What is company data worth? Prices, buyer programmes and deals in 2026

Since 2026 there has been a market for internal company data with published prices. Data vendors that supply AI labs openly solicit email archives, ticket systems, CRM and ERP data on their websites and quote specific amounts for them. The register Who Sells Data lists 170 providers of training data and training environments.^[16] This article compiles prices, documented deals and funding rounds with sources and shows how buyers value a dataset.

What do buyers pay for internal company data?

Depending on size, between US$10,000 and several million. For a mid-sized company with 50 to 250 employees, the published figures are in the six-figure range.

The most detailed public scale by company size is published by Replay, a New York buyer specialising in company data.^[7] Handshake AI, which supplies OpenAI among others, offers a calculator: for 100 employees and 10 years in operation it gives US$383,000 to US$897,000.^[5][27]

Which deals involving company data are documented?

Publicly documented are mainly cases from the US: the auction of Spirit Airlines' data and a growing number of sales by wound-up start-ups.

Spirit Airlines: US$10 million for emails and Teams messages

In the US airline's bankruptcy, around 100 million emails, 500 million Microsoft Teams entries, code and operational data were auctioned. Google won the auction with US$10 million against Mercor with US$7.5 million; micro1 then put forward US$12.5 million, combined with additional privacy commitments. Customer profiles are excluded.^[1] The court approval hearing was postponed four times and is scheduled for 6 November 2026.^[2]

Wound-up start-ups: US$10,000 to US$100,000 per company

The US wind-down service provider SimpleClosure supported almost 100 data sales in the twelve months to April 2026, with proceeds of US$10,000 to US$100,000 per company.^[4] There are now more than 150 deals; according to CEO Dori Yona, 47% of SimpleClosure's deals include a sale of code or data. SimpleClosure expects payouts of more than US$10 million to start-ups by the end of 2026.^[3] One example: a crypto start-up around two and a half years old received about US$25,000 for its code and expects around another US$15,000 for data from Google Workspace, Notion, Figma and Slack.^[3]

Ooak Data: deals from US$20,000

The buyer Ooak Data openly states its previous deals: the smallest was US$20,000, the largest several hundred thousand US dollars. Ooak buys from operating companies, companies being wound up and companies that have already closed; according to Ooak, a liquidator or insolvency administrator can also sign.^[12]

micro1: more than 50 transactions in 45 days

In its filing in the Spirit proceedings, micro1 stated that it had completed more than 50 data transactions in the preceding 45 days, without naming sellers or prices.^[1] In October 2026 the company announced that it would spend US$1 billion on company data over twelve months, financed by Citi and Hercules Capital.^[24]

What do the data buyers' programmes pay?

Most programmes quote US$100,000 as the entry point and US$1 million to US$5 million as the cap. Some pay once, others additionally pay on every further licence.

There are also marketplaces without published prices. At data.cool, buyers bid on anonymous data sheets; the seller pays a 12% fee, the buyer a 4% surcharge, plus a 10% deposit and 10% retention until acceptance.^[15] The register Who Sells Data, which usefully complements the overview of programmes, is operated by the vendor Secludy AI, which by its own account may itself sell to listed companies.^[16]

How much money is flowing into the market for training data?

Billions. The largest data vendors are valued at US$10 billion to more than US$29 billion, and individual AI labs have discussed spending more than US$1 billion a year on training environments alone.

The money is increasingly going into data from real workflows. Surge AI held talks about a valuation of at least US$25 billion.^[20] In July 2026 Mercor acquired Deeptune, which had rebuilt hundreds of business applications for training and had previously raised US$43 million.^[22] micro1 was valued at US$500 million in a US$35 million Series A in 2025.^[1] Realset AI raised US$10 million for training data from real workflows in September 2026,^[26] and the wind-down service SimpleClosure, which now brokers data sales, raised US$15 million in a Series A.^[25]

For a single training task, labs usually pay US$200 to US$2,000, and contracts between providers of training environments and labs are often in the six- to seven-figure range per quarter.^[17] According to Wired, OpenAI together with Handshake AI asked contractors to upload real work products from previous jobs, as files rather than summaries.^[27] Both show how scarce real work data are.

How do buyers value a dataset?

By headcount, years in operation, systems and data volume. Several buyers state these criteria explicitly on their websites.

How is company data paid for?

As a one-off payment, as an upfront payment with a revenue share, or as a purchase price via a marketplace.

• One-off payment: Turing pays once, up to US$1 million, with no ongoing obligation.^[8]
• Upfront payment plus revenue share: Replay and Labelbox pay on signing and give the seller a share of every further licence.^[7][10]
• Volume pricing: pre.dev pays for code by tokens, for example US$6,000 for a repository with 20 million code tokens and 100 million tokens of Git history.^[14]
• Marketplace: at data.cool buyers bid on the dataset and the platform retains fees; Grepped pays out after the sale.^[15][13]

What lowers the value of a dataset?

Unclear rights, gaps in the systems and content that cannot be reliably anonymised.

• Unclear rights, for example data processed on behalf of third parties, or licence agreements that rule out passing data on.
• A high share of sensitive data such as personnel files or health information that must be excluded.
• Gaps: individual systems are missing, mailboxes have been deleted, periods are incomplete.
• Legal uncertainty, such as unresolved approvals, delays the deal.

How data assets are treated on the balance sheet is covered in Data assets on the balance sheet. The buyer side is covered in Selling company data for AI training.