Version 10

Anthropic closed a $65 billion Series H at a $965 billion post-money valuation on May 28, 2026 [3], disclosing run-rate revenue of $47 billion in May — up from $30 billion in April and roughly $14 billion in February [3][4]. Compute agreements now total 15 gigawatts across Amazon, Google/Broadcom, and SpaceX's Colossus infrastructure, with Micron, Samsung, and SK hynix added as strategic chip partners [3]. Both The Information and the Financial Times have independently reported an IPO as early as Q4 2026 [28][29], while formally registered CVEs for Claude Code [21] and Speakeasy's continued active MCP server generation development [25][26] remain the sharpest unresolved pressures on Anthropic's infrastructure and governance narrative.

A $65 billion round at near-trillion-dollar valuation, combined with revenue that has roughly tripled in three months and confirmed Q2 operating profitability, positions a potential Q4 2026 Anthropic IPO as one of the most consequential public offerings in technology history. The same public-market due diligence that will validate the revenue trajectory will scrutinize the formally registered CVEs, enterprise adoption friction, and governance credibility of the Agentic AI Foundation with a rigor that private investors have not applied.

Anthropic's financial trajectory in the first half of 2026 has been exceptional by any historical benchmark. The company closed a $30 billion Series G at a $380 billion post-money valuation in February [1], confirmed a $30 billion annualized revenue run rate in April [2], and on May 28 announced a $65 billion Series H at a $965 billion post-money valuation [3] — a round that exceeded even the highest pre-announcement estimates. The accompanying press release disclosed run-rate revenue of $47 billion in May, tracing a trajectory from roughly $14 billion in February through $30 billion in April [3][4]. Simon Willison observed that the numbers carry particular credibility because misrepresenting them to $65 billion in new investors would constitute securities fraud, and an S-1 filing for the anticipated IPO will make the actual figures public in any case [4]. The round added Micron, Samsung, and SK hynix as strategic chip partners to help scale memory and storage supply chains [3]. An anonymous AI consultant told Axios that a single client spent half a billion dollars in one month on Claude licenses after failing to set usage limits — an anecdote that simultaneously illustrates the scale of enterprise adoption and raises questions about customer concentration risk [4].

The Series H disclosed compute agreements totaling 15 gigawatts of new capacity — a significant expansion from the 5 gigawatts announced with the April AWS deal — naming Amazon, Google/Broadcom, and SpaceX's Colossus infrastructure as partners [3]. Anthropic had earlier committed $200 billion to Google's cloud and chip infrastructure [5] and entered a 10-year arrangement with Amazon securing more than $100 billion in AWS spending, with Amazon committing an additional $25 billion direct investment [2][6]. The Information also reported early-stage talks to lease Microsoft's custom Maia 200 chips for inference workloads [7], positioning Microsoft as a potential fourth compute supplier. On talent, Andrej Karpathy joined in May 2026 to lead pretraining research [8], and Anthropic's $300 million acquisition of Stainless [9] — the builder of every official Anthropic SDK and a leading MCP server generation platform — brought a critical tooling layer in-house [10].

On protocol governance, Anthropic donated MCP to the Agentic AI Foundation under the Linux Foundation umbrella, with MCP, the goose agent framework, and AGENTS.md as founding projects [11][12]. The foundation has since added 43 new enterprise and government members [13][14], and a16z published analysis affirming MCP as foundational to 2026's defining shift from prompts to agents [15]. Against this governance momentum, the security surface has escalated from researcher-documented to CVE-tracked: OWASP formally catalogued MCP Tool Poisoning [16]; Invariant Labs, CyberArk, Unit 42/Palo Alto Networks, and academic researchers published independent vulnerability findings [17][18][19][20]; and formally registered CVEs for Claude Code — CVE-2025-59536 and CVE-2026-21852 — have appeared [21]. A dedicated certification and training industry has since emerged around MCP security [22], and enterprise practitioners at Portkey.ai and Solo.io independently document operational friction in enterprise MCP deployments that exceeds what documented attack surfaces alone capture [23][24]. Speakeasy, the direct competitor to Stainless in MCP server generation, has continued actively building — producing direct comparisons against Stainless and appearing in developer roundups as a leading alternative [25][26][27].

Both The Information and the Financial Times have reported that Anthropic is considering an IPO as early as Q4 2026 [28][29], and prediction markets actively quote odds on whether Anthropic or OpenAI will go public first [30][31]. At a $965 billion post-money valuation with $47 billion in run-rate revenue and confirmed Q2 operating profitability [32][3], a Q4 2026 listing would rank among the largest public offerings in technology history. SpaceX's assembly of a 21-bank underwriting syndicate [33][34] provides a concrete comparable for the banking activity and scrutiny likely to surround any Anthropic listing at this scale. The same due diligence that validates the revenue trajectory will also scrutinize the formally registered CVEs, the AAIF's governance authority, customer concentration, and whether $65 billion in capital can be efficiently deployed into compute commitments while sustaining the operating profitability that has only just been achieved.