Harvey announced on September 9, 2026 that it had raised $550 million at a $15.5 billion valuation — less than six months after a $200 million round valued the legal AI company at $11 billion. The raise brings Harvey's total capital to more than $1.5 billion, making it comfortably the best-funded company in legal AI. Bloomberg reported the valuation marginally higher, at $15.6 billion; most outlets cite $15.5 billion.
The round was co-led by Diffusion and Lightspeed Venture Partners. Diffusion is a new firm co-founded by Kris Fredrickson, a longtime Harvey backer and former Coatue Management investor. Existing investors including Sequoia, Kleiner Perkins, Andreessen Horowitz, Coatue, Conviction, Elad Gil, Evantic, GIC, Goldman Sachs Alternatives, Verified Capital and WNDR participated, joined by new backers Sapphire Ventures and Whale Rock.
The Revenue Behind the Multiple
The valuation is aggressive but not untethered. Harvey's reported annualised revenue has passed $400 million, more than doubling from roughly $190 million in January 2026. The company says 80% of Am Law 100 firms use its product, alongside five of the Fortune 10 in-house legal departments, across more than 3,000 organisations in 60 countries.
That penetration is the important figure. Legal is a market where the buyer set is small, conservative and highly networked — and where, historically, software has taken a decade to reach majority adoption among the largest firms. Reaching four-fifths of the Am Law 100 inside three years is the fastest enterprise adoption curve the legal technology sector has produced.
The Guardrails AI Acquisition
Announced alongside the funding, Harvey is acquiring Guardrails AI, a San Francisco security platform for AI agents. Co-founders Shreya Rajpal and Zayd Simjee and their team join Harvey's product and engineering organisation.
This is the detail that tells you where the company thinks the next competitive battle is. Harvey is no longer selling document drafting and review; it is selling agents that act on legal matters. The moment software takes an action on a privileged matter rather than drafting text for a human to check, the binding constraint stops being model quality and becomes provable control: what the agent was permitted to do, what it actually did, and how you demonstrate both to a general counsel and, eventually, to a regulator or an insurer. Buying an agent-security team is a bet that auditability, not capability, is the wedge in professional services.
The Open-Weight Turn
The funding also followed Harvey's unveiling of Harvey Tenet, its first post-trained open-weight model, built on a base from Chinese AI startup Moonshot's Kimi K3 and trained specifically for legal tasks. The company additionally launched Harvey LAB — a Legal Agent Benchmark designed to measure how AI agents perform on legal work.
Both moves point in the same direction: reducing dependence on any single frontier lab, and controlling the yardstick by which legal agents are evaluated. A vertical AI company that post-trains its own weights owns its margin structure and its data path in a way that a pure application layer on top of a third-party API does not. It also, of course, gets to define the benchmark on which it scores well — a caveat worth holding in mind when Harvey LAB results circulate.
Why It Matters
Harvey's round is the clearest evidence yet for a thesis that dominated the week's investment activity: capital is flowing to vertical AI with distribution inside existing enterprise workflows, rather than to horizontal tools asking customers to adopt a new category.
The same week logged roughly 74 funding rounds globally, with the largest concentrated in infrastructure and mission-critical verticals:
- Fluidstack, a chip-agnostic neocloud building modular AI data-centre capacity, raised $1.5 billion
- Forus, an AI prescription fulfilment network, raised a $150 million Series C at a $3 billion valuation
- Clay, an AI sales platform, raised a $115 million Series D
- Profound raised $180 million at a $1.8 billion valuation
- Inspiren raised a $70 million Series C for physical AI in senior living care
The pattern across these is consistent. None of them is a model company. All of them sit between a frontier model and a workflow that a specific industry already pays a great deal of money to execute — and the investment case rests on owning that last mile.
The Risk Nobody Is Pricing Loudly
Harvey has now roughly tripled its valuation inside nine months on revenue that has roughly doubled. That gap is the whole debate. At $15.5 billion against $400 million of ARR, the multiple assumes legal AI spending compounds for years and that Harvey holds its share against both incumbent legal publishers and the frontier labs themselves, which increasingly ship domain-tuned capability directly.
The Guardrails acquisition is arguably the most defensible thing in the announcement precisely because it addresses that risk from the direction competitors find hardest to copy. Frontier labs can match Harvey's drafting quality. Matching a security and audit posture that a risk committee at a global law firm has already signed off on takes considerably longer than a model release cycle.
