Mark Zuckerberg is betting more than $14bn that a well-connected 28-year-old is the catalyst his company needs to catch up with rivals racing to develop and commercialise artificial intelligence.
The Meta chief struck a deal last week to buy 49 per cent of data labelling business Scale AI and hire its leader Alexandr Wang. Zuckerberg paid $14.3bn, marginally ahead of Scale’s last valuation. But ownership of the start-up is “basically incidental”, according to one Scale investor. The real prize was Wang.
An executive at one of Meta’s biggest rivals said the move came after Zuckerberg decided Wang was the “wartime CEO” he needed at the centre of the Big Tech group’s “superintelligence” lab — at a time when Meta was falling behind its AI rivals.
A maths prodigy and the child of two Chinese physicists, Wang, who co-founded Scale aged 19, is neither a celebrated AI researcher nor a top engineer.
According to a dozen people who have worked with and for Wang, or invested in Scale, he is a highly effective operator who has close connections across Silicon Valley.
“There are very few companies that have deep relationships with every top research team: there’s Scale and Nvidia, and [Nvidia boss] Jensen [Huang] probably wouldn’t take the job,” said Dan Levine, a partner at venture capital firm Accel who was one of the first investors in Scale and remains on the company’s board.
Zuckerberg is hoping that Wang’s star power will draw away the best talent from rivals such as OpenAI, Anthropic, Google and Safe Superintelligence (SSI) — the start-up headed by ex-OpenAI chief scientist Ilya Sutskever.
Meta is in talks to hire Dan Gross, one of SSI’s co-founders, and his investing partner Nat Friedman, according to a person with knowledge of the matter. The discussions, which were first reported by The Information, are at an advanced stage and poaching the pair would represent a coup for Meta’s revamped AI division.
Gross and Friedman did not respond to a request for comment.
Wang’s real gift, according to multiple people who have worked with him, is an ability to cultivate powerful allies both within Silicon Valley and beyond. His social calendar has interspersed meetings with world leaders including Keir Starmer, Emmanuel Macron and Narendra Modi, alongside trips to New York’s Met Gala and Formula 1 races.
His confidantes include OpenAI chief and one time housemate Sam Altman, the ChatGPT maker’s former chief technology officer Mira Murati, and Michael Kratsios, President Donald Trump’s technology adviser.
Wang has long idolised Zuckerberg, according to someone familiar with the matter, adding that Scale had sought venture funding from Accel in part because it was an early Facebook investor.
For staff at Scale — who have joked internally that the company would take years to go public because of its high valuation — the deal represents a welcome windfall. Staff in Meta’s generative AI team are less impressed, according to a person close to the company, citing doubts over Wang’s technical prowess.
Meta is far from a neutral broker in the AI race, and Wang’s ability to pull in favours might be shortlived. His new post puts him in a war for talent against the likes of Altman and Murati, who founded her own AI start-up called Thinking Machines last year.
Altman said this week that Meta had attempted to poach OpenAI developers with the promise of $100mn sign-on bonuses and even higher pay.
The massive outlay on new employees is an admission that Meta’s existing AI strategy was stuttering.
The company’s latest release, Llama 4, underwhelmed on a number of independent performance benchmarks. Wang’s title, and his relation to Yann LeCun — the computer scientist who leads Meta’s fundamental AI research efforts — has not yet been determined, according to the company.
“This is the great bet: the most important thing short term is the research team they assemble quickly,” said someone close to Scale. “They’re making incredible offers and are trying to hire a talented team as soon as possible. Alex knows all of these people . . . It’s almost like a heist.”
Others who know Wang see opportunism behind the relentless network building. “Wang is an operator, he wants to make as much money as possible and make as big a name for himself as possible,” said one former employee.
“He wants his name on everything, even when he isn’t across it or hasn’t done the work,” said another person who has worked closely with Wang on a number of projects.
Scale contested that characterisation, describing Wang as hard working and conscientious.
At Scale, Wang proved adept at reading the market and political winds. The start-up initially focused on serving companies training autonomous vehicles, before pivoting in 2019 to focus more on artificial intelligence. It quickly became the most prominent of a group of companies doing the unglamorous work of labelling and categorising data used to train AI tools such as OpenAI’s ChatGPT.
That has involved hiring more than 100,000 “gig” workers, many in the developing world, via a subsidiary called Remotasks. It has more recently used subject experts to train and test models in highly specialised domains.
Scale said most of the company’s “contributors” are now based in the US.
Scale has pursued new business with the US government and helped companies build AI applications. That business is growing fast, according to a person close to Scale, but the bulk of its $870mn in revenue last year came from labelling data for the biggest generative AI companies, including Meta.
“Alex is never happy with his team, never happy with himself. When I met him as a 19-year-old I noticed his intelligence, but missed his relentless drive,” said one early investor in Scale.
Wang has also become a trusted messenger in Washington DC thanks to his adroit policy stances, such as extreme hawkishness on China and an advocacy for “merit-based” hiring over diversity.
Scale’s fate is now unclear. New chief executive Jason Droege said the Meta deal, which valued Scale at $29bn, validated the company’s approach and was “not a pivot or a winding down.”
“We have an exceptional team, a clear vision, and the resources to achieve truly remarkable things,” he said.
In the near-term issue Scale is fighting to retain its biggest customers, and reassure them that their data will be ringfenced from Meta. OpenAI and Google have already signalled they will wean off Scale’s services.
“Scale’s business is going to collapse, no one wants to work with a company Meta owns 49 per cent of, so [Meta] had to pay what investors valued the entire company at in the last round,” said one of the start-up’s backers.
Scale said the company’s relationships with key customers remains strong, adding that Meta’s investment indicates there is real value in the remaining business. OpenAI said the company had been in the process of moving off Scale for the last few months. Google declined to comment.
Someone close to the company was more bullish about Scale’s future without Wang. “Our customers can’t really leave us,” the person said. “We’re in a war and [Scale is] the arms dealer.”
Additional reporting by Stephen Morris in San Francisco