TypeSafe Raises $870M Series A at $7.5B for Non-Text AI Model Jev

What Happened

TypeSafe AI raised an $870 million Series A at a $7.5 billion valuation, led by Andreessen Horowitz (a16z), with Sequoia Capital, seed-lead investor DCVC and a group of angel investors participating. Bloomberg broke it on October 9, TechCrunch followed hours later with the company's confirmation, and deal counsel Wilson Sonsini published its own announcement the same day. $7.5 billion is the closed valuation; in late September The Information reported talks of "over $1 billion at a $10 billion valuation." Those talks did not close; the final amount and valuation came in below the reported figures.

The company is a 2024-founded San Francisco startup: CEO Diogo Almeida is a former OpenAI researcher and a co-inventor of RLHF who worked on InstructGPT; co-founder Sasha Sheng is a former Meta research engineer, and Erik Gafni is an engineer-turned-serial-founder. TypeSafe exited stealth on September 15, announcing a $40M seed round (led by DCVC at a ~$200M valuation) and launching Jev on the same day. Three-plus weeks later, the valuation went from $200M to $7.5B — 37.5x.

Key Facts

1. 24 days, a 37.5x valuation jump: The seed valued the company at ~$200M on September 15; the Series A at $7.5B on October 9 — 24 days apart. In late September The Information reported talks of "over $1B at a $10B valuation"; the final $870M / $7.5B came in below the reported numbers, yet it still ranks among the year's biggest AI rounds (Crunchbase's weekly roundup puts it second only to the $2B Axiom deal).
2. What Jev is: a decision model that outputs no text: TypeSafe's thesis: LLMs are for chatting with humans; software needs answers it can execute, not paragraphs to parse. Jev is a transformer-based model that does not output text — feed it application state and it returns typed questions and answers: a choice, a score, a boolean, each with a probability attached. The company calls it a "decision model," trained with RLCD (Reinforcement Learning for Calibrated Decisions); Almeida says it runs far faster with far fewer tokens than LLMs. The speed and efficiency figures all come from company benchmarks, with no independent verification.
3. The big numbers are company-reported; no customers are named: TechCrunch relays the company's claims: "about a third of Fortune 500 companies are already using Jev" and "one million users within days." AlphaMaven notes that the company declined to name any customers, so the adoption figures "cannot be independently confirmed from the available reporting." The third-party number comes from Vercel: nearly 13% of its AI Gateway paid teams used Jev within 24 hours of launch — the fastest adoption in the gateway's history; that figure measures usage, and Jev was free on the gateway for the first ten days. The 40 million launch-video views are also company-reported.
4. The close date has not been disclosed: RuntimeWire, summarizing Bloomberg's report, notes: Bloomberg did not report when the financing closed. The headline numbers — $870M, $7.5B — have been announced, but neither close timing nor terms (preferred structure, ratchets, liquidation preferences) have been disclosed.

Context

The timeline: 2024 — Almeida, Sasha Sheng and Erik Gafni found TypeSafe in San Francisco and operate in stealth for nearly two years; September 15, 2026 — the company exits stealth: a $40M seed (DCVC, ~$200M valuation) and the Jev launch on the same day; within 24 hours, Vercel reports nearly 13% of its AI Gateway paid teams used Jev — the platform's fastest adoption ever — OpenRouter usage rises noticeably, and Cloudflare, LangChain and Langfuse integrate within three days; September 24 — The Information reports talks of over $1B at a $10B+ valuation; October 9 — the close: $870M Series A at $7.5B, led by a16z.