A source-backed map of 29 companies that buy, package, broker, or distribute business records for AI, including which routes may fit a 50-person company.
Open the market mapThree distinctions should shape outreach before payout math does.
Marketplaces list it. Brokers make introductions. Managed partners prepare and sell it. Only a few companies publicly say they buy it themselves.
An established company can still have a new data program with little public proof.
A 50-employee company may qualify, but archive depth, rights, productization, and buyer demand still determine whether the route works.
Choose three measures, then rotate the map. Each axis runs through the midpoint and points toward stronger evidence at its arrowhead.
| Company | Small-company fit | Evidence-backed standing | Trades and contractor fit | Action |
|---|---|---|---|---|
| Dawex | Unlikely | Credible | Weak | |
| AWS Data Exchange | Friction likely | Established | Weak | |
| Databricks Marketplace | Friction likely | Established | Weak | |
| Snowflake Marketplace | Friction likely | Established | Weak | |
| Nomad Data | Friction likely | Credible | Weak | |
| Turing Project Lazarus | Possible | Established | Weak | |
| Halluminate | Possible | Credible | Weak | |
| AfterQuery | Friction likely | Credible | Specialist | |
| Datarade | Yes | Credible | Weak | |
| Sunset | Possible | Credible | Specialist | |
| Mercor Data | Friction likely | Credible | Moderate | |
| Appen Enterprise Data | Possible | Established | Moderate | |
| Handshake Data | Possible | Established | Moderate | |
| Labelbox Prism | Possible | Established | Moderate | |
| OpenAI Data Partnerships | Possible | Established | Moderate | |
| Toloka AI | Possible | Established | Moderate | |
| Magma | Yes | Emerging | Weak | |
| idler / ShelfLife | Yes | Credible | Specialist | |
| Protégé | Yes | Credible | Specialist | |
| SimpleClosure Asset Hub | Yes | Credible | Specialist | |
| micro1 Data Partnerships | Possible | Credible | Moderate | |
| Defined.ai Partnership Programs | Yes | Established | Moderate | |
| Scale AI Data Partnerships | Yes | Established | Moderate | |
| Ooak Data | Yes | Emerging | Specialist | |
| Hookshot Data / Protege AI | Possible | Emerging | Moderate | |
| Troveo Business Data | Yes | Credible | Moderate | |
| Polyshares | Yes | Emerging | Strong | |
| DataFactor Labs | Yes | Emerging | Strong | |
| Petrarch | Yes | Emerging | Strong |
Filter the market, compare up to three organizations, or open a company’s complete source-backed dossier.
The clearest public principal-buyer offer for specialized small-company operating histories.
Broad system coverage and a published 20-to-50 employee tier; payment and buyer identity still need contract proof.
A direct fit for industrial, construction, logistics, and bankruptcy-derived archives, with transaction claims that remain young.
A mature AI-data marketplace with serious governance, but admission does not guarantee a sale.
Strong field-service relevance and deep packaging capability, but the program prefers companies with at least 40 employees.
A real content-licensing operator extending into business archives with repeat-sale economics.
A functioning benchmark built from a retailer's operating history, with an explicit 10-to-100 employee supplier target.
Credible specialist route for healthcare, media, audio, motion, and spatial archives; broad default rights require narrowing.
A funded wind-down service licensing source code now and workspace data in beta, with clear seller and buyer routes.
Turns Gmail, Slack, Jira, Notion, Drive, and SharePoint histories into anonymized digital twins for AI labs.
Open to businesses of all sizes, but the seller must turn its archive into a governed data product.
A narrow route for existing AI-agent traces and corrections, paid through revenue share when data is licensed.
Mature delivery infrastructure and payment on acceptance, with limited public proof for this specific program.
Established parent and direct-purchase language, with a published delivery-based wire-payment milestone that still needs contract confirmation.
The clearest published compensation framework for 30-plus-employee suppliers with documented operating workflows.
A serious acquisition route for private codebases and engineering artifacts, not ordinary contractor records.
An established data platform with an emerging licensing program; default exclusivity and vendor-stated payouts need proof.
Interesting for resolved ITSM histories, but the buyer, licensee, and unconditional payment obligor remain unresolved.
A credible wind-down operator that says it now buys and licenses company archives, with unresolved purchase-versus-license wording.
A specialist route for finance and consulting workflows, with open intake but no independently verified data-license payout.
The only frontier lab found with an open general intake for private datasets; commercial terms remain entirely bespoke.
A mature data operator now hiring specifically to acquire proprietary company data and deliver it to major AI customers.
A newly explicit enterprise-data buyer with serious capital, but no public proof yet of a completed company-archive purchase.
Broad integrations and frontier-lab access, with a confirmed 2026 supply-chain incident that raises the diligence bar.
Enterprise distribution infrastructure for a finished data product, not a buyer of ordinary SMB archives.
Enterprise distribution for a Databricks-native product; commercial transactions happen off-platform.
A strong distribution channel for Snowflake-native data vendors, with significant packaging and operating burden.
A direct introduction channel for companies that already operate as data vendors; it does not package raw archives.
Software for organizations that operate governed exchanges, not a practical direct route for one SMB seller.
Model-lab procurement
OpenAI is the only frontier lab in this review with a general public intake for private datasets. Other labs buy selectively, work through partners, or publish no operational-data route.
A public form accepts proposals for large, private datasets. Pricing and deal structure are private.
The only frontier-lab door in this review that a 50-person company can approach without an introduction.
Public evidence shows negotiated acquisitions, procurement roles, or licensed content. None publishes a general operating-data seller form.
A strong archive likely needs a broker, existing relationship, or targeted introduction.
These companies publish partner networks, collection roles, or data-factory programs rather than a practical archive-seller route.
Approach through an established data partner unless a buyer requests the archive directly.
Public materials discuss training, customer data, licensing, or privacy, but do not expose a usable seller path for company archives.
Do not treat these labs as active prospects without new direct evidence.
Research watchlist
These names remain outside the scored dataset until identity, seller access, or category fit is clear.
Corpus / datacorpus.ai, Nyne, and MonterraAI were excluded after identity or category validation failed.
Each measure uses four named levels. Unknown is a separate state and removes a company from any view that needs the missing measure. No composite score is calculated.
| Measure | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Evidence-backed standingHow much independent operating and transaction evidence is visible. | Unclear | Emerging | Credible | Established |
| Commercial commitmentHow close the provider gets to taking principal buyer risk. | Infrastructure | Listing / broker | Managed licensing | Principal buyer |
| Small-company fitPublic evidence that an operating company with about 50 employees can qualify. | Unlikely | Friction likely | Possible | Yes |
| Trades and contractor fitHow directly the offer fits field-service and contractor records. | Weak | Specialist | Moderate | Strong |
| Company employee scaleEstimated company headcount, not the size of the data-provider program. | 1–25 | 26–200 | 201–1,000 | 1,000+ |
| Capital scaleDisclosed funding or public-company scale. | Bootstrapped / under $1M | Under $10M | $10M–$100M | $100M+ / public |
| Security evidenceThe strength of public assurance evidence, not a claim about actual security quality. | Not established | Controls described | Attestation visible | Program-level assurance |
| Payment certaintyHow clearly public terms establish the payment trigger and obligor. | No paid route | Downstream contingent | Acceptance / milestone | Direct obligation |
| Data-handling depthHow much of extraction, cleaning, packaging, and delivery the provider performs. | Introduce / list | Host / deliver | Package / license | Extract / anonymize |
The information was not public or not defensible at the time of review. It does not mean the company lacks the capability.
Company payroll or a conservative company-size estimate. Contributor networks, marketplace experts, and program participants are kept separate.
Programs, funding, headcount, and terms move quickly. Recheck every commercial and security claim before making an introduction.
