Disclosure: some links on this page are referral links. DataBuyout may earn a fee if you request an offer from Polyshares, at no cost to you. How we make money
DataBuyoutRequest an offer
Pricing · Updated

Value of Data: What Company Data Is Worth to AI Buyers

The value of data held by a company is now set by real buyers, not guesswork. Reported deal sizes, the factors buyers actually price on, and how to estimate your own number before you talk to anyone.

The short answer

Reported deals run from $10,000 to $100,000 (Forbes, Apr 16, 2026) for startups selling archives during a shutdown, and $100,000 to $2 million and up for direct licenses from operating companies. At the top end, Google bid $10 million for Spirit Airlines' internal emails and Teams chats (the sale awaits court approval, hearing Nov 6, 2026). Buyers price on years of history, how much work is done in writing, how many connected systems you have, and team size.

Data valuation: reported deal sizes for company data

Deal typeReported rangeExample
Shutdown archive sale$10K to $100K (Forbes, Apr 16, 2026)SimpleClosure deals reported in April 2026
Single startup, rich historyHundreds of thousandsCielo24's Slack, email and Jira history
Direct license, operating company$100K to $2M+Polyshares published typical deals
Bankruptcy auction$10MSpirit Airlines' emails and Teams chats, 2026 (winning bid; approval hearing Nov 6, 2026)

What buyers price on

FactorWhy it matters
Years of historyLong, continuous records show how work changes over time
Work in writingDecisions argued out in threads are what agents learn from
Connected systemsChat plus tickets plus CRM shows whole workflows
HeadcountMore people means more handoffs and more variety
Data qualityReal discussion, not bots, templates or empty channels
Clean ownershipRecords you clearly own, with no client data mixed in

Why history beats volume

A 20-person company with eight years of detailed Slack and ticket history can be worth more than a 200-person company with two years of thin records. Buyers want sequences of real work: a problem raised, discussed, handed off and solved.

Which data types are worth most

In our calculator, team chat carries the highest weight, followed by phone and call recordings, source code, email and meeting recordings. Product analytics, HR data and photo archives add context but carry less weight. See every category and system.

Estimate yours

The value calculator gives an indicative range from your systems, history and headcount in two minutes. For a real number, request an offer from Polyshares.

Published buyer ranges, side by side

Several buyers now publish what they pay, which makes it easier to sanity-check any offer. These are the buyers' own ranges or ranges reported by the press, not guarantees, and each one measures something slightly different: a per-company payout, a per-partnership value or a typical deal.

Published payout ranges from company data buyersLog-scale bars from $10K to $10M showing each buyer’s published or reported range: SimpleClosure $10K to $100K, Scale AI $10K to $1M plus, Turing up to $1M, Polyshares $100K to $2M plus, micro1 $150K to $2M plus, Handshake AI $100K to $4M.$10K$100K$1M$10MSimpleClosure (typical, per Forbes)Scale AI (illustrative)+Turing Lazarus (up to)Polyshares (typical)+micro1 (typical)+Handshake AI (typical)
Published or reported payout ranges on a log scale. A "+" means the buyer says deals can go higher. Turing publishes only a ceiling. Sources and dates are listed below the chart and at the end of this page.
BuyerPublished or reported rangeWhat the range coversSource
SimpleClosure$10,000 to $100,000Typical payment per shutting-down company for code and workspace dataForbes, Apr 16, 2026
Scale AI$10K to $1M+"Illustrative value per data partnership, scaling with cadence"; prefers companies with 40+ employeesScale AI, read Oct 31, 2026
Turing (Lazarus)Up to $1MA single payout for operating or winding-down companies; sample assessed usually within a weekTuring Lazarus program page (read Oct 31, 2026), read Oct 31, 2026
Polyshares$100K to $2M+Typical license deals; offer in one to two weeks, no fee to the seller, paid Net 30–60 after approval and anonymizationPolyshares published terms (read Oct 31, 2026), read Oct 31, 2026
micro1$150K to over $2MTypical payments, with tiers of $500K–$2M and $2M+; aimed at companies with 30+ employeesmicro1, read Oct 31, 2026
Handshake AI$100K to $4M"Qualified partnerships," depending on the data; de-identified, under NDAHandshake AI, undated

The spread is the point. A small startup selling an archive on the way out and a 200-person operating company licensing ten years of records are in different markets, even when they use the same tools.

What recent deals actually paid

Published ranges are marketing. Reported deals are the better check, and a few are now public:

  • A small startup's code and workspace. MIT Technology Review reported on Oct 13, 2026 that founder Keelan Miskell received about $25,000 for his startup's code, with about $15,000 more pending for its workspace data (MIT Technology Review).
  • A 13-year company. Forbes reported on Apr 16, 2026 that cielo24, with 13 years of history, received "hundreds of thousands of dollars" for its records (Forbes).
  • Wind-down volume. By April 2026, SimpleClosure had processed nearly 100 deals and recovered over $1 million for companies, per Forbes. By October, MIT Technology Review reported more than 150 deals through its Asset Hub, with the company expecting to return more than $10 million by the end of 2026, and 47% of its deals including code or data.
  • The top end. Google was the winning bidder at $10 million for Spirit Airlines' enterprise data at an August 2026 bankruptcy auction, ahead of Mercor at $7.5 million. That sale is not approved yet: the court hearing is set for Nov 6, 2026 (Stretto).

Read together, the reported deals sit inside the published ranges, with most small shutdown sales at the low end and only large or unusually rich datasets reaching seven figures.

How buyers set the price

No buyer publishes a formula, but the ones that describe their process point to the same inputs.

Size, age and how much was used

Sunset, another wind-down firm, told MIT Technology Review it values data by headcount, years of operation, data volume and how heavily each tool was used. Forbes reported that Sunset prices at levels similar to SimpleClosure, depending on a company's size, age and "data richness."

Linked records beat loose ones

Buyers pay more when records connect. Forbes gave the example of a Jira ticket tied to a code commit being worth more than either one alone, because together they show a task from request to finished work.

Industry

Healthcare and finance data command a premium, per the same Forbes report, because few companies in those fields can share records at all.

Fit with what the buyer is building

Surge AI told MIT Technology Review that one company's Slack can be worth more than another company's Slack, email and calendars combined, and that buyers quote very different numbers for the same data because each is pursuing its own research thesis. That is the strongest argument for getting more than one quote.

Why company size changes your number

Some buyers set a minimum. Scale AI says 40 or more employees is its preferred starting point, and micro1's program is aimed at companies with 30 or more. Below that, most published deals come through wind-down platforms at the $10,000 to $100,000 level Forbes reported.

Size isn't everything, though. A small team with many years of detailed, connected records can still be worth more than a larger company with thin history, which is why the earlier sections on history and data types matter as much as headcount.

Who actually gets the money

In an operating company, a license payment goes to the company like any other revenue. In a shutdown it usually doesn't reach the founders: MIT Technology Review reported that proceeds from these sales typically flow to senior creditors and investors first. If you're winding down, budget the sale as a way to pay obligations, not as a personal payout.

Supply is also growing. The same report cited Carta and AngelList figures showing startup shutdowns rising between 26% and 56% a year, so more archives are coming to market.

How to get a real number for your data

  1. List your systems. Chat, email, tickets, code, CRM, support and call recordings, with the years each one covers.
  2. Measure the history. Message counts, ticket counts and the earliest date in each system. The system guides show where to find these.
  3. Confirm you own it. Check customer contracts, vendor terms and IP assignments, and separate anything you hold for clients.
  4. Get an indicative range. The value calculator gives a rough number from your systems, history and headcount.
  5. Ask more than one buyer. Compare the net payout after any fee, when you get paid (Polyshares says Net 30–60 after approval and anonymization; micro1 says within 30 days of invoicing, after a review of about 30 days) and whether you keep your data under a license.

More in this guide series

Related step-by-step guides

Common questions

How much is company data worth to AI companies?

Reported deals range from $10K to $100K (Forbes, Apr 16, 2026) for shutdown archives and $100K to $2M+ for direct licenses. Large bankruptcies have drawn bids of $10M.

What makes company data more valuable?

Years of continuous history, work done in writing, several connected systems, and clear ownership.

Is Slack data worth more than email?

Per unit, team chat usually carries more weight because it shows internal decisions and handoffs. Most deals include both.

Can a small company's data be valuable?

Yes, if it has long, detailed records. History and depth matter more than headcount.

Sources
  1. Fast Company: Shuttered startups are selling old Slack chats and emails to AI companies
  2. Stretto: Spirit Airlines data sale hearing moved to Nov 6, 2026 (Oct 14, 2026)
  3. Polyshares published terms
  4. Gizmodo: Failed companies are selling old Slack chats and email archives
  5. MIT Technology Review: AI labs are buying data from dead startups to train models (Oct 13, 2026)
  6. Scale AI: Data Partnership program (read Oct 31, 2026)
  7. Turing Lazarus program page (read Oct 31, 2026)
  8. micro1: business data program terms (read Oct 31, 2026)
  9. Handshake AI: employer data program (undated blog post, read Oct 31, 2026)

What's yours worth?

Run the calculator, then get a written offer.