Google Ditches Scale AI After Meta’s Surprise Stake Grab

In a major shake-up within the AI infrastructure world, Google is planning to sever ties with Scale AI — its largest AI data-labeling vendor — following the announcement that rival Meta is acquiring a hefty 49% stake in the company. According to five insiders who spoke to Reuters, the move could reshape the competitive landscape for AI data services and spark a scramble among competing vendors to fill the gap.

The decision comes as a blow to Scale AI, which had been set to earn approximately $200 million from Google this year alone for human-annotated training data. That data fuels the development of advanced AI models like Google’s Gemini, its flagship competitor to OpenAI’s ChatGPT. Sources say Google has already begun reaching out to alternative vendors this week, signaling an urgent shift in its AI data strategy.

The Meta-Scale deal, which values Scale AI at a staggering $29 billion — double its previous valuation of $14 billion — has sent ripples through the AI community. As part of the deal, Scale AI’s high-profile CEO, Alexandr Wang, will jump ship to Meta, taking several key employees with him. Wang is expected to take a prominent leadership role overseeing Meta’s AI initiatives, a move that underscores how aggressively the social media giant is now chasing AI dominance.

A Risky Bet for Scale AI

While Scale AI insists its business remains strong and secure — working with governments, enterprises, and top AI labs — the fact remains that much of its revenue depends on a small number of major clients. Losing a customer the size of Google is a seismic hit. The company pulled in an impressive $870 million in revenue in 2024, with Google alone contributing around $150 million last year.

Compounding the issue, several other big tech players appear to be backing away from Scale. Microsoft, another key client, is reportedly scaling down its relationship. Elon Musk’s xAI is also exploring alternatives, while OpenAI quietly reduced its business with Scale months ago, though its CFO confirmed the company still works with Scale as one of several vendors.

AI Labs Fear for Their Secrets

At the heart of the fallout is a growing concern within AI labs about data security and competitive exposure. By outsourcing to Scale AI, companies often share highly sensitive proprietary data and prototype products for annotation. With Meta now poised to own nearly half of Scale, rivals fear that their AI roadmaps and technical secrets could inadvertently become visible to one of their fiercest competitors.

One source told Reuters that “it’s no longer just about the quality of the labels — it’s about protecting your intellectual property.” AI firms are increasingly anxious that working with Scale might compromise their competitive edge in the high-stakes AI arms race.

Rivals Move In for the Kill

The Meta deal is creating opportunities for Scale AI’s competitors to snap up panicked customers. Jonathan Siddharth, CEO of Turing, a direct rival, described the situation as a “turning point” in the AI infrastructure space. “Neutrality is no longer optional — it’s essential,” Siddharth said.

Labelbox, another major player in the data-labeling space, is forecasting hundreds of millions in new revenue by the end of 2025, fueled by clients defecting from Scale. Similarly, Handshake, which focuses on assembling networks of PhDs and domain experts for complex labeling tasks, has seen its workload triple in the past week.

And it’s not just established vendors benefiting from the upheaval. Brendan Foody, CEO of Mercor, a data-labeling startup that uses automation to recruit and vet human labelers, says many AI companies are now rushing to build in-house labeling teams. “The top labs want full control over their data security,” he added.

A High-Stakes AI Battle Intensifies

Founded in 2016, Scale AI became a central infrastructure provider for the AI boom, offering massive volumes of curated, human-labeled data crucial for training state-of-the-art models like ChatGPT. But this week’s developments illustrate how fragile even the most successful AI service providers can be in an ecosystem where every player is also a potential competitor.

For Meta, the acquisition of a nearly 50% stake in Scale AI is a bold bet to catch up in the AI race. After the release of its Llama 4 models earlier this year received a tepid response from the AI community, Meta is clearly ramping up its efforts to improve its large language model capabilities.