Google is pulling the plug on its biggest AI partnership. The tech giant plans to cut ties with Scale AI after learning that rival Meta is taking a 49% stake in the AI data company. Five sources familiar with the matter confirmed the breakup to Reuters.
This isn’t pocket change we’re talking about. Google had planned to pay Scale AI roughly $200 million this year for human-labeled training data. That’s the stuff that powers Gemini, Google’s answer to ChatGPT. Last year alone, Google dropped $150 million on Scale’s services.
The math is simple. Meta gets a piece of Scale AI, and suddenly Google’s biggest AI competitor has a window into Google’s research. Not happening.
Alphabet’s Google already started shopping around this week. The company held talks with several Scale competitors as it scrambles to replace that $200 million worth of work. Sources say Google wants out of all its key contracts with Scale.
Google isn’t alone in the exodus. Microsoft is backing away too. Elon Musk’s xAI is looking for the exit door. Even OpenAI pulled back months ago, though they spent far less than Google anyway.
The concern is obvious. When you hire Scale AI, you’re sharing proprietary data and prototype products. Workers at Scale see your technical blueprints. Your research priorities. Your roadmap. Now Meta owns nearly half the company.
“Companies that compete with Meta are concerned that doing business with Scale could expose their research priorities and road map to a rival,” sources explained. Makes sense.
Scale AI built its empire on human intelligence. The company charges AI makers for access to a network of human trainers. Historians. Scientists. Some with doctorate degrees. These people annotate complex datasets that “post-train” AI models. As AI gets smarter, demand for this sophisticated human input has exploded. A single annotation can cost $100.
The company pulled in $870 million in revenue last year. But that business model just hit a wall.
Scale’s CEO Alexandr Wang is moving over to Meta as part of the deal. He’ll take a top position leading Meta’s AI efforts. A few employees are following him.
Meta valued Scale at $29 billion in this deal. That’s more than double its previous $14 billion valuation. Nice payday for Scale’s investors at Accel and Index Ventures.
Scale says it’s staying in business despite losing its CEO and biggest customers. A company spokesperson insisted the business “remains strong” and that Scale is “committed to protecting customer data.” They wouldn’t comment specifically about Google.
The company still has other revenue streams. Self-driving car companies. The U.S. government. But generative AI model makers were the biggest money-makers. Losing Google, Microsoft, and others? That’s going to hurt.
Scale’s competitors are already circling. Jonathan Siddharth, CEO of rival Turing, sees an opportunity. “The Meta-Scale deal marks a turning point. Leading AI labs are realizing neutrality is no longer optional, it’s essential.”
Labelbox CEO Manu Sharma told Reuters his company will “probably generate hundreds of millions of new revenue” by year-end from customers fleeing Scale.
Garrett Lord, CEO at competitor Handshake, reported immediate results. “Our demand has tripled overnight after the news.”
Some AI companies are ditching the middleman entirely. Brendan Foody, CEO of Mercor, says many labs now want to hire in-house data-labelers. Keeps the data secure. No risk of leaks to competitors.
Founded in 2016, Scale AI became essential infrastructure for the AI boom. The company provides the labeled data that trains sophisticated AI tools like ChatGPT. But the industry runs on paranoia about competitive advantages. Meta’s investment just triggered every alarm bell.
Meta needs this deal. The company is fighting perceptions that it’s falling behind in the AI race. Its initial Llama 4 language models released in April disappointed. Performance fell short of expectations.
Getting Wang and a stake in Scale gives Meta a boost. But it comes at the cost of Scale’s biggest relationships. In AI, trust matters more than technology. Once that’s broken, the money follows.
Google’s already talking to Scale’s rivals. The data-labeling business is about to get very competitive.
