Expert data marketplaces
Placement: a judgment drawn from the evidence, open to correction (dispute it).
The model
- Definition
- Platforms recruit and pay vetted domain experts (engineers, doctors, lawyers, linguists) on a marketplace/gig basis to produce or grade the specialized human demonstration, evaluation, and reinforcement-learning data that AI labs use to post-train frontier models, taking a margin on the labor.
- In one line
- Platforms pay doctors, lawyers, engineers and other experts to create and grade the data that trains AI.
- What compounds
- Better models need harder data, which raises the value of scarcer expertise.
- Who accelerates
- Experts earn from knowledge that was previously unpaid; the work may train their own replacements.
- What it displaces
- Low-skill data labeling.
- Stops working when
- When a lab-side ownership stake makes a vendor untrustworthy to rival labs (Scale AI lost customers after Meta's investment).
Evidence
Companies 4
Evidence items 7
Grades 1A 6B
Latest dated 2026-08-20
Counter-evidence 1
Evidence items 7
Grades 1A 6B
Latest dated 2026-08-20
Counter-evidence 1
The three labels
2. Rides · loop gain positive. Each step of the curve makes its own loop stronger.
Owning the scarce human · money lands in original human signal, expertise, likeness.
Human-signal loop · more machine output makes original human signal scarcer and worth more.
Same rung and family
Named companies 4 companies · grade per item
| Company | Claim | Date | Gr | Source |
|---|---|---|---|---|
| Mercor US · startup · best B |
Annualized revenue crossed $1 billion in February 2026 and $2 billion by June 2026 | 2026-07-09 | B | Forbes |
| In talks to raise $500 million at a $20 billion valuation (expected to close later in July 2026), up from a $10 billion valuation in September 2025 on a $350 million prior round | 2026-07-09 | B | Forbes | |
| Surge AI US · scaleup · best B |
2024 revenue of $1.2 billion, profitable and bootstrapped with no VC funding to date; named customers include Google, Meta, Microsoft, Anthropic, and Mistral, and it helped train Gemini and Claude | 2025-09-17 | B | Forbes |
| Estimated valuation around $24 billion, with a pending raise reported at roughly $1 billion at a $30 billion valuation | 2025-09-17 | B | Forbes | |
| Micro1 US · startup · best B |
Grew gross annual run rate from $100 million to $500 million over eight months, reaching the $500 million milestone in August 2026; net run rate estimated at $150 to $200 million after payouts to experts | 2026-08-20 | B | TechCrunch |
| Raised a Series A at a $500 million valuation around September 2025, with reports of a subsequent higher-valuation round unconfirmed in specifics | 2025-09 | B | TechCrunch | |
| Karya India · startup · best A |
$4.42 million in wages directly distributed to Karya's worker network as of July 2026, with 200,000+ workers onboarded on the platform, and a 6.8x year-over-year increase in wages distributed from 2024 to 2025; datasets built in 115+ languages across 28 Indian states and deployed in 7 countries | 2026-07 | A | Karya |
Counter-evidence Reversals, failures and disputes
| Within weeks of Meta's $14.3 billion investment in Scale AI closing, Google, OpenAI, and xAI cut or significantly reduced their Scale AI contracts, unwilling to route proprietary training and evaluation data through a vendor now half-owned by a direct frontier-AI competitor; that business shifted to Mercor and Surge AI, and Scale AI's own 2026 revenue guidance fell to about $1 billion, down from the 2025 estimate of $2 billion | 2025-06 to 2026 | Value Add VC |
Sources 5
Forbes forbes.com · 2026-07-09
Forbes forbes.com · 2025-09-17
TechCrunch techcrunch.com · 2025-09
Karya karya.in · 2026-07
Value Add VC valueaddvc.com · 2025-06 to 2026
Forbes forbes.com · 2025-09-17
TechCrunch techcrunch.com · 2025-09
Karya karya.in · 2026-07
Value Add VC valueaddvc.com · 2025-06 to 2026