The largest customer of Scale AI, Google of Alphabet, is planning to cut the relationship with the scale after the breakdown of the news that the rival Meta AI is taking 49% stake in data-labeling startups, with five sources familiar to the matter reported to Reuters.
Google planned to pay around $ 200 million this year for human-labeled training data, which is important for developing technology, which includes sophisticated AI models, which said Power Gemini, its flat competitive, one of the sources.
The search giant already interacted with several AI rivals this week, as it tries to overcome that charge, the sources said.
Scale’s significant business loss comes when the meta takes a large stake in the company, giving it a price of $ 29 billion. The scale was $ 14 billion before the deal.
Scale AI intends to keep its business running, while its CEO, Alexandra Wang, along with some employees, proceed to the meta. Since its main occupation focuses around some customers, if it loses major customers like Google, it can suffer greatly.
In a statement, a spokesman for A Scale AI said that its business, which works with major companies and governments, remains strong, as it is committed to the protection of customer data. The company refused to comment on the nuances with Google.
Sources said Scale AI recorded revenue in 2024 for $ 870 million, and Google spent $ 150 million on AI’s services last year.
Other major technical companies who are scal customers, including Microsoft, are also retreating. Elon Musk’s Zai is also looking to get out, a source said. According to sources familiar with the case, Openai decided to pull back from the scale several months ago, although it spends much less money than Google. Openai’s CFO said on Friday that the company will continue to work with Scale AI as one of its many data vendors.
Five sources said that companies competing with Meta in developing state -of -the -art AI models are concerned that doing business with a scale can expose their research preferences and road maps to an opponent. By contracting with Scale AI, customers often share ownership data as well as prototype products, for which the scales of the scale are providing data-lingering services. Meta is now taking 49% stake with Mata, AI companies are concerned that one of their main rivals can gain knowledge about its business strategy and technical blueprint.
Google, Microsoft and Openai refused to comment. XAI did not respond to the remarks request.
Rivids see the opening of the opening
The larger part of the revenue of Scale AI comes from charging generic AI model manufacturers to provide access to a network of human trainers with special knowledge – from historians to scientists, with some doctoral degrees. Humans anote complex datasets that are used for the “post-train” AI model, and as the AI models have become smarter, the demand for sophisticated human-derived examples has increased, and the cost of an anotation can be $ 100.
The scale also does data-labeling for self-driving car companies and enterprises such as US government, which is likely to live according to sources. Sources said that its biggest money-maker is partnership with generic AI model manufacturers.
Google had already demanded its data service providers to diversify for more than a year, saying three sources. Sources said that this week, Meta’s steps have inspired Google to move Scale AI on all its major contracts. The way the data-labeling contracts are structured, that process can be quickly done, two sources said.
This scale will provide an inauguration to the AI rivals to jump.
“Meta-Skele Deal is a turn,” Jonathan Siddharth said, CEO of Turing, a scale AI competitor. “Leading AI Labs feel that neutrality is no longer alternate, it is necessary.”
Labbox, another competitor, “possibly hundreds of crores of new revenue”, which by the end of the year to the end of the year, its CEO, Manu Sharma told Reuters.
Handsheakes, a competitor focusing on the creation of a network of PhDs and experts, saw a bounce of work from top AI laboratories that compete with meta.
Handshek CEO Garat Lord Lord Lord said, “Our demand has been three times overnight after the news.”
Many AI Labs now want to hire in-house data-laberes, which allows their data to be safe, Mercer CEO Brendon Foody said, a startup that besides competing directly with a scale AI also makes technology around the AI, AI is able to make technology, AI Labs is also able to handle it.
Established in 2016, Scale AI provides large amounts of label data or curated training data, which is important to develop refined tools such as openiI chats.
The Meta Deal Scale AI will have a boon for Asec and Index Ventures, as well as its current and former employees.
As part of the deal, Scale AI’s CEO, Wang, will take a top position for AI efforts of Meta.
Meta is fighting the assumption that the Lama released in April after the initial set of 4 big language models may be behind in the AI race, decreasing with the expectations of performance.
© Thomson Reuters 2025
(This story is not edited by NDTV employees and auto-generated from a syndicated feed.)