11 SEP 2026 — Harvey has raised 550 million dollars at a valuation of about 15.5 billion, co-led by Lightspeed Venture Partners and Diffusion. Annual recurring revenue has passed 400 million dollars, which is the number that makes this round unusual: most AI valuations at this level are not attached to a revenue figure at all.
What the numbers say
More than 3,000 customers, including 80 per cent of the top 100 law firms, 20 per cent of the Fortune 500 and half of the Fortune 10. Total raised is now above 1.55 billion dollars, and the valuation has roughly doubled in about nine months from 11 billion.
Diffusion is a new firm cofounded by Kris Fredrickson, formerly of Coatue and a long-standing Harvey backer. Sapphire Ventures and Whale Rock Capital Management joined, alongside existing investors.
Bloomberg reports the valuation as 15.6 billion and most other outlets as 15.5. The gap is a rounding difference on a private valuation, and it matters only because nobody can check either figure.
Thirty-nine times revenue
Divide 15.5 billion by 400 million and you get about 39. For comparison, mature enterprise software trades in the high single digits to low twenties on forward revenue, and the fastest-growing public software companies occasionally reach the thirties.
The multiple is aggressive but not absurd, which is a different position from most of what we have covered this month. Forus tripled to 3 billion with no public revenue figure at all. Harvey has put a number on the table that can at least be argued about.
What the multiple assumes is that the growth rate holds. Doubling a valuation in nine months prices continued expansion, not the current book, and legal software has historically been a slow-replacement market where firms change systems rarely.
The customer list cuts both ways
Eighty per cent of the top 100 law firms is close to saturation of the segment that matters most. That figure represents a strong position. Large firms are hard to sell to, procurement takes months, and once a tool is embedded in document review it is awkward to remove.
It also means the easy growth is finished. The next 400 million dollars has to come from mid-market firms, from in-house legal departments at the 80 per cent of the Fortune 500 that are not yet customers, or from selling more to the firms already signed.
The third of those is the likeliest and the least visible from outside. Seat expansion inside an existing account looks identical to new growth in an ARR figure, and only the company can see the difference.
An open-weight model from a legal vendor
One detail most coverage skipped is Harvey's first post-trained open-weight model, released alongside Harvey LAB, a benchmarking product for legal agents.
An application company publishing open weights is a strategic statement. It says Harvey does not intend to compete on the model and would rather commoditise that layer than depend on a frontier lab's pricing — the same reasoning that has pushed several application companies toward open weights this year.
The benchmarking product is the more interesting half. A vendor that publishes a benchmark for its own category is trying to define how that category is measured, which is a powerful position and an obvious conflict. We made the same point about a security vendor reporting six of six on a benchmark it designed.
Why law was always going to be first
Legal work is a good fit for language models, both technically and economically. The output is text, the input is text, the work is billed by the hour at rates that make a software subscription look cheap, and the errors are caught by people whose job is to catch errors.
That last point about review is the most important. A legal AI product operates inside a review process that already exists, which is why adoption ran ahead of sectors where a mistake reaches the customer directly.
It is also why the 400 million dollars is believable in a way that similar figures elsewhere are not. Law firms buy software that saves associate hours because they can calculate the saving precisely.
What to watch
Net revenue retention. That is the figure separating seat expansion from new logos, and Harvey has not published it. Whether the open-weight model is adopted by anyone outside Harvey's own product, since that is the test of whether it was a strategy or a press release.
And whether a frontier lab moves directly into legal workflows. Harvey's position is as an application layer above a model. The labs have shown this year that they will climb into applications when the revenue is visible enough.
View author profile → · Editorial policy
