as Italian Enterprise Anchors Sign On

Anthropic formally opened its Milan office on May 27, its sixth European location after London, Dublin, Paris, Zurich, and Munich. Thomas Remy, Head of Southern Europe, leads the team; Chris Ciauri (MD International) described the mission as supporting "Italian enterprise, Italian research, and Italian culture through a safe AI transition." The announcement comes immediately after Pope Leo XIV's encyclical on artificial intelligence, which Anthropic co-founder Chris Olah addressed at the Vatican — giving the Italian launch an unusually prominent cultural backdrop.

Enterprise anchor customers

• Finance: Generali Group, Unipol Group
• Life sciences: Angelini Pharma, Bracco Group
• Energy: Enel Group
• Automotive / mobility: Pirelli
• Consulting & digital: JAKALA — 3,000+ Claude seats deployed across client delivery
• Consumer / fintech: Satispay (6M+ Italian users), Bending Spoons (majority of code changes now co-authored with Claude Code)

EMEA growth signal

EMEA is now Anthropic's fastest-growing region with run-rate revenue up roughly 9× year-on-year and large-business accounts up 10×. The Milan opening is the third European office added in 2026 (after Zurich and Munich) and coincides with the EU AI Act's general-purpose AI provisions coming fully into force — a compliance context that Anthropic is positioning Claude's governance tooling to address directly.

🧭 Claude Code v2.1.157–158: Local Plugin Auto-Loading and Auto Mode on Bedrock, Vertex, and Foundry

Two rapid-fire releases over May 29–30 ship features developers have been waiting for: v2.1.157 decouples plugins from the marketplace entirely, and v2.1.158 brings Auto mode to Bedrock, Vertex, and Claude Platform on AWS (Foundry) for the first time.

v2.1.157 — Local plugin auto-loading

Plugins placed in a .claude/skills/ directory inside any project now load automatically without marketplace registration. This means team-specific or company-internal skills can be version-controlled alongside the codebase, installed via git clone, and used immediately without touching the marketplace. Two companion features arrived alongside:

• claude plugin init <name> — scaffolds a new plugin with the correct directory layout, SKILL.md, and example tool definitions
• Autocomplete for /plugin arguments — subcommands, installed plugin names, and marketplace sources now all tab-complete in the CLI
• EnterWorktree can now switch between Claude-managed worktrees mid-session, enabling multi-branch agent workflows without restarting

# Scaffold a new team plugin
claude plugin init my-company-deploy

# Structure created:
# .claude/skills/my-company-deploy/
# ├── SKILL.md         ← skill description and invocation instructions
# ├── tools.json       ← optional tool definitions
# └── examples/        ← example prompts (optional)

# Auto-loads in any project that has .claude/skills/my-company-deploy/
# — no marketplace step required

v2.1.158 — Auto mode on Bedrock, Vertex, and Foundry

Auto mode — which lets Claude dynamically select the right model tier (Haiku, Sonnet, or Opus) based on task complexity — was previously limited to direct Anthropic API access. v2.1.158 brings it to all three managed-cloud endpoints as an opt-in feature:

# Enable Auto mode on managed cloud endpoints (opt-in)
export CLAUDE_CODE_ENABLE_AUTO_MODE=1

# Works on: Amazon Bedrock, Google Vertex AI, Claude Platform on AWS (Foundry)
# Supports: Opus 4.7 and Opus 4.8 as the maximum tier
# Falls back gracefully if a lower tier model is not available in your region

Auto mode is off by default on managed clouds because enterprise procurement teams often need to control which models are billed to which cost centre. The env-var opt-in lets platform teams enable it selectively per deployment. Anthropic noted a bug fix in the same release: v2.1.156 (also May 29) resolved thinking-block modification errors that were triggering API errors for some Opus 4.8 users — upgrade from v2.1.154 or earlier is recommended.