Legal AI Valuations Soar as Harvey Hits $15.5B Amid Sanctions

The money: a valuation curve with almost no precedent

Harvey has become the clearest measure of how much investors now believe in legal AI. In September 2026 the company announced a $550 million round at a $15.5 billion valuation. Diffusion and Lightspeed Venture Partners co-led the round. Sequoia, Kleiner Perkins, a16z, Coatue, GIC and Goldman Sachs Alternatives returned as investors, and Sapphire Ventures and Whale Rock joined for the first time.21 Bloomberg put the figure slightly higher, at $15.6 billion, and said the money is meant to help Harvey build its own AI models. The gap between those two numbers is a rounding question. The speed of the climb is the real story.

The company's own timeline makes that clear. Harvey was worth $3 billion in February 2025, $5 billion in June 2025, $8 billion in December 2025 and $11 billion in March 2026.27 The March round alone, a $200 million raise co-led by GIC and Sequoia, made Harvey one of a small group of AI startups worth more than $10 billion, alongside OpenAI, Anthropic, Perplexity and Sierra.35 It has now raised more than $1.55 billion in total.25

Revenue is growing quickly but has not been audited. Co-founder Winston Weinberg said annual recurring revenue had passed $400 million, up from about $190 million in January. That number came from interviews, not from Harvey's funding announcement.24 Earlier in September, outside reporting put the figure closer to $350 million.24 In August, The Information reported that Harvey was negotiating the round with annualized revenue above $350 million.26

Analysts don't agree on how expensive that makes Harvey. One deal tracker said the round valued the company at about 14 times ARR.34 The arithmetic doesn't support that: $15.5 billion divided by roughly $400 million is close to 39 times, the multiple another analysis reported.25 The higher figure is the right one, and it is a high price by any normal enterprise-software standard.

The rest of the herd

Harvey is not the only legal AI company pulling in huge sums. Stockholm-based Legora closed a $550 million Series D led by Accel in March at a $5.55 billion valuation. That was about triple its value five months earlier. An April extension led by Nvidia's venture arm and Atlassian brought the round to $600 million and the valuation to $5.6 billion.31 Estimates of Legora's revenue range from just over $100 million6 to about $150 million.29 The Financial Times reported that the company plans to grow from 650 to about 1,500 employees by the end of 2026.4

Below those two, the list of unicorns keeps getting longer. Norm AI, which turns regulations into rules software can apply, raised a $120 million Series C at a $1.2 billion valuation in July. Khosla Ventures led the round, and New York Life, TIAA and Vanguard took part. Personal-injury specialist EvenUp reached a valuation above $2 billion with a $150 million Series E.6 GC AI, which sells to corporate legal departments, raised a $60 million Series B at a $555 million valuation.33 Crunchbase counts more than $7 billion invested in legal tech over the past two years, most of it in AI companies. It also says 2026 is running only slightly behind the record year, with at least 12 startups raising rounds of $50 million or more.36

The money is spread very unevenly. One industry analysis found that Relativity, Harvey and Legora took nearly two-thirds of first-quarter 2026 funding while the median round dropped to around $1 million.4 Another estimated that rounds at Series B or later account for about 80% of legal AI dollars, and that the ten largest deals take a growing share. In practice, this market rewards companies that are already winning and is getting tougher on generic AI wrappers.

Acquisitions: the land grab moves to M&A

The largest companies are also spending heavily on acquisitions. Legora has made at least five this year and Harvey at least three.33 Recent deals include Legora buying Vancouver's Walter AI and Harvey buying Hexus. Thomson Reuters picked up Noetica, and in an unusual move, the law firm Cleary Gottlieb bought Springbok AI.3

Clio, a practice-management company founded in 2008, has the most complete acquisition strategy in the sector. In November 2025 it completed its $1 billion purchase of vLex, the largest deal legal tech has seen. It also closed a $500 million Series G at a $5 billion valuation and secured a $350 million debt facility.10 Some coverage still gives Clio's valuation as $3 billion, which is out of date.6 In June 2026 Clio bought Toronto-based Jurisage to fill gaps in vLex's Canadian legal content, and an executive described the deal as an "elbows-up moment" against Harvey and Legora.8 On September 30 it bought Learned Hand, a startup that builds AI tools for judges. That gives Clio's new judiciary business a product already in use in some courthouses.9 Chief executive Jack Newton told Business Insider that getting judges onto Clio's data and tools could eventually help it win large law firms, the market where Harvey and Legora are spending heavily.7

The strategic pattern is clear. Harvey and Legora are competing for elite law firms with AI-first products. Clio is building wide infrastructure on top of data it owns: more than a billion legal documents across 110-plus jurisdictions and about a billion U.S. court records.9 Both are betting that complete platforms will beat single-purpose tools. One analysis predicts that more practice-management and document-management companies will buy AI vendors over the next two years.2

The risk: the courtroom keeps the score

Investors are paying for these products to be dependable. Courts are publishing a running record of what happens when legal AI output isn't checked. A public database had logged 2,046 court cases involving AI-fabricated citations or content as of September 21, 2026, up from about 200 a year earlier.12 Other trackers report lower numbers depending on when and how they count: about 1,227 worldwide in April20 and 1,600 across 35 countries by June.15 All of them show the same steep rise.

The penalties have also become more severe. The Sixth Circuit ordered two lawyers in Whiting v. City of Athens to pay $15,000 each, plus the other side's appellate fees and double costs, and said lawyers must personally read and verify every citation they file.11 In April, Nebraska's Supreme Court indefinitely suspended a lawyer whose brief contained 57 defective citations out of 63.11 In June, a federal judge in Mississippi cancelled a trial after both sides filed fake citations, and barred the two lead attorneys from practicing in that district for two years.12 U.S. courts imposed at least $145,000 in such sanctions in the first quarter of 2026 alone.16 The record for a single matter is about $109,700, in an Oregon case12, though one report gives a much lower figure for the same case.16

The problem is international. England's Solicitors Disciplinary Tribunal struck off a lawyer over AI-generated citations in September. India's Supreme Court overturned a customs penalty of ₹425.27 crore because the order relied on authorities that don't exist.16

For anyone weighing these valuations, the most important detail is that specialist legal tools are not immune. In 2026 the Fifth Circuit sanctioned a lawyer who said he drafted arguments with vLex and Thomson Reuters CoCounsel, and those arguments cited fabricated authority.17 vLex is now part of Clio. A Stanford RegLab study found that even paid legal research tools hallucinate 17% to 34% of the time.12 So a premium product is not, by itself, a guarantee of accuracy.

The reading

Investors are not ignoring this risk. They appear to be betting that it will push buyers toward the large, well-funded platforms. A market analysis that was otherwise upbeat still marked reliability as getting worse, said verification remains a major unsolved problem, and pointed to governance and verification tools as the next big opportunities.4 Clio describes its court product as "judiciary-grade," which suggests it sees trust as something it can sell.7 Harvey's new spending on its own models is a large capital bet. It only pays off if legal-specific models clearly outperform what OpenAI and Anthropic already offer.25

The most likely outcome is that the hype is partly justified but the field won't survive intact. Lawyers are adopting these tools for real: Harvey says 80% of the Am Law 100 use its product.24 But valuations of nearly 40 times revenue assume a profession that still bills by the hour will turn time saved into profit, and so far law firms have struggled to do that.4 Expect the leaders to keep buying smaller rivals, and expect thinly funded startups to either get acquired or fade. For lawyers and their clients, the courts have already set the rule that matters: AI can draft, but the lawyer who signs is responsible, and the sanctions record shows that responsibility is being enforced.