⚡ Palantir's US commercial AI revenue surpassing government for the first time is a landmark signal that enterprise adoption is now the primary growth engine for applied AI platforms.
Happy Tuesday. Anthropic dropped a bombshell revised S-1 today, revealing $17 billion in Q3 revenue and nearly $900 billion in purchase commitments — numbers that reframe the AI IPO calculus entirely. The same day, the company's researchers warned they are 'very close' to recursive self-improvement capability, a safety threshold the AI doomsday crowd considers the point of no return. Meanwhile, Marvell's $115B acquisition of optical component maker Lumentum signals that the race to wire AI data centers is entering a new, consolidation-driven phase.
When Anthropic filed its original S-1 last week, the market got the story structure but not the numbers. Today's amended filing fills in the blanks — and the figures are staggering: $17 billion in a single quarter, with a forward-commitment backlog of nearly $895 billion. To put that in context, that backlog alone exceeds the entire annual revenue of most Fortune 50 companies. The losses are also growing, which tells you Anthropic is spending aggressively to maintain its technical lead and build out inference capacity ahead of the IPO. What this really signals is that the enterprise market has decisively picked sides: Claude is no longer a niche safety-first alternative to GPT; it is a primary infrastructure bet for large organizations. The IPO could value Anthropic north of $200 billion, reshaping how every AI startup — and every AI investor — thinks about terminal value. One caution: purchase commitments are not revenue; watch for the fine print on cancellation terms and ramp schedules when the roadshow begins.
Enterprise AI adoption is generating both enthusiasm and cost sticker shock: Y Combinator's Garry Tan is personally spending up to $200K a year on AI agent tokens, while Palantir's commercial AI revenue has overtaken government contracts for the first time — a milestone that validates the enterprise market's maturation.
Two distinct middleware stories stand out today: OpenAI quietly publishing its Pando Git-at-scale system reveals how Codex-level code generation is straining even the best developer tooling, while Google's Mandiant report on LLMjacking puts a concrete dollar figure — $100K per day — on the new threat of stolen AI API access.
Anthropic is dominating the model-layer narrative today on two fronts simultaneously: blockbuster IPO financials that validate Claude as enterprise infrastructure, and a sobering internal warning that recursive self-improvement capability may be closer than anyone expected.
The AI infrastructure build-out is accelerating on multiple vectors at once: a landmark $115B optical-networking M&A deal, SpaceX/xAI expanding its Colossus 2 supercluster, Google eyeing a 1.2GW campus, and Microsoft snapping up hundreds of acres in Iowa — all pointing to a data center arms race with no ceiling in sight.
Beyond Anthropic's IPO numbers, today's deal flow shows continued appetite for AI-adjacent infrastructure and tooling: a $115B optical networking mega-merger, a $75M robotics warehouse raise, a $62.5M behavioral AI launch, and a $45M no-code model training startup all point to investors betting across every layer of the stack.
Content governance and AI copyright are colliding today: the New York Times is suing browser Brave for allegedly scraping and redistributing its content to AI companies including Anthropic and Perplexity, while India's MeitY faces scrutiny over opaque website blocking practices that are increasingly catching journalists in the crossfire.
India's digital governance tensions are front-and-center today, with MeitY's blocking practices under scrutiny across multiple cases — from opaque takedown orders targeting journalists to questions about Meta silently restricting hashtags at government request — signaling that India's AI and internet policy environment remains unpredictable for platforms and publishers alike.
Anthropic warns that it's very close to RSI capability, aka self-building AI — Most IPO filings are financial documents; this one doubles as a safety warning that the company building one of the world's most capable models believes it is approaching the threshold where AI can autonomously improve itself. Senior strategists need to understand what RSI actually means — and what governance frameworks, if any, exist to manage it — before it becomes an unavoidable board-level question. Read →