HARVEY, A LEGAL AI startup named after a fictional television lawyer, has managed something rather nonfictional: quadrupling its valuation in thirteen months. On Wednesday, the company announced $200 million in fresh capital at an $11 billion valuation, led by Singapore's GIC and Sequoia Capital. It is the kind of number that, in an earlier era of enterprise software, would have required a decade of compounding growth. Harvey has been in business for three and a half years.
The round's headline figures are impressive but not, by 2026 standards, especially surprising. Harvey reached $190 million in annual recurring revenue by January, up from $100 million announced just five months earlier. Its tools — which handle contract analysis, due diligence, compliance, and litigation support — are now used by more than 100,000 lawyers across 1,300 organizations, including HSBC and NBCUniversal. Sequoia has led three of Harvey's funding rounds, a level of repeat conviction that Pat Grady, a partner at the firm, compared to Salesforce's early cloud trajectory. The investor roster reads like a who's who of Sand Hill Road: Kleiner Perkins, Andreessen Horowitz, Coatue, the OpenAI Startup Fund, and Elad Gil have all written checks at various stages. Total capital raised now exceeds $1.4 billion.
What makes Harvey interesting is not the company itself — though its growth is formidable — but what it represents. For the past two years, a nagging fear has haunted the AI startup ecosystem: that OpenAI and Anthropic, now collectively valued north of $1.2 trillion, are vacuuming up so much capital and capability that they will eventually collapse the application layer beneath them. Why pay for a vertical AI product, the argument goes, when the foundation model can do the job itself?
Billable powers
But the vertical AI counterargument is gaining traction, and Harvey is its most persuasive exhibit. The company's trajectory — from a cold email to Sam Altman and a prototype built on GPT-3's California tenant-law capabilities to an $11 billion enterprise — suggests that domain specificity creates defensibility that raw model intelligence cannot easily replicate. Winston Weinberg, Harvey's CEO and a former securities litigator, co-founded the company with Gabriel Pereyra, a research scientist who had worked at both DeepMind and Meta. That pairing — deep legal domain expertise fused with frontier ML knowledge — has become something of a template for the nascent vertical AI category.
Yet the valuation demands scrutiny. At $11 billion on $190 million in ARR, Harvey trades at roughly 58 times revenue — a multiple that would make even the most ebullient SaaS investor pause. For comparison, Thomson Reuters, which owns Westlaw and commands a $59 billion market capitalization on $7.4 billion in trailing revenue, trades at approximately 8 times. Clio, the Canadian legal-tech stalwart with $300 million in ARR, last raised at a $3 billion valuation. Harvey's premium rests entirely on the assumption that AI-native legal software will not merely complement existing tools but replace entire categories of billable labor — a shift from software budgets to labor budgets that would dramatically expand the addressable market.
There is some evidence to support this thesis. Case studies published by Harvey show mid-sized firms achieving 35% or greater increases in case capacity, with initial document reviews that once took a week collapsing to hours. One family-law practice reported that Harvey's analysis of five years of financial records identified $2.2 million in commingled marital assets within minutes, securing an additional $4 million for the client. When AI can demonstrably move the needle on case outcomes (which is to say, on revenue), the pricing conversation changes entirely.
The broader vertical AI cohort is benefiting from the same dynamics. There are now at least 47 AI-native application companies generating more than $25 million in ARR, and the category is growing at roughly 400% according to Bessemer's latest analysis — competing at roughly 80% of traditional SaaS average contract values. In a market where three companies absorbed 83% of all global venture capital in a single month this February, Harvey's ability to raise $200 million from a sovereign wealth fund and a top-tier venture firm suggests that investors are actively seeking application-layer bets as a hedge against model-company concentration.
Still, the risks are not trivial. Harvey scrapped its own fine-tuned legal model in favor of multi-model agentic workflows last year — a pivot that improved performance but also raised questions about defensibility. If the intelligence layer is rented rather than owned, the moat depends on workflow integration, data flywheel effects, and switching costs rather than on proprietary technology. Sacra's research has noted that Harvey's legal research capabilities route queries through LexisNexis rather than accessing the corpus directly, positioning it as an orchestration layer atop someone else's data. That is a viable business, but it is a different kind of business than the valuation might imply.
For the legal profession, Harvey's ascent poses a question that is equal parts exciting and unsettling: what happens when the tool that makes lawyers more productive also makes fewer of them necessary? Weinberg insists that the firms succeeding with Harvey are using it to take on more work, not to reduce headcount. That may well be true in a demand-rich environment. Whether it remains true in a downturn is another matter — and at 58 times revenue, investors are betting heavily that they will not have to find out. ■
For more, join 75,000 subscribers getting tech's favorite brief here
