Matt O’Brien
Meta has invested $14.3 billion in the size of an artificial intelligence company, and is joining the team that will develop “Superintelligence” with Tech Giant, employing CEO Alexandr Wang.
The deal, announced Thursday, reflects a push by Meta CEO Mark Zuckerberg to revive AI efforts at Facebook and Instagram’s parent company as it faces tough competition from rivals such as Google and Openai.
Meta announced what was called “Strategic Partnerships and Investments” on a scale late Thursday. According to Scale, the $14.3 billion investment will have a market value of over $29 billion.
Scale said it will remain an independent company, but the agreement “will significantly expand the commercial relationship between scale and meta.” Meta holds a 49% stake in the startup.
Wang will set out for a small group of employees of other sizes and meta, but will remain on the board of sizes. He will be replaced by Jason Droege, the company’s chief strategy officer and CEO of the new interim scale who previously took on past executive roles at Uber Eats and Axon.
Zuckerberg’s focus on the abstract idea of ”super intelligence” (what rivals call artificial general information, or AGI) is the latest pivot for technology leaders who made the idea of metaverse all over the board in 2021.
It’s not the first time ChatGpt’s 2022 debut has sparked an AI Arms race. Microsoft has hired key staff at the startup’s Refracted AI, including Mustafa Suleyman, co-founder and CEO who currently runs Microsoft’s AI division.
Google has drawn in the leaders of AI Chatbot Company Characher.ai, and Amazon has signed a deal with San Francisco-based Adept, sending its CEO and key employees to the e-commerce giant. Amazon has also licensed Adept’s AI systems and datasets.
Wang was a 19-year-old student at the Massachusetts Institute of Technology when he and co-founder Lucy Guo began the scale in 2016.
They won influential support from the startup incubator Y combinator that summer. This was led at the time by Sam Altman, now CEO of Openai. Wang dropped out of MIT after following a similar trajectory to Zuckerberg, who left Harvard University to start Facebook more than a decade ago.
The pitch of the scale provided the human labor needed to improve AI systems, employing workers to draw boxes around pedestrians and dogs in street photographs, allowing self-driving cars to better predict what is right in front of them. General Motors and Toyota are among the customers of Scale.
The scale provided to AI developers was a more customized version of Amazon’s mechanical Türkiye. It was a go-to service to match freelance workers with temporary online jobs.
Recently, the growth in the commercialization of AI large-scale language models – the technology behind Openai’s ChatGpt, Google’s Gemini and Meta’s Llama have brought new markets for scale annotation teams. The company is serving “all major large-scale language models” including humanity, Openai, Meta and Microsoft, and insists on fine-tuning training data to test performance. It is not clear what meta transactions mean for other customers on scale.
The King also tried to develop close ties with the US government, win a military contract to supply AI tools to the Department of Defense, and attend President Donald Trump’s inauguration. Michael Kratzos, director of Trump’s Science and Technology Office, was a major executive for four years between Trump’s Article 1 and Second Terms. Meta has also started offering AI services to the federal government.
Meta takes a different approach from many of its rivals, releasing the flagship Llama system for free and free as an open source product that allows you to use and modify some of its key components. Meta says that over 1 billion people use AI products each month, but it is widely seen as lagging behind competitors such as Openai and Google in encouraging consumer use of large-scale language models, also known as LLMS.
Despite previewing it in April as “one of the smartest LLMs in the world and one of our most powerful LLMs,” it has not released its most advanced model, the Llama 4 Behemoth.
Meta AI scientist Yann Lecun, who won the top award in computer science for his pioneering AI work in 2019, has expressed his skepticism about his focus on the technology industry’s current large-scale language model.
“How do you build an AI system that understands the physical world and has permanent memory, which can reason and plan?” asked Lecun at the French Technical Conference last year.
These are all traits of intelligent behaviors where large-scale linguistic models can “basically not, or can only do them in a very superficial and approximate way,” Lecun said.
Instead, he emphasized Meta’s interest in “tracing the path to human-level AI systems, or perhaps even superhuman.” When he returned to France’s annual Vivatec conference again on Wednesday, Lekun dodged questions about a pending size deal, but said his AI research team’s plans “always reached and exceeded human intelligence.”
“Now, we have a clearer vision of how to achieve this,” he said.
Lecun co-founded Meta’s AI research division with Rob Fergus, a fellow professor at New York University over a decade ago. Fergus later left for Google, but after five years of absence last month, he ran the lab and replaced longtime director Joel Pineau.
Last month, Fergus wrote on LinkedIn that meta’s commitment to long-term AI research is “unwavering,” describing the work as “building human-level experiences that change the way we interact with technology.”
Original issue: June 13th, 2025, 5:29pm EDT
