Last updated: 23 September 2026. Comparables refreshed to Harvey’s September 2026 round and Legora’s April 2026 extension; the sizing model now uses live eCourtsIndia price points and every input is shown.
Is Indian legal AI a ten billion dollar category? Not on today’s numbers. A bottom-up model built on the Bar Council of India’s enrolment figure, real Indian price points and Harvey’s current revenue multiple puts the category at roughly USD 0.2 billion to USD 4.4 billion of enterprise value. Ten billion needs about Rs 2,270 crore of annual revenue, more than twice our stretch case. This post shows every input so you can argue with it.
An earlier version of this post claimed ten billion dollars was the conservative case. When we reran the numbers, the three parts it stacked (USD 5.7 billion, USD 0.5 to 1 billion and USD 0.5 billion) added up to USD 6.7 to 7.2 billion, not ten, and the biggest part used a lawyer price point we do not charge. A sizing post that does not add up is worse than no sizing post, so we rebuilt it from the bottom. The comparables below are current as of September 2026. The price points are the ones on our pricing page.
Key takeaways
- Harvey raised USD 550 million at a USD 15.5 billion valuation in September 2026 with ARR above USD 400 million, a multiple of about 39 times revenue.
- Legora extended its Series D to USD 600 million at a USD 5.6 billion valuation in April 2026, with ARR above USD 100 million.
- India’s bar is large: the Bar Council of India told the Supreme Court in April 2023 that about 25.70 lakh advocates were enrolled. How many practise, and where, is an estimate.
- At Indian price points (AI Clerk starts at Rs 250 a month), our model gives Rs 160 crore to Rs 987 crore of annual revenue across lawyers, enterprise data and background checks.
- That is USD 0.2 to 4.4 billion of value depending on the multiple. Ten billion needs roughly 2.3 times our stretch revenue, so treat it as an ambition, not a floor.
What the global comparables say in September 2026
Four reference points frame the category. Two price the application layer, one prices a strategic acquisition and one prices a data layer business in India.
Harvey. On 9 September 2026 Harvey announced a USD 550 million round at a USD 15.5 billion valuation, co-led by Diffusion and Lightspeed, as TechCrunch reported. It said ARR had crossed USD 400 million and that it serves more than 3,000 customers, including 80 percent of the top 100 law firms. Six months earlier it had raised at USD 11 billion. At USD 15.5 billion on USD 400 million plus of ARR, the multiple is about 39 times, down from the high fifties implied by the March round. Growth has caught up with price.
Legora. The Stockholm company raised a USD 550 million Series D led by Accel at USD 5.55 billion in March 2026, then extended it to USD 600 million at USD 5.6 billion in April, when it said ARR had passed USD 100 million. That is up to about 56 times revenue. Two companies with the same product shape, priced at a premium in two different home markets, tell you legal AI is not a one-winner, one-geography category.
CaseText. Thomson Reuters bought CaseText, maker of the CoCounsel assistant, for USD 650 million in cash in 2023. It was the incumbent research publisher paying up so as not to be left behind. We read a similar signal in the Clio and vLex deal in Reading the Clio-vLex Deal.
TransUnion CIBIL. The closest Indian analogue for a data layer that turns scattered records into structured intelligence sold to lenders. The cleanest public mark is Bank of India’s March 2017 sale of its 5 percent stake for Rs 190.62 crore, which implied a value of roughly USD 590 million for the whole bureau. That is a data layer, with no AI application on top, valued at over half a billion dollars nearly a decade ago.
| Comparable | Date | Value | Revenue marker | Implied multiple |
|---|---|---|---|---|
| Harvey | Sep 2026 | USD 15.5 billion | ARR above USD 400 million | About 39x |
| Harvey (previous round) | Mar 2026 | USD 11 billion | ARR about USD 190 million (reported) | About 58x |
| Legora | Apr 2026 | USD 5.6 billion | ARR above USD 100 million | Up to about 56x |
| CaseText (acquired by Thomson Reuters) | 2023 | USD 650 million | Not disclosed | Not disclosed |
| TransUnion CIBIL (5% stake sale) | Mar 2017 | About USD 590 million implied | Not used | Not used |
The inputs, one by one
The old version of this post leaned on three soft figures: a practising bar of “seventeen lakh”, an AI subscription at Rs 24,000 a year, and a flat 57 times multiple. None survives a check. Here is what we use instead.
Bar size. In April 2023 the Bar Council of India told the Supreme Court that enrolled advocates, about 16 lakh at an earlier point, were estimated at almost 25.70 lakh, according to the court’s order of 10 April 2023. Enrolment is not practice. Some enrolled advocates are in-house, in government, in other careers or retired. We do not have a sourced practising figure, so the model runs penetration against the enrolled number and keeps the rates low. For comparison, the American Bar Association counts roughly 13 lakh active lawyers in the United States. By headcount, India’s bar is the largest in the common law world. We looked at what that buyer pool actually looks like in The Bar Is Seventeen Lakh People.
Price. The eCourtsIndia AI Clerk costs Rs 250 plus GST a month (500 credits), Rs 5,000 for six months or Rs 10,000 a year (20,000 credits). There is a free tier of 50 credits a month, and search, cause lists and directories are free. Case tracking costs Rs 5 per case per month. So the realistic range for a paying advocate is Rs 3,000 a year at the entry plan to about Rs 10,000 a year for a heavy user. Rs 24,000 was never a number the district bar would pay, and we have written about that bar in Most Indian Lawyers Practise in District Courts.
Enterprise demand. Banks, NBFCs, insurers, background verification firms, RegTech vendors and corporate legal teams all need court data. RBI-registered NBFCs alone run into the thousands, so the pool is not small. We do not have a census of organisations that buy court data today. The model assumes 500 to 1,500 paying organisations at Rs 10 lakh to Rs 15 lakh a year each. Treat those as assumptions, not findings.
Background checks. This leg is live now. LegalCheck is our identity-first legal background check, and its partner API costs Rs 99 per check pay-as-you-go or Rs 33 with a subscription. The model assumes 1 crore checks a year at Rs 33 in the conservative case and 2.5 crore a year at Rs 99 in the stretch case.
Exchange rate and multiple. We convert at an assumed Rs 88 to the dollar; the conclusion does not change at Rs 85 or Rs 90. For multiples we show three: 10 times (a mature software or data business), 39 times (Harvey today) and 56 times (Legora’s top end).
The model, with the arithmetic shown
| Revenue leg | Conservative case | Stretch case |
|---|---|---|
| Advocate subscriptions | 10% of 25.70 lakh = 2.57 lakh advocates x Rs 3,000 = Rs 77.1 crore | 20% of 25.70 lakh = 5.14 lakh advocates x Rs 10,000 = Rs 514 crore |
| Enterprise court-data API | 500 organisations x Rs 10 lakh = Rs 50 crore | 1,500 organisations x Rs 15 lakh = Rs 225 crore |
| Background checks | 1 crore checks x Rs 33 = Rs 33 crore | 2.5 crore checks x Rs 99 = Rs 247.5 crore |
| Total annual revenue | Rs 160.1 crore (about USD 18 million) | Rs 986.5 crore (about USD 112 million) |
| Multiple applied | Conservative case | Stretch case |
|---|---|---|
| 10x (mature software or data) | About USD 0.18 billion | About USD 1.1 billion |
| 39x (Harvey, Sep 2026) | About USD 0.71 billion | About USD 4.4 billion |
| 56x (Legora, Apr 2026, top end) | About USD 1.0 billion | About USD 6.3 billion |
Two things stand out. First, even the stretch case, which assumes one in five enrolled advocates pays about Rs 10,000 a year, lands under USD 5 billion at Harvey’s multiple. Second, the lawyer leg dominates only in the stretch case. In the conservative case the three legs are closer in size, which is a sign that enterprise data and background checks matter more to the Indian category than they do to Harvey.
What would have to be true for ten billion dollars
Work backwards. At 39 times revenue, USD 10 billion needs about USD 258 million of annual revenue, or roughly Rs 2,270 crore at Rs 88. That is 2.3 times our stretch case. At a 10 times multiple it needs Rs 8,800 crore a year, which no Indian legal business is close to.
Getting to Rs 2,270 crore needs some mix of the following. Higher revenue per advocate as AI drafting and research become daily tools and heavy users move up the credit ladder. Enterprise contracts that are larger than Rs 15 lakh because court data becomes part of every credit, KYB and collections workflow. Background checks that become routine for hiring, tenancy and vendor onboarding. And products that do not exist yet, such as litigation analytics for insurers or lenders. None of that is impossible. None of it is in the numbers today.
So the honest version of the claim is this. India’s legal AI category is a multi-billion dollar opportunity on today’s price points, and it becomes a ten billion dollar category only if revenue per user and per enterprise rises well beyond where it sits in 2026. The investment case for Indian legaltech is argued at more length in The $793 Million Question. Tracxn lists more than 1,000 legal tech startups in India, so the question is less whether the money arrives and more which layer captures it.
Why the Indian AI lawyer is likely to be built in India
A fair question in venture meetings is whether a Western platform with a multi-billion dollar valuation will simply arrive and take this market. Harvey is already in India in one sense. In January 2026 it acquired Hexus, a small product-demo startup, and said it would open a Bangalore engineering office, TechCrunch reported. That office builds for Harvey’s global in-house customers. Nothing Harvey has announced targets the Indian district bar, and the reasons are structural rather than a matter of will.
The data is not in their corpus
Harvey and Legora grew up on American and European legal material that publishers and court systems spent decades digitising and curating. Indian court data is a different engineering problem: 25 High Court portals, district and taluka courts in every state and union territory, scanned orders in regional languages, and party names spelt five different ways. eCourtsIndia’s index holds 32 crore+ case records and 125 crore+ orders and judgments across the Supreme Court, all 25 High Courts, district and taluka courts in all 36 states and union territories, and 18 tribunal and commission types. We explain why that layer is the defensible part of the business in The Data Moat in the Age of Commodity LLMs. Anyone entering India seriously either builds that layer or partners with someone who has.
Indian procedure is its own system
Article 226 writs against Article 32 writs. Section 482 of the CrPC, now Section 528 of the BNSS. Order XXXVII summary suits. Section 138 of the Negotiable Instruments Act and the Section 139 presumption. The twin conditions for bail under Section 45 of the PMLA, which we covered in PMLA Bail Reality Check. A general model can read the text of these provisions. Reasoning about how they play out in a sessions court needs Indian judgments, Indian procedure and the statute book in a structured form. Ours carries 10,084 Acts and 2,69,602 provisions on IndiaCode.
Much of the work happens in regional languages
District court proceedings, orders and cause lists often run in the language of the state: Hindi, Marathi, Tamil, Telugu, Bengali, Kannada, Gujarati and more. Frontier models handle these languages far better than they did in 2023. The bottleneck is the pipeline that turns a scanned, typewritten order into clean text and matches the names in it to the right parties. That pipeline sits on the critical path for an Indian team and nowhere on a Western roadmap.
The price points do not translate
Harvey sells to large firms and enterprises with enterprise contracts. That is rational for a firm billing by the hour in dollars. It does not work for a sole practitioner in a district court. The product that reaches the wider Indian bar starts free, charges Rs 250 a month for the first paid tier, lets an advocate track a case for Rs 5 a month and sends WhatsApp alerts for 50 paise each. That is a different cost structure and a different sales motion.
Distribution looks nothing like the West
Indian legal practice runs on WhatsApp and the phone. Juniors get briefs over WhatsApp, cause-list changes travel over WhatsApp and clients ask for updates there. So our alerts go to WhatsApp and email, and eCourtsIndia ships an Android app, an iOS app and a Chrome extension that shows tomorrow’s listings. Enterprise SSO on a desktop web app, the default Western rollout, does not reach a two-person chamber in a tier 2 town.
Partnership is the likelier shape
The probable future is not Harvey shipping a rival Indian product. It is either Western platforms staying focused on markets where their unit economics work, or a partnership in which a frontier model sits on top of an Indian data layer and an Indian product team. The Conglomerate Litigation Map we built for Adani, Reliance, Tata and Birla, in The Conglomerate Litigation Map, is an example of work that only exists because that Indian layer exists.
The window
Two things shape the timing. Model quality is converging across the frontier labs, so the advantage moves to whoever owns fresh, structured, local data. And legal workflow tools are sticky. Once an advocate has a year of cases, orders and client notes inside a tool, switching means rebuilding that memory. Surveys suggest most Indian lawyers already use general AI tools, and the gap is the court data behind them, which we covered in Most Surveyed Indian Legal Professionals Already Use AI, But Can’t Reach Court Data. The team that makes that data usable inside the advocate’s daily routine sets the default for the category.
What this means for eCourtsIndia
The global comparables are real: Harvey at USD 15.5 billion, Legora at USD 5.6 billion, CaseText at USD 650 million, CIBIL implied at nearly USD 600 million in 2017. India has the largest bar in the common law world and very little incumbent tooling below the High Courts. On today’s price points that is a multi-billion dollar category, not yet a ten billion dollar one. We are building the data layer and the products on top of it, from free search to the AI Clerk to LegalCheck, and we would rather size the market honestly than inflate it.
If you want to see what sits under the model, start with free case search, try the AI Clerk plans, or build on the eCourtsIndia API (23 endpoints, Rs 200 free credits on signup). For the wider architecture, read The Operating System for Indian Law.
TL;DR
- Harvey: USD 15.5 billion, ARR above USD 400 million, about 39x (Sep 2026). Legora: USD 5.6 billion, ARR above USD 100 million (Apr 2026). CaseText: USD 650 million (2023). CIBIL: about USD 590 million implied (2017).
- India’s bar: about 25.70 lakh enrolled advocates, per the Bar Council of India to the Supreme Court in April 2023. The practising share is an estimate.
- At live Indian prices, the model gives Rs 160 crore to Rs 987 crore of annual revenue across advocates, enterprise data and background checks, worth about USD 0.2 to 4.4 billion at 10x to 39x.
- Ten billion dollars needs about Rs 2,270 crore a year at Harvey’s multiple, 2.3 times the stretch case.
- Harvey has a Bangalore engineering office, but its product targets global firms and in-house teams. Indian data, procedure, languages, prices and distribution favour a locally built AI lawyer.
Sources
- TechCrunch, Harvey hits USD 15.5B valuation, 9 September 2026
- TechCrunch, Legora hits USD 5.6B valuation, 30 April 2026; Legora Series D announcement, March 2026
- TechCrunch, Harvey acquires Hexus, 23 January 2026
- Thomson Reuters acquisition of CaseText, 2023, USD 650 million
- Bank of India disclosure of its 5% TransUnion CIBIL stake sale for Rs 190.62 crore, March 2017
- Supreme Court of India order, 10 April 2023, recording the Bar Council of India’s estimate of about 25.70 lakh advocates
- American Bar Association, National Lawyer Population Survey
- Tracxn, Legal Tech startups in India
- eCourtsIndia pricing and LegalCheck partner API pricing, September 2026
- eCourtsIndia corpus figures verified against the eCourtsIndia MCP, 23 September 2026
Read next: The Operating System for Indian Law and Most Surveyed Indian Legal Professionals Already Use AI, But Can’t Reach Court Data.
Frequently Asked Questions
How big is the Indian legal AI opportunity?
On live Indian price points, a bottom-up model gives Rs 160 crore to Rs 987 crore of annual revenue across advocate subscriptions, enterprise court data and background checks. At 10 to 39 times revenue that is roughly USD 0.2 to 4.4 billion of value. Ten billion dollars would need about Rs 2,270 crore a year. The inputs are explained on the eCourtsIndia pricing page.
What are the global comparables for legal AI valuations?
Harvey raised at USD 15.5 billion in September 2026 with ARR above USD 400 million, about 39 times revenue. Legora reached USD 5.6 billion in April 2026 with ARR above USD 100 million. Thomson Reuters bought CaseText for USD 650 million in 2023, and a 2017 stake sale implied about USD 590 million for TransUnion CIBIL. Read our Clio-vLex analysis.
How many lawyers does India have?
The Bar Council of India told the Supreme Court in April 2023 that about 25.70 lakh advocates were enrolled. Not every enrolled advocate practises, and no current official figure splits the bar by court tier, so practising and district-tier numbers are estimates. By headcount it is still the largest bar in the common law world. Find advocates on the eCourtsIndia lawyer directory.
Is Harvey AI building for India?
Harvey acquired Hexus in January 2026 and said it would open a Bangalore engineering office, but that team builds for Harvey’s global law firm and in-house customers. Nothing announced targets the Indian district bar, whose needs are Indian court data, regional languages and prices near Rs 250 a month. Search Indian cases free on eCourtsIndia search.
Which enterprise buyers need Indian court data?
Banks, NBFCs, insurers, background verification firms, RegTech vendors and corporate legal teams all check litigation exposure before they lend, hire or sign. Our model assumes 500 to 1,500 such buyers at Rs 10 to 15 lakh a year each. They reach the data through the eCourtsIndia API, the MCP server or LegalCheck, the background check product.
What would make Indian legal AI a ten billion dollar category?
At Harvey’s multiple, about Rs 2,270 crore of annual revenue, more than twice our stretch case. That needs higher revenue per advocate as AI drafting and research become daily tools, larger enterprise contracts, routine legal background checks and new analytics products. It is a plausible ambition, not today’s math. See the LegalCheck launch.