Unlock Editor’s Digest Lock for Free
FT editor Roula Khalaf will select your favorite stories in this weekly newsletter.
Meta has invested $15 billion in data-learning startup scale AI and hired co-founder Alexandr Wang as part of a bid to attract talent from rivals in a fiercely competitive market.
The transaction value is $29 billion, twice as much as last year. Without elaborating, Scale said it would “significantly expand” its commercial relationship with Meta “to accelerate the deployment of scale data solutions.” Scale helps businesses improve their artificial intelligence models by providing labeled training data.
Size distributes revenue from Meta’s investments to shareholders, Meta owns 49% of the capital on the post-trade scale.
Wang, who was also the CEO of Scale, will “work on Meta’s AI efforts” in his new role. However, Meta did not reveal his new role.
Wang, 28, will remain on the large board of directors that nominated Jason Droege as Chief Strategy Officer. Droege launched Uber Eats as interim CEO in September and helped him join Scale.
“Meta’s investments recognize the achievements of scale so far and reaffirm that our progress, like AI, is endless,” Wang said. “We are delighted that Jason will lead the next step in his Scale journey.”
Meta’s investment is the latest attempt by CEO Mark Zuckerberg to gain an edge in the race to develop stronger AI models for a 1.8-ton social media company. Zuckerberg is trying to poach top researchers and engineers from rival groups as they are trying to build a new “Superintelligence” team.
The deal is one of the biggest of its kind as tech companies increasingly agree to invest in startups while attracting top staff. Last year, Microsoft paid $650 million to hire Mustafa Suleyman and his top Li, and licensed technology for the startup. Google paid $2.7 billion for a similar arrangement to Character.ai.
Meta has invested heavily in generator AI, investing most of the $720 billion of capital expenditure allocated to data centers and servers this year. The agreement highlights AI companies willing to pay for data that can be used to train AI models.
Zuckerberg pledged last year that his company’s model would surpass rival efforts in 2025, but Meta’s latest release, Llama 4, reduced performance with a variety of independent reasoning and coding benchmarks.
Meta’s researchers’ long-term goals “we’ve always reached and surpassed human intelligence,” said Yann Lecun, AI scientist at the Vivedch Conference this week in Paris.
Building artificial “general” intelligence – AI technology with human-level intelligence – is a popular goal for many AI companies. In addition, the number of Silicon Valley groups is increasingly trying to reach “super intelligence,” a virtual scenario in which AI systems outweigh human intelligence.
Recommended
The core of the business of scale is data labels, a manual process that guarantees images and text, and is labeled and classified accurately before being used to train AI models.
Wang has built Silicon Valley’s biggest investors and engineers (relationships with engineers including Sam Altman from Openai). Scale AI’s early customers were autonomous vehicle companies, but the majority of the expected $2 billion in revenue this year comes from labelling the data used to train large-scale AI models built by Openai and others.
The deal will provide significant paydays for early venture capital investors on scale, including Accel, Tiger Global Management and Index Ventures. According to those with knowledge of the issue, Tiger’s $200 million investment is worth more than $1 billion in the company’s new valuation.
Additional Reports by Tabby Kinder in San Francisco
