- Google and others drop Scale AI after Meta takes 49% stake.
- Companies shift to other vendors amid data security fears.
Google is planning to pull out of its partnership with Scale AI after learning that Meta is buying a large stake in the company, according to people familiar with the matter.
Google had been preparing to spend about $200 million this year on Scale’s data-labeling services, work used to train AI systems like Gemini, Google’s answer to ChatGPT. But after Meta agreed to take a 49% stake in Scale, Google started looking for other vendors, the sources said. The company has already spoken with several of Scale’s competitors this week.
Meta’s deal values Scale AI at $29 billion – more than double the $14 billion valuation it had before. The investment is a major shift for Scale, whose business depends heavily on a few large tech customers. With CEO Alexandr Wang and several team members moving to Meta, Scale’s future as an independent data supplier is now in question.
In a statement, Scale said it works with both governments and major companies and is committed to protecting customer data, and declined to discuss Google specifically.
Scale made $870 million in revenue last year, with Google’s reportedly worth $150 million. Other technology companies are rethinking their partnerships with Scale, too. Microsoft is stepping back, sources said. xAI, Elon Musk’s AI company, is also planning to leave. OpenAI started to move away months ago, though it spends much less than Google’s outlay. On Friday, OpenAI’s CFO said the company still works with Scale as one of many vendors.
Many AI companies are concerned that working with Scale could put their research at risk. Customers often send proprietary data and early versions of products to Scale’s contractors for labelling. With Meta now owning a large share, some fear the data could be used to Meta’s advantage.
Google, Microsoft, X, and OpenAI have declined to comment.
Scale’s main business is providing high-quality training data labelled by humans. That includes people with advanced degrees who help fine-tune AI models by labelling tricky datasets. Some annotations can cost up to $100 each. Scale also works with industries like self-driving cars and government agencies, but most of its money comes from working with AI labs, sources said.
Even before the Meta deal, Google had started spreading its data-labeling contracts among different providers. But Meta’s move has pushed Google to accelerate its exit, with two sources saying the switch could happen quickly due to how the contracts are set up.
Other companies in the space are already seeing new opportunities.
“The Meta-Scale deal marks a turning point,” said Jonathan Siddharth, CEO of Turing, a rival data-labeling firm. “Leading AI labs are realising neutrality is no longer optional, it’s essential.”
Labelbox, another competitor, expects to bring in hundreds of millions in new revenue this year, its CEO Manu Sharma told Reuters. Handshake, which works with a network of PhDs and specialists, said it’s already seeing a jump in work. “Our demand has tripled overnight after the news,” said Garrett Lord, the company’s CEO.
Scale AI’s new deal with Meta could mean big returns for investors like Accel and Index Ventures, along with past and current employees. As part of the agreement, CEO Alexandr Wang will take on a major role leading Meta’s AI team. Meta, meanwhile, is trying to catch up in the AI race after its latest set of Llama 4 models – released in April – didn’t perform as well as expected.
