There are all sorts of ways large corporate entities attempt to weaponise capital, quite apart from throwing a ton of money at the founders of unproven start-ups. While WeWork became a cautionary tale, some of SoftBank’s other investments, including TikTok’s owner, ByteDance or British chip design firm ARM Holdings, made it a fair bit of money.
In January, Softbank announced it was teaming up with ChatGPT creator OpenAI Inc to invest up to US$500 billion in a joint venture dubbed Stargate, which will build massive AI data centres across America. So far, though, only one data centre in Ohio is under construction.
Where there is a promise of outsized returns, there is someone ready to pour an infinite amount of capital to corner those boatloads of profits. Take Mark Zuckerberg, the founder and CEO of social media colossus Meta Platforms Inc, which owns Facebook, Instagram and WhatsApp. Over the past two months, Zuckerberg has been literally splashing around billions of dollars to lure top AI talent, mostly software engineers, to build superintelligence, or the form of AI that can match or even exceed human intelligence. Indeed, he has gone so far as to say that Meta is willing to spend whatever it takes to get there.
Meta’s focus is on what it calls “personal superintelligence for everyone” to give people tools for their own empowerment and to better the world. Unlike Amazon, Microsoft and Google, which rent out their servers to customers looking to outsource cloud infrastructure needs, Meta is using its data centres and expensive AI chips for its own use. Meta expects to spend US$72 billion this year and nearly US$100 billion next year. Imagine what sort of return Meta needs on US$100 billion of capital expenditure (capex) on AI infrastructure.
See also: Microsoft says Chinese hackers are exploiting SharePoint flaws
Moreover, the advertising industry’s total global social media spending between now and 2030 is likely to grow 9.3% per year or by a total of US$142 billion. Meta’s revenues are projected to grow from US$295 billion this year to US$362 billion by the turn of the decade. Clearly, Meta’s Math looks challenging.
Spending spree on talent
Meta began its current superintelligence spending spree through an acquihire in June. It plonked down US$14.8 billion for a 49% stake in Scale AI Inc, which uses gig workers to manually label data, because it wanted to hire its 28-year-old founder, Alexandr Wang. Acquihire is a term used to describe the purchase of a company principally to lure its top talent.
See also: Nvidia hits US$4 trillion value as rally notches another milestone
Meta’s swoop on Scale AI, clearly designed to bring Wang into its fold, is the biggest acquihire in history. In 1996, Apple Inc paid US$400 million to purchase NeXT Computer. The move was the acquihire of its founder, Steve Jobs, who had been fired by its board 11 years earlier. Jobs’ return enabled Apple, which was weeks away from bankruptcy, to get its act together and eventually become a tech giant worth US$3.5 trillion.
Unlike an outright acquisition or a transaction that might give Meta a controlling stake, the Scale AI deal was deftly structured so that it does not require a review by antitrust regulators. The US Federal Trade Commission under the Biden administration probed Amazon’s deal to hire top executives and researchers from AI start-up Adept and Microsoft’s US$650 million deal with Inflection AI Inc to lure its top talents, London-born ethnic Syrian Mustafa Suleyman and British-Armenian Karen Simonyan, but took no action.
Aside from the acquihire of Wang, Meta has hired Daniel Gross, CEO of Safe Superintelligence and Nat Friedman, who once ran developer platform GitHub. Gross and Friedman co-founded VC fund NFDG and other top AI talent to help drive its new Meta Superintelligence Labs, a key priority for Zuckerberg, who has been focused on pushing Meta to pole position in the race to dominate AI.
Until recently, Meta was seen as the number four player in the race behind OpenAI and its partner, software supremo Microsoft, search giant Google and billionaire Elon Musk’s xAI, which created Grok chatbot. Its efforts to build foundational models reportedly fell way short of its lofty expectations. Its latest foundational model, Llama 4, missed key benchmarks.
According to media reports, internal tests showed Llama 4 underperformed rivals like OpenAI’s GPT-4 and Anthropic’s Claude 3. Indeed, some staffers went as far as accusing top Meta executives of inflating Llama 4 scores. Morale plunged as infighting and finger-pointing led many Meta engineers to jump ship and join other AI start-ups.
That eventually forced Zuckerberg to intervene and relaunch Meta’s AI initiative. The founder took direct control of recruitment to rebuild its AI lab. He reportedly got hold of a list of top AI talents in America and began personally messaging and cold-calling individuals on the list.
Potential recruits from Apple, OpenAI, Google’s DeepMind, Anthropic, Microsoft, Tesla and other firms initially get a WhatsApp message from Zuckerberg. Those who respond are then personally called by the world’s third richest person (net worth US$271 billion) for a long chat, followed by an invite to his home in Palo Alto or his holiday home in Lake Tahoe, where they are offered compensation packages of over US$100 million.
Sink your teeth into in-depth insights from our contributors, and dive into financial and economic trends
The New York Times recently reported that Matt Deitke, a 24-year-old AI engineer, was initially offered a cash-and-stock options package of US$125 million over four years. He turned the offer down. Zuckerberg then upped the ante with a US$250 million offer with US$100 million paid in the first year.
Dietke wasn’t the one who got the biggest offer. Zuckerberg tried to lure Mira Murati away from her AI start-up, Thinking Machines Lab, with a US$1 billion deal. Murati, the former chief technology officer of OpenAI, launched her start-up earlier this year. By last month, she had already raised over US$2 billion at a US$2 billion valuation when Meta came calling. She turned it down.
Zuckerberg then reportedly approached almost all of Murati’s key staffers with outrageous offers, with at least one reportedly as high as US$1 billion. Wired magazine reported that many other offers were between US$200 million and US$500 million over four years.
How do compensation packages for 20-something AI engineers compare with what the world’s top athletes or movie stars make? Football superstar Cristiano Ronaldo brings in US$225 million a year, basketball icon Steph Curry makes US$150 million and tennis champion Novak Djokovic earned US$44.6 million last year. Hollywood idol Tom Cruise made around US$45 million last year. Of course, younger Hollywood stars earn far less. Sydney Sweeney, 27, currently the hottest actress in Hollywood and the star of Anyone But You, reportedly made about US$28 million last year.
As it tries to reshape the AI landscape in a fierce talent war, Meta has reportedly lured nearly 40 top AI engineers, including six from Google, five from Apple and about 14 from OpenAI. Though it has been using its chequebook aggressively, Meta isn’t the only firm poaching AI talents from competitors. Google, xAI, OpenAI and Microsoft have been luring AI engineers from each other in recent months. Indeed, media leaks have cited a Microsoft “list” reportedly part of a grand plan to poach Meta’s own top AI engineers.
Is Meta concerned that it is throwing too much money at AI talent? Or is the quantum of money irrelevant? “We do take very seriously that this is just a massive amount of capital,” Zuckerberg told analysts recently.
Superintelligence is a powerful tool
During the firm’s earnings call on July 30, Zuckerberg said Meta would continue paying top dollar for AI talent “because we have conviction that superintelligence is going to improve every aspect of what we do”.
Here’s another way to look at the huge gobs of money Meta and others are throwing at AI talent. Capex at the hyperscalers — Amazon, Microsoft and Google, plus Meta, Oracle Corp and Musk’s xAI — will top US$400 billion this year, and probably US$480 billion next year. Meta alone will spend up to US$72 billion in capex this year and US$100 billion next year. If you are spending US$100 billion on buying AI chips or building data centres, it doesn’t matter if you set aside an extra US$5 billion to hire the best global talent. Why spend US$100 billion on hardware if your rival has the best talent? Moreover, since most of the money is being paid in stock options rather than cash, and those options vest over several years, companies like Meta can let go of the employees at any time they realise the talent is not the right fit for the firm.
While Meta has lagged behind OpenAI, Anthropic and Google in building cutting-edge models, it is now ready to use its open-source strategy to undercut OpenAI’s proprietary models. Zuckerberg believes that Meta can commoditise the technology by releasing open-source models that directly compete with ChatGPT maker OpenAI or Claude’s maker Anthropic.
Not everyone believes offering hundreds of millions of dollars to a handful of very talented 20-year-olds will work out. For one thing, despite hiring the best people for top dollar, there is no guarantee that Meta will emerge as the AI leader. Moreover, the huge pay disparity between its new top star hires and 76,000 other employees could turn into a “cultural challenge”.
While stars like to work with other talented co-workers, they are also fiercely competitive and turf-conscious. Michael Dell, the founder of PC maker Dell Inc, in a recent podcast, noted that the generous salaries of the new AI hires could lead to unhappiness and resentment among existing employees. That in turn could potentially create a long line of “complaining” employees outside Zuckerberg’s office, demanding fair treatment, Dell noted. While he conceded that Meta’s strategy and the “math” might work if the focus is only on the “race to superintelligence”, Dell argued that there was potential for internal conflict and dissatisfaction, which could work against Meta.
In an advertising business like Meta, the key to success is accurate targeting of the ads. The Facebook founder’s vision is to use AI to help generate cheaper, better and more-targeted ads for the three billion users of Facebook, two billion users each of Instagram and WhatsApp. Zuckerberg has hinted at building an array of wearables to help drive Meta’s revenue growth.
Those “personal devices like glasses that understand our context because they can see what we see, hear what we hear, and interact with us throughout the day will become our primary computing devices,” he noted recently. Anyone can make devices. Selling them globally at scale and making money is a different story. Meta sold two million glasses over the past two years, while Apple sold 275 million iPhones in the same period.
Whether consumer-focused Meta will be able to convert its huge AI spending into bigger profits than business-focused players like Microsoft or Amazon remains to be seen.
Assif Shameen is a technology and business writer based in North America
