Meta plans an investment of around 15 billion dollars in the data labeling startup Scale AI and also wants to take over key figures of the company. With a 49 percent stake in Scale AI and a valuation of approximately 28 billion dollars, the deal would be one of the largest in the still young AI industry.
The goal is to build a so-called "superintelligence" unit that can compete with the leading models of OpenAI, Google, and Anthropic. Meta is responding to the weak performance of its current model, Llama 4, which fell significantly short of expectations in benchmarks for logical problem-solving and programming performance.
Co-founder Alexandr Wang, who founded Scale AI in 2016, is set to play a key role in Meta's new lab. Wang, now a billionaire on paper, is considered well-connected in Silicon Valley, including with OpenAI CEO Sam Altman. The operational leadership of Scale AI is to be transferred to the previous Chief Strategy Officer Jason Droege.
Scale AI specializes in the manual labeling of training data for AI models and supplies customers from autonomous mobility and increasingly from the government sector. Recently, the company began developing customized AI solutions for businesses to reduce strong dependence on a few large customers.
The deal would be further evidence of the strategic shift in the industry: Big Tech is increasingly investing not only in its own development teams but also in purchasing targeted start-ups or integrating key personnel. Microsoft paid 650 million dollars last year for the Inflection AI team, Google invested 2.7 billion dollars in Character AI—each under structural conditions meant to evade regulatory scrutiny.
Meta is under pressure because, in addition to major competitors, open-source models like those of the Chinese provider DeepSeek are entering the market with high quality and lower costs. The planned entry into Scale AI is primarily intended to improve the quality of training data, a leverage often underestimated so far in the development of powerful AI models.
