WSJ: DeepMind’s founder thinks Zuckerberg doesn’t understand AI—so he gave up an expensive acquisition and instead joined the Google camp

The Wall Street Journal recently disclosed a secret from 13 years ago. In 2013, when artificial intelligence had not yet become the main theme of the capital markets, Meta and Google competed to acquire the AI company DeepMind. The founder of DeepMind judged that Zuckerberg did not understand AI and only regarded it as a technology similar to virtual reality. Ultimately, Google acquired the AI startup DeepMind for approximately $650 million.

Google’s acquisition of DeepMind began with a gathering in 2013.

The report indicates that the key figure in this transaction was DeepMind founder Demis Hassabis. Long before AI ignited market enthusiasm, Hassabis had set his sights on building artificial general intelligence (AGI) and recruited top scientists to develop intelligent agent systems capable of operating across multiple Atari games in a manner reminiscent of the “Manhattan Project.”

A crucial turning point occurred in the summer of 2013. The report describes how Google founder Larry Page directly suggested to Hassabis at a private gathering that if the ultimate goal was to achieve AGI, rather than spending decades building an independent company, it would be better to leverage the vast resources and infrastructure Google had already established. This advice struck at the long-standing financial and computational bottlenecks faced by DeepMind and prompted Hassabis to seriously consider selling the company.

Subsequently, the DeepMind team discreetly traveled to Google’s headquarters to begin negotiations. Notably, both sides deliberately downplayed price discussions early on, focusing instead on research resources and AI safety governance. Co-founder Mustafa Suleyman proposed that if an acquisition were to happen, an independent oversight mechanism composed of scientists and external experts should be established to restrict the use of the technology and even exclude military applications, ensuring that the development of AGI would not be misused.

DeepMind’s founder believed that Zuckerberg did not understand AI, and thus decided to forgo a higher-priced acquisition.

At the same time, another tech giant, Meta (formerly Facebook), founded by Mark Zuckerberg, also actively entered the bidding. The report notes that Facebook not only offered more favorable financial terms but also designed a high signing bonus mechanism to attract the founding team. However, after meeting with Zuckerberg, Hassabis concluded that his understanding of AI was insufficient, viewing it as one of several technologies alongside virtual reality and 3D printing, and ultimately chose to abandon the higher bid.

Facebook responded by recruiting talent through headhunting and successfully brought in Turing Award winner Yann LeCun to build an AI team, which put pressure on DeepMind. At a critical moment, Hassabis revealed to core team members that the company was about to be acquired by Google, in order to stabilize the risk of losing key personnel, and prompted Google to accelerate the deal.

In the end, Google completed the acquisition in early 2014 for about $650 million. The report states that while this price may not seem high by today’s standards, the true value lay in subsequent investments. Over the following decade, Google continued to pour billions of dollars into DeepMind, driving it to achieve key breakthroughs in AI, making this transaction a milestone that reshaped the global AI competition landscape.

This article, WSJ: DeepMind founder felt Zuckerberg did not understand AI! and thus abandoned a high-priced acquisition, turning to Google first appeared in Chain News ABMedia.