OpenAI has confirmed it will continue collaborating with Scale AI despite Meta's massive $14.8 billion investment that values the data labeling startup at $29 billion.
Speaking at the VivaTech conference in Paris on June 13, OpenAI CFO Sarah Friar emphasized the importance of maintaining an open AI ecosystem even as ownership changes reshape the competitive landscape. "We don't want to ice the ecosystem because acquisitions are going to happen," Friar stated. "And if we ice each other out, I think we're actually going to slow the pace of innovation."
Founded in 2016 by Alexandr Wang, Scale AI has become a critical infrastructure provider in the AI industry, specializing in producing and labeling high-quality training data essential for developing sophisticated AI models. The company's services are used by major tech firms including OpenAI, Google, Microsoft, and Meta itself.
The deal, which gives Meta a 49% ownership stake, also includes Wang joining Meta to lead its new "superintelligence" unit focused on artificial general intelligence. This arrangement creates an unusual dynamic where direct competitors OpenAI and Meta will rely on the same data labeling service, even as their respective AI models—ChatGPT and Llama—compete for market dominance.
For Meta, the investment represents a strategic effort to catch up with competitors in the AI race. According to industry sources, CEO Mark Zuckerberg has grown frustrated that rivals like OpenAI appear to be ahead in both underlying AI models and consumer-facing applications. Meta's recent Llama 4 release was considered disappointing, and the company has delayed its more powerful "Behemoth" model due to performance concerns.
OpenAI, backed by Microsoft, maintains that it works with multiple data vendors beyond Scale AI. However, the company recognizes Scale's importance in providing the labeled and curated training data crucial for developing its AI systems. As models become more sophisticated, the demand for high-quality data has increased, with OpenAI now engaging specialists including historians and scientists with doctoral qualifications to train models more effectively.