What is your company's data worth to AI labs?

There’s no list price for business data. Reported deals in 2026 range from tens of thousands of dollars for a small, shuttered startup to $10 million for an airline’s corporate records. Where your company lands depends on how big it is, how far back its records go, how well those records connect, what you have the right to license and how many buyers want it.

This guide gathers the public price signals we could verify, explains what moves the number, and shows how our estimate works.

What have companies been paid for business data?

Most data deals are private, but a handful have been reported in detail. These are the figures we could confirm from published sources:

What was sold or offered	Reported amount	Source
Data from shuttered startups, sold through a wind-down service	Typically $10,000–$100,000 per company	Forbes, April 2026
A 13-year-old company’s Slack, Jira, email and shared-drive archive	“Hundreds of thousands of dollars”	Forbes, April 2026
An AI training company’s program for 50 midsize companies	$100,000–$2 million each, for access to anonymized data	Business Insider, August 2026
Spirit Airlines’ internal data and custom software, in bankruptcy	$10 million winning bid; a rival bid of $7.5 million	Business Insider, August 2026
An individual’s “high-quality written documents”	Up to $30,000	Business Insider, August 2026

These numbers come from very different situations, so treat them as signposts, not comparables. A company winding down is often selling everything at once to a single buyer. An operating company licensing selected records is in a different position entirely.

What drives the value of business records?

The same factors come up again and again, from buyers, sellers and the lawyers who advise them.

Size of the organization

More people doing skilled work means more of that work on record. Team size is the simplest proxy for how much a business has documented, which is why it anchors most estimates.

Years of history

Long histories show how work evolves: new clients, new systems, recurring problems and how they were solved. Buyers value both the volume and the change over time.

Data richness: how well records connect

This is the factor owners most often underestimate. Brendan Mahony, CEO of Sunset, which buys data from shuttered startups, told Forbes that price depends on a company’s size, its age and its “data richness,” which he described as internal traceability and links across platforms. In his example, a Jira ticket tied to a specific code commit carries more value than a standalone document.

The same logic applies outside software. An invoice dispute that runs from an email to a chat thread to an adjusted entry in the ledger is worth more than any of those records alone.

Variety of record types

Email, chat, documents, tickets, accounting records, CRM data and call transcripts each capture a different side of the work. Several connected sources usually beat one large silo.

Industry and specialization

Work that’s hard to find elsewhere is worth more. Mahony told Forbes that certain industries, like healthcare and finance, command a premium. Regulated industries also take more rights and privacy work, so the premium has to be earned.

Rights, exclusions and terms

You can only be paid for what you can actually license. Records tangled up in client confidentiality, vendor claims or privacy commitments may need to be excluded. Terms move the price too: exclusive licenses, broad permitted uses and longer retention are worth more to a buyer, and cost you more in flexibility.

Raw versus prepared records

Some buyers will pay more for raw records they prepare themselves. In the Spirit auction, the losing bidder offered more if it could obtain the raw data first and then anonymize it, Business Insider reported. That moves control of privacy to the buyer, which is a real trade-off. Our guide to how business records are de-identified explains what’s at stake.

Competition among buyers

Price discovery needs more than one bidder. In the Spirit auction, bids moved from $5 million to $5.2 million to $10 million as two buyers competed, according to Business Insider. A single unsolicited offer tells you what one buyer is willing to pay, not what your records are worth.

Why are the published ranges so wide?

Four reasons:

1. Different sellers. Distressed companies selling in a wind-down have little leverage. Operating companies can take their time and say no.
2. Different packages. Raw or prepared, exclusive or not, training or evaluation only: each changes the value.
3. Different buyers. Labs, training-data companies and simulated-workplace builders want different things and pay differently.
4. Few public comparables. Most licenses are confidential, so published figures are a small and uneven sample.

How does Cascade estimate the value of your records?

Our value estimator asks four questions: how many people work at your company, how many years of digital records you keep, which systems your team uses, and your industry. Team size and years of history set the starting point. The variety of record types you use adjusts it, because connected sources across several systems are worth more. The result is an indicative range, rounded so it doesn’t imply false precision.

For qualifying businesses, those with more than 20 employees and years of saved digital records, our partner estimate is typically $100,000 to $1 million or more. Smaller teams or shorter histories are handled case by case.

An online estimate only knows what you tell it. A reviewed estimate, after a conversation, also weighs how well your records connect, their quality, the rights and privacy constraints involved and current buyer demand. Even then, buyer interest and final value are never guaranteed.

How can you increase what your records are worth?

Some things you can do before any buyer is involved:

• Know what you have. Inventory your systems and how far back each goes.
• Keep records connected. Linkage is value. Prepare records so stand-ins stay consistent across sources rather than blanking out names.
• Clear rights early. Find the client, vendor and platform terms that limit what you can license.
• Decide exclusions up front. It’s easier to price a clean scope than to carve material out late.
• Create competition. Compare more than one buyer before you accept an offer.
• Negotiate the terms, not just the price. Exclusivity, permitted uses and retention are part of the value. Our guide to license agreement terms covers what to look for.

For the full process, start with how to license your business data to AI labs. If a buyer has already reached out, read the questions to ask before you say yes.

When you’re ready, get your indicative range. It takes about a minute, and nothing is sent until you choose to share it. Cascade works on a success fee only, so you pay nothing unless a license is signed.

Figures above are as reported by the cited publications. This guide is general information, not financial or legal advice.