Deep dive: Legal AI's $8.8B Price Tag: What the Numbers Leave Out

Why today

On October 21, Legora announced a $550M Series E at an $8.8B post-money valuation. Legora sells AI agents to lawyers and was founded in Stockholm in 2023. Its previous round valued it at $5.6B. Atreides Management led the round. Every major existing investor came back in, including Accel, Benchmark and Salesforce Ventures. Artificial Lawyer, a legal tech trade site, puts Legora's total funding at about $1.4B. It puts Harvey, Legora's main rival, at roughly $1.8B. Harvey raised $550M at $15.5B in September. On the same day, the investor Keith Rabois posted on X that Spellbook, a contract-drafting tool, "is saving a Fortune 100 company $21 m in outside counsel legal fees." Those are two headline numbers backed by very different kinds of evidence. Together they raise today's question. What actually supports legal AI pricing right now? And which numbers would tell a durable leader apart from a company riding momentum?

Plainly

Start with what the price implies. Legora says its annual recurring revenue now stands at more than $200M. Annual recurring revenue, or ARR, is the yearly value of the subscriptions in force today. It is not cash collected, and it is not profit. Divide $8.8B by $200M and you get roughly 44 times ARR. Investors pay that multiple only if they believe the revenue will keep compounding for years. It also has to come at good margins, from customers who stay and expand. A rough calculation makes this concrete. Legora says more than 140,000 legal professionals use it. For a profession that sells its time by the hour, that is not a hard sale. So the question is not whether lawyers will pay. The question is whether they keep using it and keep paying more. A gym is a useful analogy. A gym can sell thousands of memberships in January, and its recurring revenue looks superb. What matters is who is still coming in by March, and who upgrades to personal training. Seat utilisation is the March attendance: what share of paid seats are actually used each week. Net revenue retention is the upgrade question. Take last year's customers and compare what they pay now with what they paid then, after cancellations and downgrades. Gross margin is what is left after paying to run the gym. For legal AI, the biggest running cost is mostly the model inference bill. Agents that carry out multi-step work make many model calls per task, so margin is not a given. The savings claim is a different kind of number altogether. Any savings figure needs a baseline: what the customer would have spent on outside lawyers without the tool, over what period, and at what contract volume. Without a baseline, it could be a forecast written up as a result.

What the evidence shows

Here is what is established. The round size, the valuation, the lead investor and the investor list come from Legora's own announcement, and Artificial Lawyer reports them the same way. The price is a real fact. Fidelity, Capital Group and J.P. Morgan all came in as new investors, and investors at that stage usually see far more data than the public does. Then there is what is self-reported. Legora says it went from $1M to $100M in ARR in under 18 months. It says the next $100M took less than six months. It calls that pace faster than any other enterprise software company with a direct sales team, but gives no method or comparison set. It says more than half of the AmLaw 50, the fifty largest US law firms by revenue, are customers. A customer could mean one practice group or the whole firm. It says in-house teams made up more than 40% of new customers in the third quarter. That counts customers, not revenue. Now the missing pieces. There is no gross margin, net revenue retention, churn or utilisation. There is also no definition of ARR: whether it means live usage or contracted commitments that ramp up later. Legora's CFO, David Eckstein, talks about wanting "healthy gross margins, high retention." That is a goal, stated without a single number. Gavin Baker of Atreides calls it "a sustainable business," but he led the round. The Harvey comparison is useful here. Legal Business reports Harvey serves nearly 2,500 organisations. Legora says it serves more than 2,400. So the market is pricing something other than customer count. It is an investor's social post, with no named customer, no period and no method. The post does not say whether Rabois has a stake. Spellbook's own in-house legal page makes more modest claims. One named customer says outside counsel spend fell by "hundreds of thousands of dollars over a relatively short period of time." The page also claims a 75% cut in drafting and review time, "depending on agreement complexity." Both are vendor claims too.

How people read it

The same facts support quite different readings. In the AGI-forward reading, agents that do multi-step legal work are going after the budget for legal labour, not the budget for legal software. On that view, 44 times ARR is cheap if revenue doubles a couple more times. Legora's own pace, $100M to $200M in under six months, is the evidence this reading points to. The AI-realist reading starts with how law firms buy. Firms under competitive pressure buy seats so they are not left behind, and ARR can outrun daily use. Firms that bill by the hour have mixed incentives to adopt tools that cut hours. Agents run many model calls per task, which puts pressure on margins. And the model providers underneath could move into the category themselves. None of that is disproven by the announcement. None of it is confirmed either. The market reading focuses on who bought. Fidelity, Capital Group and Baron are crossover investors, funds that hold both private and public stocks and often come in before an IPO. Their presence suggests the price reflects how scarce a clear number two in a two-company race is, as much as measured economics. One detail fits several of these readings. In-house teams do not bill hours, so cutting outside counsel spend saves them money directly. Legora's growing in-house share and the savings claims point the same way: the most measurable value may sit with corporate legal departments rather than law firms.

What it means for builders and investors

Our read is that the price is real and the economics are undisclosed. Legora did publish ARR, which is more than many private companies do. But ARR without retention and margin is half the picture. Using this round as a comparable for a seed or Series A legal AI company means borrowing a multiple without seeing what supports it. For builders, the opening may be measurability. In-house buyers can see an outside counsel invoice go down. A vendor that shows a baseline, a period and a referenceable customer has a stronger story than one that shows a big number. That matters more if the product is mostly orchestration on top of frontier models, where differentiation is thin and margins depend on someone else's pricing. One question worth exploring in diligence might be: of last year's customers, what do they pay today, and what share of their paid seats were active weekly last quarter? For savings claims, a natural follow-up is what the customer spent on outside counsel in the twelve months before deployment, and in the twelve months after. We hold this view provisionally.

What to watch

Three things would settle a lot of this. First, any disclosure of net revenue retention or gross margin from Legora or Harvey. The most likely venue is an IPO filing, and neither company has given a timeline. Second, renewals. Many firm-wide deals were signed in the past eighteen months. If those contracts are annual, renewal rates and seat counts will start to show in the coming year, through customer interviews and trade press. One named customer, a stated period and a described baseline would turn it from a marketing number into evidence. Until then, both stories show what the market is willing to pay, not what the businesses have proven.