Meta is currently struggling in the AI race, but had CEO Mark Zuckerberg been more bullish on AI a decade ago, it could’ve been where Google is today.
In early 2014, DeepMind was the most sought-after AI lab in the world. Google and Facebook were both circling it, chequebooks open. Facebook reportedly offered more money. It still lost — and according to a new book, the reason came down to a single dinner in Palo Alto.
The story is told in Sebastian Mallaby’s The Infinity Machine: Demis Hassabis, DeepMind and the Quest for Superintelligence. As acquisition talks intensified, Hassabis flew to the West Coast to have lunch with Larry Page, Google’s strongest suitor. Zuckerberg got wind of the visit and invited him to dinner.
What followed wasn’t a pitch — it was a test. Arriving at Zuckerberg’s home, Hassabis steered the conversation toward AI and registered how Zuckerberg responded. The Facebook CEO expressed genuine enthusiasm. So far, so good. But then Hassabis pivoted — bringing up virtual reality, augmented reality, 3-D printing. Zuckerberg lit up about all of them equally.
That was the tell. “Facebook offered more money, but I wanted somebody who really understood why AI would be bigger than all these other things,” Hassabis said later. Zuckerberg’s undifferentiated excitement signalled to Hassabis that AI was, for him, just another emerging technology on a list — not the defining bet of the century.
Google’s acquisition of DeepMind, reportedly for around $500 million, now looks like one of the shrewdest moves in tech history. The lab’s culture of fundamental research — AlphaGo, AlphaFold, the transformer architecture itself — merged with Google’s engineering infrastructure to create what is now Google DeepMind, a dominant force in frontier AI. Hassabis was awarded the Nobel Prize in Chemistry in 2024 for his work on AlphaFold.
Meanwhile, Facebook — now Meta — spent years treating AI as a feature rather than a foundation. It built recommendation algorithms and content moderation tools, but ceded the frontier research ground. By the time Zuckerberg declared AI an existential priority, Google and OpenAI had a decade-long head start.
The irony is compounded by what happened after Hassabis declined. Spurned, Zuckerberg redoubled his recruitment efforts and lured Yann LeCun — the deep learning pioneer at NYU — to build an internal AI lab. LeCun agreed to join without even leaving his professorship or New York, conditions Zuckerberg accepted without hesitation. The mission: use Facebook’s war chest to poach researchers from smaller labs, starting with DeepMind.
At that December’s machine learning conference, Hassabis literally bumped into LeCun. “You’re not going to poach all my guys, are you?” he asked. LeCun’s honest answer: he had basically just signed on to do exactly that. Within weeks, LeCun had called Koray Kavukcuoglu — a key DeepMind contributor and former student — and offered him a substantial raise to defect. “That was the moment,” Demis’s co-founder Mustafa Suleyman recalled, “I thought DeepMind might really fail.”
Hassabis’s countermove was to accelerate the Google deal. He let Kavukcuoglu in on the secret — DeepMind was on the verge of being acquired, and the stock options everyone had written off might be worth a fortune. Kavukcuoglu stayed. Hassabis then pushed Google to close the acquisition as fast as possible.
The episode reveals something important about how consequential acquisitions are actually decided. It wasn’t valuation or term sheets that settled DeepMind’s future — it was a founder’s read of a prospective acquirer’s conviction. Hassabis didn’t just want capital. He wanted a home that understood, at a fundamental level, what he was building and why it mattered. In Zuckerberg, he saw enthusiasm but not depth. In Google, he saw a company that got it.
For Meta, the cost of that dinner has only grown. The company has spent the years since playing catch-up — launching Llama, offering nine-figure packages to poach researchers, and reportedly even exploring a joint bid on OpenAI with Elon Musk. None of it has closed the gap that opened on a Palo Alto evening in 2013 when Zuckerberg answered a question about AI with equal enthusiasm for 3-D printing.