Google, the primary client of Scale AI, is poised to sever its relationship with the AI data-labeling startup. This move follows reports that rival Meta has acquired a 49% stake in Scale, according to five insiders who spoke with Reuters.
The search engine giant had earmarked approximately $200 million for Scale AI this year for essential human-labeled training data, which plays a crucial role in the development of advanced technologies, including the AI models that drive Gemini, its competitor to ChatGPT, as indicated by one of the sources.
In light of recent developments, Google has engaged in discussions with multiple competitors of Scale AI this week to redistribute much of its data-labeling workload, sources revealed.
Scale AI’s potential loss of a major client comes as Meta’s investment values the company at $29 billion, a significant rise from its $14 billion valuation prior to the deal.
Despite the impending changes, Scale AI plans to continue its operations. CEO Alexandr Wang and several employees are expected to transition to Meta, yet the company’s reliance on a limited number of clients raises concerns about its future viability if it loses key partners like Google.
A spokesperson for Scale AI asserted that the company’s relationships with various major institutions and governments remain robust, emphasizing its commitment to safeguarding customer data. Specifics regarding Google’s partnership were not disclosed.
In 2024, Scale AI generated revenue of $870 million, while Google reportedly spent around $150 million on its services last year, according to sources.
Other prominent tech firms that currently utilize Scale AI’s services, including Microsoft, are also reconsidering their partnerships. Additionally, Elon Musk’s xAI is reportedly seeking to withdraw, as one insider noted. OpenAI, another key player, has reduced its dependence on Scale over the past few months, although it maintains it continues to utilize Scale as one of several data vendors, as confirmed by its CFO.
Companies competing with Meta in the AI landscape are wary that collaborating with Scale may reveal sensitive research priorities and strategic plans to a competitor. Engaging with Scale often requires sharing proprietary data and product prototypes that Scale’s personnel assist in annotating. With Meta’s new stake, the stakes are heightened, leading to concerns that a main competitor could gain insights into their operational tactics and innovations.
Google, Microsoft, and OpenAI have opted not to comment on the situation, while xAI did not respond to requests for input.
Rivals Seize Opportunities
Scale AI’s main revenue source stems from servicing generative AI model creators by providing access to a network of human annotators with specialized knowledge, ranging from historians to scientists, many of whom hold advanced degrees. These human contributions are integral for annotating complex datasets used to enhance AI models, with some annotations costing up to $100 due to increased demand for high-quality human examples.
While Scale AI also contracts with organizations such as self-driving vehicle companies and governmental entities, its primary revenue driver is partnerships with generative AI firms, insiders stated.
Google has been actively looking to diversify its data service providers for over a year. However, Meta’s recent actions have prompted Google to consider severing ties with Scale AI across its major contracts, according to sources. The structure of data-labeling contracts allows this transition to occur swiftly, as revealed by two informants.
This situation creates a significant opportunity for Scale AI’s competitors.
“The Meta-Scale deal signifies a pivotal moment,” remarked Jonathan Siddharth, CEO of Turing, a competitor. “Leading AI labs are now recognizing that neutrality is not just preferred, but essential.”
Labelbox, another rival, anticipates it will “probably generate hundreds of millions of new revenue” by the year’s end due to clients migrating from Scale, CEO Manu Sharma shared with Reuters.
Handshake, which focuses on building a network of doctoral experts, has experienced a tripling of work requests from top AI labs that compete with Meta, according to its CEO, Garrett Lord.
As many AI labs explore hiring in-house data-labelers to protect their data, Brendan Foody, CEO of Mercor, noted that his startup not only competes with Scale AI but has also developed technologies to automate the recruitment and vetting of candidates, helping AI labs rapidly expand their data-labeling capacities.
Established in 2016, Scale AI specializes in delivering large datasets and curated training data, which are vital for creating sophisticated applications like OpenAI’s ChatGPT.
The partnership with Meta is anticipated to benefit Scale AI’s investors, such as Accel and Index Ventures, as well as its current and former staff members.
Following the agreement, Scale AI’s CEO Wang is set to assume a leading role in spearheading Meta’s AI initiatives.
Meta is addressing the narrative that it may have lagged in the AI race, especially after its initial rollout of Llama 4 large language models in April did not meet performance expectations.
© Thomson Reuters 2025
(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)
