Analysis
Harvey, the legal AI startup, raised $550 million at a $15.5 billion valuation, the company announced September 9. Diffusion and Lightspeed Venture Partners co-led the round, with new investors Sapphire Ventures and Whale Rock joining a returning roster that includes Sequoia, Kleiner Perkins, Andreessen Horowitz, Coatue, Conviction, Elad Gil, GIC, Goldman Sachs Alternatives and Verified Capital. The company plans to spend the money on product development and hiring as it pushes law firms, in-house legal teams and professional-services firms to build and manage their own AI capabilities.
Harvey was founded in 2022 by CEO Winston Weinberg, a former securities and antitrust litigator at O'Melveny & Myers, and President Gabriel Pereyra, a former DeepMind and Google Brain research scientist. The two began experimenting with GPT-3 on legal workflows shortly after its public release, and a cold email to OpenAI's general counsel led to Harvey becoming one of the first companies backed by the OpenAI Startup Fund, which anchored its $5 million seed round that November. The product is now used by more than 100,000 lawyers across 1,300 organizations, and the company is headquartered in San Francisco.
A valuation that keeps re-rating
The new mark caps an unusually fast run: Harvey was valued at $8 billion in December 2025, $11 billion by March 2026 on a $200 million raise, and now $15.5 billion -- nearly doubling across two rounds in nine months. That pace outstrips even 2026's generally frothy AI-application valuations, and it comes without Harvey disclosing hard ARR figures alongside the announcement, unlike some peers.
The more interesting detail is technical, not financial: Harvey just shipped Harvey Tenet, its first proprietary model, built by post-training the open-weight Kimi K3 rather than building on top of OpenAI's or Anthropic's proprietary APIs. That's notable given Harvey's own origin story runs through the OpenAI Startup Fund. The company is reportedly encouraging its customers to adopt and post-train their own open-weight models too -- a bet that legal AI's real moat is workflow and data, not access to whichever lab has the best frontier model that quarter.
The competitive field
Legal AI has become one of the most closely watched vertical-AI categories precisely because it shows an application layer can out-earn the model layer it sits on:
- Legora -- a Swedish-founded legal AI rival expanding aggressively into the US market, seen as Harvey's closest head-to-head competitor for large-firm contracts.
- Casetext -- acquired by Thomson Reuters in 2023 for $650 million, representing the prior generation of legal-tech AI tools Harvey's broader platform has largely eclipsed in enterprise deals.
- OpenAI and Anthropic -- both remain Harvey's technical benchmark and its indirect competitive threat: either lab could bundle legal-specific tooling into a general subscription at a fraction of Harvey's implied price per seat.
What the multiple assumes
Public detail on Harvey's actual revenue remains thin -- the company hasn't disclosed ARR alongside this raise, a contrast with a rival that published hard numbers the same week:
- Harvey -- $550M raised at $15.5B valuation, no ARR disclosed.
- [Cognition](/pulse/company/cognition) -- run-rate revenue jumped from $492M to $900M alongside its own $48B valuation, TechCrunch reported.
Without a disclosed multiple, outside investors are left pricing Harvey largely on customer count and lawyer-seat growth rather than a clean revenue benchmark.
The bear case is straightforward: legal AI's addressable market is real but not infinite, and $15.5 billion assumes Harvey keeps winning enterprise deals against both dedicated rivals like Legora and the two frontier labs that could each release a cheaper, bundled legal-agent product at any time. Winning that fight without disclosed revenue numbers to point to makes the multiple harder to underwrite from the outside than Cognition's or Clay's.
Harvey's next test isn't the fundraising environment, which has clearly stayed open to it -- it's whether Harvey Tenet's open-weight approach actually produces better legal outputs than a frontier-model wrapper, a claim only customer retention data, not funding announcements, can settle.
