NVIDIA just funded Harvey's $5.6B Swedish rival

The numbers tell a tighter story than the headlines. Harvey closed $200M in March 2026 at an $11B valuation, co-led by GIC and Sequoia, with $190M ARR — a 58x revenue multiple. Customers: a majority of the AmLaw 100, 500+ in-house legal teams, 50 asset management firms across 60 countries. Legora closed $550M Series D in March at $5.55B, then took a $50M NVentures extension in late April pushing it to $5.6B. ARR: $100M, up from roughly $30M six months ago. Harvey is twice the valuation but only twice the ARR. Legora is growing faster off a smaller base. The category winner is not decided.

NVIDIA's NVentures rarely leads on application-layer companies — their portfolio is heavy on infra (CoreWeave, Recursion, Figure). Legora is one of three apps NVentures has touched in 2026. That is not just capital. It is GPU allocation, GTM co-selling into NVIDIA's enterprise channel, and a public signal to general counsels that NVIDIA sees Legora as a workflow worth optimizing — not a model layer to be obsoleted in 18 months. Same playbook NVIDIA ran with CrowdStrike (security) and Recursion (drug discovery): back the vertical specialist after the generalist hits saturation. They are quietly calling Harvey near-saturated in big law.

Most operators picked Harvey because it was first and looked obvious. The pattern repeats every cycle — Salesforce vs. ServiceNow, Snowflake vs. Databricks, Twilio vs. MessageBird. The challenger usually wins the bottom 80% of the market while the leader compounds in the top 20%. Legora's Stockholm DNA is its moat: 80% of UK Magic Circle firms run it, it was built around European data residency and multi-jurisdictional workflows, and the core UX is collaborative — junior associates editing AI output in shared workspaces, not single-shot Q&A. That product shape maps to mid-market US law firms (95% of the market by headcount) better than Harvey's AmLaw 100-first design.

The vertical AI playbook is now repeatable, and this is what your paid subscription buys you to see early: pick a knowledge-work vertical, build the collaborative workflow (not the chatbot), undercut the leader 30-40% on price, and let champions inside the firm drive bottom-up adoption. The same shape is playing out right now in accounting (Pilot vs. Bench AI), construction (Trunk Tools vs. Procore Co-pilot), insurance (Sixfold vs. Zelros), revenue ops (Clay vs. Apollo AI), and tax (Trullion vs. the Big 4 in-house stacks). The leader gets the marquee logos in year one. The challenger compounds in the mid-market and wins the platform fight 24-36 months later when consolidation hits.

If you sell into a regulated knowledge-work vertical, the door is open RIGHT NOW for the next 18 months — after that, NVIDIA-backed vertical specialists will own the workflow layer and you will be selling around them, not into them. The trap most operators fall into is betting on the leader and standardizing too early. Don't. In legal AI today the smarter play is dual-deploy: Harvey for litigation and M&A workflows where AmLaw distribution matters; Legora for collaborative drafting, compliance review, and multi-jurisdictional work. Pay both for 6 months. The one your associates open every morning wins. Standardize then — not before.