Five sources familiar with the matter told Reuters that Scale AI’s largest user, Scale AI’s largest user, plans to reduce relationships with such a scale that rival Meta AI is taking 49 % of the shares in the data labeling startup, according to five sources.
One of the sources said that Google plans to pay about $ 200 million to Scale AI for human -labeled training data this year, which is very important for developing technology, which includes its Chat GPT’s rival, Gemini, the latest AI models in Power Gemini.
Sources added that Search Dev had already talked to several AI rivals this week as she is trying to eliminate the burden of the work.
The loss of the main business of the scale has come to the fore when the Meta Company is at a big stake, costing $ 29 billion. Prior to the deal, the scale was worth $ 14 billion.
Scale AI intends to continue its business, while its CEO, Alexander Wang, goes to the meta with some employees. Since its main business is focused around some consumers, it may hurt a lot if it loses key users like Google.
In a statement, an AI spokesperson said his business, which works with major companies and governments, is strong, as it is committed to protecting customer data. The company refuses to comment on the details with Google.
Sources said that Scale AI increased revenue of 70 870 million in 2024, and Google spent about $ 150 million last year on Scale AI’s services.
Other major tech companies, including Microsoft, are also backing it. A source said that Elon also wanted to get out. According to sources familiar with the matter, Openi decided to withdraw from the scale several months ago, though it spends less than Google. Openai’s CFO said on Friday that the company will continue to work with Scale AI, as it is one of many of its data dealers.
Five sources said that companies that compete in the development of the latest AI models with Meta are concerned that doing business with a scale can expose their research priorities and road maps to the competitors. By contracting with Scale AI, consumers often share proprietary data as well as prototype products for which scale workers are providing data labeling services. Meta has now taken 49 percent of the shares, AI companies are concerned that one of her major rivals can find information about her business strategy and technical blueprint.
Google, Microsoft and Openi refused to comment. Zee did not respond to a comment.
Rivals saw the opening
Most of the revenue of Scale AI is obtained by charging generating AI model makers to provide access to a special knowledge network to the network of human trainers – from historians to scientists, with some doctoral degrees. Humans interpret complex datases that are used for the “after the train” for the AI model, and as the AI models have become more smart, the demand for sophisticated human provided examples has increased, and an interpretation can cost up to $ 100.
According to sources, the scale also labels data for self -run car companies and businesses like the US government, which is likely to remain according to sources. Sources said that its biggest money maker is in partnership with the Generative AI model makers.
Google has already tried to diversify its data service providers for more than a year, three sources said. Sources added that this week Meta’s tricks have tried to remove Google on all of its important contracts on a scale. Two sources said that the way the data labeled contracts have been developed, this can be done fast.
This scale will provide an opening task to jump to AI rivals.
“Meta -scale contracts indicate a turning point,” said Jonathan Siddhartha, CEO of Tourist, Scale AI. “The well -known AI labs realizes that neutrality is no longer optional, it is important.”
His CEO, Manu Sharma, told Reuters, the label box, another competitor, “will probably generate hundreds of new revenue” by the end of the year, his CEO, Manu Sharma, told Reuters.
Handshake, a contestant focusing on a PhD and a network of experts, saw an increase in work burden from high AI labs competing with Meta.
“After the news, our demand has increased threefold overnight,” said Handshek CEO Garrett Lord.
Many AI labs now want to hire the data labors, which allow their data to be safe, said Merkel’s CEO Brendon Foody.
Established in 2016, Scale AI provides labeled data or curatic training data, which is very important to produce sophisticated tools like Openi’s chat GPT.
Meta Dale Scale and Index Ventures will have an honor for Scale AI investors as well as its current and former employees.
As part of the agreement, Scale will take a high position under the AI’s CEO, Wang, Meta’s AI efforts.
Meta is fighting the impression that Lama, released in April, has fallen behind in the AI race after the opening set of the opening set of the initial set of 4 major language models.
© Thomson Writers 2025
(This story has not been edited by the NDTV staff and has been made auto from the Syndicate Fed.)
