Legal AI Is Now a Two-Horse Race Priced at 100x Revenue

Legora, the Stockholm-based legal AI company, is in early talks with investors to raise at a valuation above $10 billion — roughly double the $5.6 billion it commanded in April 2026, four months ago.

Its principal competitor, Harvey, is reportedly repricing toward $15.5 billion in the same window.

Two companies serving the same professional services vertical, both nearly doubling their marks inside a single month. That is either the clearest signal yet that AI has found genuine product-market fit in a large industry, or the clearest signal yet that a specific corner of the venture market has stopped anchoring on revenue. The evidence points, uncomfortably, to both.

The revenue curve is real

Legora's growth is not a projection. It is one of the steepest documented enterprise software ramps on record.

The company went from about $3 million in ARR at the end of 2024, to $50 million at the end of 2025, to $100 million in April 2026. Bessemer Venture Partners has characterized the $3M-to-$100M pace — roughly eighteen months — as the fastest any enterprise software company has achieved.

The customer base backs it up: more than 1,000 customers across 50 markets, with tens of thousands of legal professionals actively using the platform. Earlier reporting has named Linklaters, Deloitte, and Heineken among them, and the company's Q2 growth has been described as roughly 50%.

Those are not pilot logos. Magic Circle firms and Big Four accountancies do not deploy tools across thousands of fee-earners as an experiment.

Why legal was always the best vertical

There is a reason legal is the first professional services category where AI has produced billion-dollar businesses rather than demos.

The output is text. Legal work product is documents — contracts, memos, briefs, diligence reports. This is the modality language models are best at, with no translation layer required.

The inputs are structured and available. Contracts, filings, case law, and precedent documents are exactly the kind of corpus that retrieval works well over.

The billing model rewards it. Law firms bill for time. Any tool that compresses a first-year associate's six-hour document review into forty minutes converts directly into margin, and the buyer can calculate the ROI on the back of an envelope.

The work is genuinely awful. Document review and diligence are the tasks junior lawyers hate most and the tasks firms have the least emotional attachment to automating. There is no internal constituency defending them.

That combination is unusual. Most enterprise verticals have one or two of those properties. Legal has all four.

The multiple problem

At $10 billion against roughly $100 million of ARR, Legora would price at about 100x revenue.

There is no comfortable way to describe that number. High-growth enterprise software has historically topped out around 20-30x forward revenue at the peak of enthusiastic markets, and those multiples were applied to companies with proven retention and years of cohort data.

The bull case rests entirely on the growth rate. A company compounding this fast makes today's multiple look different in eighteen months. If Legora reaches $300 million ARR next year, $10 billion is 33x — expensive but recognizable. If it reaches $500 million, it is 20x, and the round looks prescient.

The bear case is that 100x prices in that outcome as a certainty rather than a scenario, and leaves no room for the thing that eventually happens to every fast-growing software company: deceleration.

The competitive knife-edge

Legora and Harvey are converging on the same customers with similar products, and both are repricing upward at the same time. That is a specific and unstable market structure.

Two well-funded competitors in a category with a finite number of very large buyers — the global elite law firms, the Big Four, the largest corporate legal departments — produces intense pressure on pricing and an expensive race for logos. Each new marquee account is worth more to the winner and costs more to acquire than the last.

The capital both are raising is largely for that fight. Which means the valuations are not just marks on a business. They are war chests, and the size of each is partly a response to the size of the other.

There is also a structural risk neither round fully prices. Both companies build on foundation models they do not own. The frontier labs have shown a consistent willingness to move up the stack into vertical applications when a vertical proves valuable enough — and $10 billion and $15.5 billion are exactly the kind of numbers that prove it.

The defensibility argument is workflow depth: integrations with document management systems, firm-specific precedent libraries, security and confidentiality architecture that survives a general counsel's review, and the accumulated institutional knowledge of how a particular firm actually works. Those are real switching costs, and they are not the kind of thing a general-purpose model release erases.

Whether they are worth 100x revenue is a different question.

The read

Legal AI has produced the clearest genuine adoption story of the enterprise AI era. The revenue is real, the customers are the hardest-to-sell institutions in professional services, and the growth rates are documented rather than projected.

What is being priced now is not whether the technology works in law. That question is settled. What is being priced is which of two companies ends up owning the category, and how much of the value created by AI in legal accrues to the application layer rather than to the labs underneath it.

Both rounds are bets on the second question having a favorable answer. Nothing in the market has confirmed it yet.