Google to End Contract with Scale AI Amid Meta's Stake Acquisition

Google, the primary client of data-labeling firm Scale AI, is reportedly planning to terminate its contract. This development follows Meta's announcement of acquiring a 49% stake in the startup, which has significantly increased Scale's valuation. Google had initially planned to allocate approximately $200 million to Scale this year for human-labeled training data, essential for training its Gemini AI model.

Sources indicate that Google has already initiated discussions with other companies to transfer a substantial portion of its workload away from Scale AI. This shift coincides with Meta's investment, which has elevated Scale's value from around $14 billion to $29 billion. Despite the potential loss of a major client, Scale AI intends to continue operations. The company's CEO, Alexandr Wang, and a few employees will transition to Meta, while the rest of the team will remain. However, the departure of Google, one of its largest clients, could significantly impact Scale's revenue.

Scale AI reported $870 million in revenue for 2024, with Google contributing approximately $150 million of that amount. Other leading tech firms are also reassessing their relationships with Scale AI. Microsoft is reportedly reconsidering its collaboration, and Elon Musk's xAI plans to discontinue its use of Scale's services. OpenAI, which had reduced its reliance on Scale several months ago, still utilizes the company as one of its multiple data vendors.

Concerns among rival AI developers center on the potential exposure of strategic plans and technical details. When companies contract Scale, they often share prototype tools and private datasets for labeling. With Meta now owning nearly half of Scale, competitors fear that their research roadmaps and technical information could be accessed by a major rival.

Scale AI's primary revenue stream comes from services for generative AI firms, connecting these companies with experts to "post-train" models. As AI systems have become more advanced, the demand for detailed human input has surged. Scale also provides data labeling services for self-driving car manufacturers and U.S. government agencies, clients expected to remain. However, the majority of its income is derived from partnerships with generative AI developers.

Google has been diversifying its data-labeling services for over a year, and Meta's recent investment has accelerated this transition. Built-in clauses in many contracts allow for a rapid shift away from Scale. This change presents opportunities for Scale's competitors. Leading AI labs are increasingly prioritizing neutrality, recognizing it as essential rather than optional. Some AI labs are opting to hire in-house data-labelers to ensure the security of their information. This trend reflects a growing concern among AI developers about the potential risks associated with outsourcing data-labeling services to third-party providers.