![Harvey Hits $15.5B: AI Owns the Am Law 100](https://cdn.slatesource.com/d/0/9/d0909e79-4e37-42ca-a5f8-d86c2e5764c1.webp) # Harvey Hits $15.5B: AI Owns the Am Law 100 - [Made in Slatesource](https://slatesource.com/@steph/harvey-hits-15-5b-ai-owns-the-am-law-100) - By [Steph](https://slatesource.com/@steph) - Created on Sep 11, 2026 ## $400M ARR. From $10M three years ago. That is the number. Harvey raised $550 million on September 9, 2026, at a valuation of $15.5 billion, co-led by Diffusion and Lightspeed Venture Partners. Six months ago, a $200M round put the company at $11 billion. The new figure adds $4.5 billion of value in roughly 180 days, a pace that makes almost every other private software company look slow. The revenue arc is starker. Harvey launched in November 2022 with zero revenue. It hit $10M ARR by the end of 2023, $65M by the end of 2024, $100M in August 2025, $190M by January 2026, and $400M by September 2026. That is 40x in three years. Sacra, which tracks the company closely, calls it one of the fastest ARR trajectories recorded in enterprise software history. ![](https://cdn.slatesource.com/d/0/9/d0909e79-4e37-42ca-a5f8-d86c2e5764c1.webp) ## Eight in ten of the hundred biggest firms already pay The customer base tells the story more plainly than the growth charts. Harvey says 80% of the Am Law 100, the ranking of the hundred highest-grossing law firms in the United States, now use its tools. Five Fortune 10 in-house legal teams are clients. More than 3,000 organisations across 58 countries subscribe, and there are over 700 named partners on the platform. That is not a startup selling to early adopters. That is a product that has moved through BigLaw's procurement committees and cleared every conflict check. The firms paying are the same firms that bill $1,200 an hour for associate time and spend decades defending that rate. They are paying for Harvey anyway. [Harvey.ai: official announcement of the $550M round, September 9, 2026](https://www.harvey.ai/blog/harvey-raises-dollar550m-at-a-dollar155b-valuation-to-help-legal-teams-own-their-intelligence?utm_source=slatesource) ## What the $550M is actually buying Co-founders Winston Weinberg and Gabe Pereyra built Harvey out of a simple observation: legal work is almost entirely document processing, pattern matching, and precedent retrieval, tasks that large language models do well. The early product was a wrapper. The current one is not. Harvey Tenet, launched alongside this round, is the company's first post-trained open-weight model. It starts from a Kimi K3 base and was trained on roughly 1,750 agentic legal task environments using asynchronous reinforcement learning, on about 150 NVIDIA B300 GPUs over two months. Harvey LAB, its Legal Agent Benchmark, is now the public measure the field uses to compare legal AI systems. ## The fourth acquisition of 2026 The round also funds Harvey's fourth corporate transaction this year: the acquisition of Guardrails AI, a San Francisco security platform specialising in testing the behaviour of AI agents. Co-founders Shreya Rajpal and Zayd Simjee join the product and engineering teams. The capital will go toward model research, compute, and commercial expansion, particularly in markets outside the US, where the Am Law 100 brand means little but the need to process contracts in seventeen languages is real. ARR, end of 2023 $10M ARR, end of 2024 $65M ARR, August 2025 $100M ARR, January 2026 $190M ARR, September 2026 $400M+ Valuation, March 2026 $11B Valuation, September 9, 2026 $15.5B Total raised since founding $1.55B+ > The opportunity for companies to accelerate their competitive advantage with AI has never been higher. Harvey wants to be the global partner legal teams turn to for this work. Winston Weinberg and Gabe Pereyra, Harvey co-founders > > — Steph · 11th of September 2026 ## The model providers are watching and moving The competitive threat is no longer from other legal startups. Anthropic launched Claude for Legal in August 2026, a product with dozens of connectors and plugins built specifically for law firm workflows. OpenAI has similar ambitions. Both companies have the advantage of training on every legal document Harvey has ever processed, having supplied the underlying models for much of Harvey's early architecture. Kris Fredrickson, who led Diffusion into this round after years at Coatue, put the thesis plainly in a conversation with Bloomberg: Harvey's moat is not the model. It is the 1,750 task environments, the benchmark, the training data built from live legal work, and the relationships inside the firms. You cannot rent those from a hyperscaler. [Artificial Lawyer: Harvey raises $550M and acquires Guardrails AI, September 9, 2026](https://www.artificiallawyer.com/2026/09/09/harvey-raises-550m-at-15-5bn-val-buys-guardrails-ai/?utm_source=slatesource) ## What the billing partner loses sleep over The question lawyers ask quietly is not whether Harvey is useful. It is whether it eventually prices out the junior associate. A first-year at a top-50 US firm costs roughly $250,000 to employ for a year and bills perhaps 1,800 hours. Harvey completes the same document review in hours for a fraction of that cost. The firms are absorbing the tension by charging clients the old rate while using Harvey internally, effectively pocketing the margin on the time that used to go to junior staff. That works until a client notices and starts asking why AI costs are not flowing through to lower bills. That conversation is already happening. At $15.5 billion and 40x ARR growth, Harvey is not a threat to the legal industry. It is already the legal industry's infrastructure. The question that the next round will need to answer is what happens to the lawyers who built that infrastructure's value.