Meta's Scale stake triggers customer exodus
$1.1B/year Indeed for data labelers
Meta’s Scale deal turned data labeling from a supplier quality contest into a trust contest overnight. Frontier labs hand vendors their unreleased model prompts, eval sets, product roadmaps, and fine tuning workflows, so once Meta took a 49% stake in Scale and brought in Alexandr Wang, rivals had a clear reason to move spend elsewhere. In a market where the work itself is similar across vendors, perceived independence became the fastest way to win budget.
-
Google, described as Scale’s largest customer, planned to cut ties after the Meta deal because labeling vendors often see proprietary datasets and prototype products. OpenAI also wound down its work with Scale soon after. That is the core mechanism behind the reshuffle, customer data sensitivity, not just pricing.
-
The opening mattered because these buyers switch quickly. Handshake’s April 2026 update describes data labeling as a commodity service with low switching costs, and Prolific said the Scale disruption pushed a wave of work back into the market that benefited multiple vendors, including Handshake and Surge.
-
Scale still remained enormous, with estimated 2024 revenue of $1.5B and an implied $29B valuation at the June 12, 2025 strategic deal. But the episode showed that in human data, a vendor can lose share even while the market is growing fast, if one major customer views the vendor as too close to a direct competitor.
Going forward, the winners in human data will be the vendors that look neutral, plug directly into lab workflows, and expand beyond simple annotation into evals, red teaming, and custom participant pools. The more sensitive model development becomes, the more independence will function like product, not branding.