How Much Is Company Data Worth to AI? The 2026 Price Evidence
Published offers for company data run from $10,000 to $4 million and above, but documented payouts are mostly far below the headlines. A wind-down startup netted about $25,000 for its code, one direct buyer's four published deals ran $110,000 to $450,000, and the median documented public sale across the market is about $31,600 (public-record compilation, checked November 2026).
MyCompanyDataWorth.com helps U.S. companies with 20+ employees license their operating records — email, chat, CRM and project data — to AI labs and training-data buyers. Owners keep ownership of their data; valuation starts with a metadata-only manifest, and no raw data is shared to get an estimate.
What do buyers publish?
Buyers publish tiers and ranges, not price lists. micro1's three tiers ($100K–$500K, $500K–$3M, $3M+) and Mercor's per-industry listings ($20K–$1.5M) are the market's reference points, and both describe programs — what a qualifying company may be offered after review — not what any company will be paid.
| Buyer / program | Type | Published payout | Published fit | Source, checked |
|---|---|---|---|---|
| micro1 | Training-data company | $100K–$500K (qualified partnerships); $500K–$3M (large-scale or ongoing); $3M+ (highly unique data) | 30+ full-time staff; mainly English records; US first, then UK and Canada | micro1 data-partnerships page, Nov 2026 |
| Mercor | Training-data company | $20K–$1.5M by industry across its listings | Varies by industry — some listings from 5 staff, most 20+ | Mercor referral listings, compiled Oct–Nov 2026 |
| Handshake AI | Training-data company | $100K–$4M+ | 20+ staff, 3+ years (its referral page) | Handshake AI program pages, Nov 2026 |
| Scale AI | Training-data company | $10K–$1M+ | Broad; range labelled “illustrative” by Scale | Scale AI data-partnership page, Nov 2026 |
| Polyshares | Direct buyer | $100K–$2M+ stated typical; four published licenses of $110K–$450K | No published headcount minimum | Polyshares site and track-record page, Nov 2026 |
| Endless Data | Buyer / program | Up to $4M; no typical range published | Around 30+ staff in published material | Endless Data site, Nov 2026 |
| SourceX | Buyer (exclusive licenses) | Not published — one-time fee, exclusive license structure | Best fit 50+ full-time staff | SourceX supplier/referral terms, Nov 2026 |
| SimpleClosure AssetHub | Wind-down platform | $10K–$100K for closing-company archives | Companies shutting down; smaller archives | AssetHub reporting, Oct–Nov 2026 |
| Spirit Airlines archive (market evidence, not a program) | Bankruptcy auction | $10M winning bid (Google); $7.5M backup bid (Mercor); customer/loyalty data excluded | Large bankrupt enterprise — not typical of any program above | Court record, Aug 2026 |
{FIGURE_DISCLAIMER}
Read this before the table: these are published ranges and stated figures, not completed-deal prices. AIData.Broker's compilation of public prices puts the median published offer at about $280,000 per company and the median documented public sale at about $31,600 (checked November 2026). Most real outcomes sit far below the ceilings in this table. That gap is exactly why a written bid on your own manifest — not a table — is the only number that counts.
The full, maintained table lives in the Price Index; it is reproduced here in the report because this report is where the numbers are argued with, and the Index is where they are kept current.
Why is the documented record so much lower than the published ranges?
Because published ranges describe the top of a program and completed deals include everyone else: smaller archives, partial datasets, wind-downs and non-exclusive terms. The market is also young — few completed sales are public at all, and the ones that are skew small.
Three documented anchors keep expectations calibrated. A shut-down crypto startup received about $25,000 for its code base, with about $15,000 more pending for workspace data (reported October 2026). Polyshares' four published licenses — real, completed, perpetual exclusive deals — ran $110,000 to $450,000 each. And at the very top, Google's $10 million court-approved bid for Spirit Airlines' de-identified archive shows what a vast, unique enterprise dataset fetches at auction — against a $7.5 million backup bid from Mercor for the same records, which is itself the cleanest public evidence that competition moves price.
An owner reading only ceilings will misprice their own archive in both directions: expecting a ceiling disappoints, and accepting a first offer without competition undersells. The evidence supports one behaviour — get the manifest in front of every buyer whose published floor you meet, and let written bids replace ranges.
What moves a company up its range?
Team size first (more people produce more connected history), then years of records, the number of systems included, industry demand, and exclusivity. The factors page takes each in turn, and the industry table shows the published ranges by sector.
Indicative market reporting puts operating companies at roughly $200K–$500K for 20–49 staff, $400K–$900K for 50–149 and $800K and up for 150+ — indicative patterns from market guides, not offers, and consistently below the program ceilings above. The estimator on this site is anchored to the conservative end of that pattern and clamped to the $100K–$5M band the underlying programs publish.
Estimate your range
The estimator maps your staff band, years of records, number of connected systems and industry onto the ranges buyers publish. It is part of the fit check above: pass the floor and your illustrative range appears with the factors that moved it, what would move you up a band, and the disclaimer below — before any booking is offered.
This is an illustrative estimate based on ranges buyers publish. It is not an offer, quote, or valuation of your company. Actual value — if any — is set by buyer bids on a metadata-only manifest of your records, after rights and de-identification checks. Many companies will not qualify or will receive offers below published ceilings. Nothing is owed by either side until a license is signed.
Start the fit check to see your rangeStart with your own fit check
Going to one buyer alone means negotiating against a single price. Start here instead: run the fit check and book the 7-minute interview, and one metadata-only manifest can be reviewed by the buyers you fit — so you compare real offers before you sign anything.
Questions owners ask first
Is this a sale or a license?
A license. You keep ownership of your records and keep using them in your business. A buyer pays for the right to use a de-identified copy for AI training, on terms — length, exclusivity, permitted use — that you approve before anything is signed.
Do I have to send my data to get an estimate?
No. The estimate and the first call use only what you describe: systems, years and rough counts. Buyers bid on a metadata-only manifest. Full data moves only after you sign a license directly with a buyer, and personal data is de-identified before transfer.
What does it cost me?
$0 upfront. There is no retainer and no listing fee. We are paid a success fee only if a deal closes — the terms are stated in writing before any submission is made.
How long does a deal take?
Buyers’ published timelines put most deals at six to twelve weeks from first contact to payment. That range comes from the buyers’ own pages, checked November 2026, and individual deals vary.
What if no buyer bids, or I don’t like the offers?
You walk away and owe nothing. The manifest stays yours. Comparing offers before signing anything is the entire point of putting several buyers in front of the same manifest.
Do you buy personal or phone data?
No. This is for companies: the operating records of a business with 20+ staff. We do not buy personal, browsing, phone or survey data, and buyers exclude customer and employee personal data from these deals.