DISCO vs Relativity

DISCO and Relativity both fold generative AI into the base price now, so the argument that used to decide this — one vendor bundles AI, the other bills it as an add-on — is dead. What still separates them is the unit the meter counts. DISCO charges per GB of processed data and lets you run AI as much as you want. Relativity includes aiR in the subscription but counts every AI run in aiR Units, one per document. That difference decides the bill on a chat-heavy matter, and it decides how much a second review pass costs you.

What changed, and when

Relativity announced at Relativity Fest in October 2025 that aiR for Review and aiR for Privilege would become standard in RelativityOne pricing rather than paid add-ons. The RelativityOne pricing page now lists aiR for Review, aiR for Privilege, and aiR for Case Strategy at no additional cost.

DISCO answered on 25 February 2026 with the DISCO Platform: ediscovery, Cecilia AI, deposition management, and timelines — the last two previously sold as Case Builder — combined at one per-GB price on processed data with no ingest fees and no per-user fees.

Any comparison written before late 2025 gets this backwards. If you are reading a review that prices aiR as a per-document add-on or Relativity as per-GB-plus-per-seat, it predates both announcements.

Where DISCO wins

• AI usage does not move the bill. The per-GB hosting fee covers unlimited Cecilia Q&A, single-document Q&A, document summaries, deposition summaries, and auto timelines. Re-running a prompt across the corpus after a bad first pass costs nothing extra.
• One line item to negotiate. No ingest fee, no processing fee, no per-user fee. When deduplication is selected at ingest, duplicate data is not billed at the global family level.
• Review throughput with a published number. Cecilia Auto Review runs at up to 32,000 documents per hour.
• Depositions and timelines share the review database. Transcript review, annotations, and case timelines sit against the same documents, so fact investigation does not need a second tool and a second export.

Where Relativity wins

• AI outside document review. aiR for Case Strategy, aiR for Data Breach Response, Relativity Contracts, Relativity FOIA, and Relativity Legal Hold run on the same platform. DISCO’s AI is pointed at discovery and depositions.
• The service-provider channel. Relativity reports more than 300,000 users. Almost every litigation service provider runs it, so changing vendors mid-matter is a hand-off rather than a re-platforming.
• Proven volume on the AI itself. Relativity reported over 200 customers using aiR to review more than 25 million documents in aiR’s first 20 months, at up to 3 million documents per day.
• Recent acquisitions extend the surface. Gavel brings drafting into Microsoft Word; VerQu covers communications-data capture and migration.

The pricing math that actually differs

Neither vendor publishes a rate. Both pricing pages route you to a quote, so “transparent pricing” here describes the structure, not a number you can look up. Structure still decides the bill.

DISCO bills one per-GB rate on processed data, and AI consumption does not enter the calculation. Relativity bills per-GB hosting and includes aiR, but consumption runs against aiR Units: one unit per document for aiR for Review, aiR for Privilege, and aiR for Case Strategy, and one unit per five successful insights for Custom Analyses, rounded down at the job level. Jobs queued or canceled before producing results do not consume units.

The consequence is arithmetic. Hosting scales with gigabytes; unit consumption scales with document count. Two 200 GB corpora burn wildly different unit counts depending on what is in them — Slack and Teams exports are many small documents per GB, scanned PDFs are few large ones. Messaging-heavy matters are where DISCO’s flat structure wins hardest, and they are also the matters growing fastest.

Second passes are the other place the models diverge. Recoding a corpus after the issue list changes is free at DISCO and consumes units again at Relativity.

Ask both vendors for the per-GB rate on processed data — not collected, not ingested — and ask Relativity for the aiR Unit allowance in the subscription plus the overage rate. Without those three numbers the two quotes are not comparable, whatever the headline per-GB figures say.

The on-prem question is closing

Relativity Server support ends at the end of 2027. From 1 January 2028, customers cannot create new workspaces or new matters, or repurpose existing workspaces, in Server; matters created on or before 31 December 2027 stay supported. Server pricing was adjusted effective 1 April 2026 to reflect the cost of running both products.

DISCO has never sold on-prem. So “we need on-premises” is now a reason to plan a migration with a deadline, not a reason to buy Server. If on-prem is a hard regulatory requirement past 2028, neither of these is your answer.

Vendor risk cuts both ways

DISCO is public as NYSE: LAW. Q1 2026 revenue was $41.9 million, up 14% year over year, with a GAAP net loss of $9.6 million and full-year 2026 guidance of $169.25 million to $178.75 million. The numbers are readable, and they are small next to the incumbent.

Relativity is private and confidentially filed a draft registration statement for a US IPO on 19 March 2026 — the first legal tech IPO attempt since 2021. Press coverage has put the raise near $750 million at roughly $4 billion; treat those as reported figures, not confirmed terms. Packaging and list pricing move around an IPO, so a three-year commit signed in 2026 is worth more than a pay-as-you-go rate.

Watch-outs

• DISCO’s agentic Cecilia Q&A was announced on 9 February 2026 for delivery later in 2026, at no additional cost. Guard: do not sign on it. Make general availability a contractual milestone with a remedy, or price the deal as though the feature does not exist.
• “Included at no additional cost” is not the same as unlimited. Guard: get the aiR Unit allowance and the overage rate onto the order form, not into a sales email.
• Per-GB on processed data is not per-GB on collected data. Guard: ask each vendor which volume the rate applies to and whether deduplication runs before metering, then price the quote against your own last matter’s collected-to-processed ratio rather than the vendor’s example.
• Service-provider gravity is real. Guard: if outside counsel and your review vendors all run Relativity, confirm production formats and load-file specs in both directions before you run a matter in DISCO.

Verdict

• Pick DISCO if you run matters in-house, your data is messaging-heavy, and you want an AI bill that does not move with usage.
• Pick Relativity if service providers or outside counsel run your matters, you need AI beyond review — breach response, contracts, FOIA, legal hold — or you are on Server and have to land somewhere before 2028.
• The default, if you cannot decide: DISCO for corporate legal departments running their own matters under about a terabyte a year, Relativity for everyone else. Volume and channel decide this, not features.
• If neither fits: Everlaw is the mid-market in-house alternative with a similar per-GB model, and Reveal is where teams that need a deployment model outliving the Server sunset end up looking.

The single mistake to avoid is comparing quotes on the headline per-GB rate. Same gigabytes, different meters — one counts storage, the other counts storage and documents.