How Buyers Compare: Direct Programmes, Marketplaces, Wind-Down Platforms and Brokers

Companies that buy business data for AI fall into four routes: direct buyer programmes (micro1, Mercor, Scale AI, Handshake AI, Polyshares, SourceX) that publish ranges of $10K to $4M+; marketplaces (Datarade, Defined.ai) where sellers set prices and pay 15-30% commission; wind-down platforms (SimpleClosure, Sunset) paying a reported $10K-$100K for closing companies' archives; and brokers that run one manifest in front of several buyers for a success fee or buyer-paid referral.

See the Full Buyer Database

Route 1: direct buyer programmes

One buyer, one price, fastest path when you clearly fit one programme. Published floors differ sharply - Handshake AI at 20+ staff, micro1 at 30+, SourceX at 50+ at peak, Polyshares with no minimum - so 'the market' has no single door. Exclusivity is the trap to read first: Polyshares is exclusive by default and its published deals are perpetual; SourceX builds exclusivity for AI training into its structure.

Route 2: marketplaces

Datarade publishes the only transparent seller fees among the routes (30% commission-only down to 15% on a $12,000/yr plan). Marketplaces suit companies that can package and describe a dataset themselves; Defined.ai's Data Partner track publishes no per-company range.

Route 3: wind-down platforms

If the company is closing, its archive still sells: SimpleClosure reports 150+ deals at a reported $10K-$100K each. This route is priced for speed, not for maximum value.

Route 4: brokers and the multi-buyer manifest

A broker's manifest puts the same records in front of several buyers. DataSupply publishes its banded success fee (20% of the first $50K of a license sliding to 8% above $1M) and illustrative worked examples; AIData.Broker publishes free introductions and a 12%-of-first-$250K sale fee. This site's route is the same shape: one interview, one manifest, offers compared on price, term and exclusivity. Published ranges are each buyer's own figures, checked on the date shown. They are not an offer, quote, valuation, or guarantee of payment. AIData.Broker's compilation of public figures puts the median published offer at about $280,000 per company and the median public sale at about $31,600 — most companies should expect an actual offer, if any, only after a buyer reviews a metadata-only manifest, and it may be below published ranges or none at all.

Timelines, honestly

Buyers' published timelines put most deals at six to twelve weeks from first contact to payment (Sell My Data To AI's compilation of buyer pages, checked 2026-11-10). Handshake publishes 7-14 business days for its process; Mode's programme is summarised at roughly three months. Anyone promising days is selling something other than a data license.

Frequently asked questions

Is it better to sell direct to one buyer or use a multi-buyer manifest?

Direct gives one buyer's price for your records. A multi-buyer manifest gives competing offers on the same records, which is why published medians for offers ($280K) sit far above medians for completed public sales ($31.6K): most sellers who took the first offer never saw a second one. Neither route guarantees any offer.

Are exclusive licenses worth more?

Compilations of buyer terms (ddiy.co, 2026) put exclusive deals at roughly 2.5 to 5 times non-exclusive ones - which is why Polyshares' perpetual-exclusive deals and SourceX's exclusive-for-AI-training structure deserve a lawyer's review before signing, not after.

Sources - checked 2026-11-10: buyer programme pages linked from the Buyer Database; route and fee figures from the programmes' and platforms' own published pages as compiled Oct-Nov 2026.