Month after month, message after message, the AI engineer was listening to from Meta recruiters. The recruiters have been pestering him to go away his employer and swap over to help the corporate’s AI efforts, and so they have been providing a large wage bundle to take action. However he wasn’t so certain.
The engineer, who works for a startup that was acquired by a number one AI firm and requested anonymity from The Verge, stated he had heard from pals that the corporate anticipated a number of private sacrifices in alternate for its excessive salaries, whether or not on staff’ worth methods in relation to Al or work-life steadiness. Engineers there, he heard, have been working across the clock to meet up with rival corporations like OpenAl, Anthropic, Google, and Microsoft.
With so many corporations desperately scrambling for AI expertise, Meta was providing between $1–1.4 million in whole annual compensation (which is often measured as a mixture of wage, annual bonus, and amortized inventory worth) for a lot of AI roles. However, he suspected, its affords could be much less beneficiant than they sounded — tied closely to subjective efficiency metrics that might be weaponized towards staff. And simply as importantly, he wasn’t prepared to surrender a semblance of work-life steadiness and a wholesome work surroundings to make a couple of hundred thousand {dollars} extra. He didn’t pursue the chance.
In current months, Meta has launched on an AI hiring spree after making its largest-ever exterior funding: a $14.3 billion acquisition of a 49 p.c stake in Scale AI, an {industry} big that gives coaching information to gasoline the know-how of corporations like OpenAI, Google, Microsoft, and Meta. As a part of the deal, Meta spun up a brand-new superintelligence lab led by Scale AI CEO Alexandr Wang — and to employees the lab, it began poaching.
Meta has reportedly poached as many as 10 of OpenAI’s high researchers and mannequin builders, with some pay packages reportedly including as much as $300 million over 4 years, together with fairness. (Meta disputes this determine.) It’s additionally approached a slew of different high AI expertise throughout the {industry}. Ruoming Pang, who heads up Apple’s basis AI fashions crew, reportedly departed for Meta, and at the least two Anthropic staff and two DeepMind staff have reportedly joined the crew as nicely. The purpose is to safe Meta’s spot within the race to attain synthetic common intelligence, or AGI: a hypothetical AI system that equals or surpasses human cognitive talents, and the shifting goal that just about each AI firm is presently chasing at breakneck pace. Meta’s main weapon is huge quantities of cash. However some sources throughout the AI {industry} query whether or not that will likely be sufficient.
Meta, to this point, has not been essentially the most thrilling vacation spot for budding AI engineers
Meta, to this point, has not been essentially the most thrilling vacation spot for budding AI engineers. CEO Mark Zuckerberg has been making an attempt to make up for misplaced floor within the AI race, having spent years and important sources over-indexing on the metaverse whereas rivals like Google, Microsoft, and Amazon invested billions in AI startups and signed cloud contracts and different offers. The corporate’s Llama AI fashions typically rank low on publicly maintained efficiency leaderboards; at time of writing, Meta’s first look on one such leaderboard, Chatbot Arena, was at No. 30. In Could, it reportedly delayed the launch of its new flagship AI mannequin as builders struggled to ship efficiency upgrades, and executives have been public on earnings calls about the necessity to aggressively make investments and shore up towards competitors. The Scale AI deal, and Meta’s subsequent sky-high price range for hiring AI expertise, is Zuckerberg’s Hail Mary: paying a premium for a number of the brightest minds within the AI world to safeguard Meta’s future.
However though Meta is plying AI employees with staggering salaries, a mountain of cash can’t purchase everybody. Anthropic and DeepMind have reportedly had far fewer defections to Meta than OpenAI has, and that’s been an ongoing development. The rationale, to these inside the sector, is clear: the AI world is full of true believers, and even the most important corporations want greater than a money supply to get a lot of them on their facet.
Business insiders emphasised to The Verge that in a sector the place nearly any firm will supply job safety and wage, skilled AI engineers and researchers wish to work someplace that aligns with their values, whether or not their high precedence is AI security and the dangers the tech poses for humanity’s future, the moral issues of AI’s impression on society at present, or accelerating and advancing the tech faster than anybody else. Some engineers, researchers, or scientists the corporate has approached have turned down Meta’s advances, {industry} sources inform The Verge.
Competitors for AI researchers is stiff, and constructing loyalty is important. “At this level, at the least a couple of hundred high researchers and engineers within the subject are what’s generally referred to as ‘post-money’ — they might retire, and also you’re solely going to draw or retain them in the event that they consider in your imaginative and prescient, management type, and so on.,” one AI {industry} supply says.
However particularly at OpenAI, Meta appears to have discovered the greenback worth of firm loyalty — and exceeded it. OpenAI has been uniquely affected by Meta’s mission to poach main AI expertise. As many as 10 of its high researchers and mannequin builders have reportedly joined Meta, with some receiving giant signing bonuses and fairness. Whereas its measurement and expertise make it an inevitable prime goal, the corporate can also be susceptible attributable to a controversial restructuring from a nonprofit to for-profit enterprise and the departures of high-profile executives who went on to begin competing AI corporations. It underwent an enormous upheaval throughout Sam Altman’s November 2023 ouster by OpenAI’s board and his subsequent Uno Reverse-style rehiring, which noticed many of the board members who opposed him resign. Workers have additionally shared concerns about non-disparagement agreements and insurance policies that raised questions on whether or not they would be capable to entry their fairness, eroding some belief in management even when the insurance policies have been walked again.
“A whole lot of the individuals engaged on this are genuinely satisfied they’re constructing transformative know-how that can reshape the world,” one supply aware of the state of affairs tells The Verge. “For the those who have been that mission-driven, there’s already been a lot organizational turbulence” — he talked about Altman’s firing and rehiring, OpenAI staff defecting to Anthropic, and governance adjustments — “that persons are much less anchored to the establishment itself, so it’s simpler to poach [from OpenAI] than the opposite labs.”
In lieu of an official remark, OpenAI directed The Verge to a blog post from its world affairs crew, which states, “Some eye-popping affords are being prolonged as of late to a handful of terrifically proficient researchers, together with to of us at OpenAI. A few of these affords are coming with deadlines of just some hours – actually ‘exploding affords’ – or with restrictions on whether or not or how they are often mentioned.” The weblog submit goes on to say the corporate plans to domesticate expertise throughout not simply analysis, but in addition product, engineering, infrastructure, scaling, and security. On Wednesday, the corporate announced it had employed 4 engineers away from corporations like Tesla, xAI, and Meta.
Now, OpenAI’s greatest protection towards the losses is its personal monetary leverage. Those that joined OpenAI early on, and even earlier than the tip of 2023, have had important inventory appreciation — the unit value jumped from $67 in a Could 2023 tender supply to $210 on the finish of 2024, in response to a supply aware of the state of affairs. And through the finish of 2023, across the time of the OpenAI board curler coaster, there was a window wherein OpenAI rushed to make hires from different corporations who would signal on on the $67-per-unit determine, since there was a near-immediate 2.5x multiplier anticipated, the supply says.
With so many corporations competing to rent AI expertise, there have been high-level departures at many various corporations. However even earlier than this most up-to-date hiring frenzy, OpenAI’s staff have been being lured elsewhere at a higher-than-average charge.
A 2025 SignalFire report analyzed retention patterns in AI and located that Anthropic was greatest at holding individuals round, with 80 p.c of staff employed at the least two years in the past at Anthropic remaining on the firm on the conclusion of their second yr. DeepMind got here subsequent at 78 p.c, whereas OpenAI’s retention charge was markedly decrease, at 67 p.c — corresponding to Meta’s 64 p.c. The report, which got here out in Could earlier than Meta’s Scale deal occurred, additionally discovered that engineers have been “8 instances extra more likely to depart OpenAI for Anthropic than the reverse,” and 11 instances extra more likely to defect from DeepMind to Anthropic than the reverse.
Anthropic was based by ex-OpenAI analysis executives with the purpose of fastidiously growing AI know-how, describing itself as an “AI security and analysis firm that’s working to construct dependable, interpretable, and steerable AI methods.” The corporate’s “mission-driven security focus” is a convincing recruitment pitch and a key cause for its low turnover, one supply aware of the state of affairs tells The Verge.
“The priorities of the businesses turn into very totally different, and it’s not one thing that they’re pleased with”
“I’ve seen, amongst people who find themselves slightly bit extra seasoned in {industry} … they’ve watched tech as an {industry} change, and so they’ve watched the individuals within the {industry} turn into very totally different, and the priorities of the businesses turn into very totally different, and it’s not one thing that they’re pleased with,” says Rumman Chowdhury, a longtime chief within the subject of accountable AI at corporations like Accenture and Twitter, who now heads up the AI nonprofit Humane Intelligence. When she’s hiring AI engineers, she says, they typically say that wage hikes are much less vital to them than to “not be contributing to a worse world.”
AI engineers and researchers can afford that idealism, and issues about security and the frenzy to commercialize the know-how have dogged most main AI corporations. At OpenAI, as an illustration, months of controversy and public stress about its upcoming transition right into a for-profit entity led to the corporate altering its plans, ceding some management to its nonprofit arm even after the restructuring. The choice adopted a public letter written by ex-employees and civic leaders to the California and Delaware attorneys common, with one former worker writing, “OpenAI could at some point construct know-how that might get us all killed.”
There are questions, too, in regards to the tempo at which Meta is shifting and its analysis priorities. Meta’s Scale AI funding got here on the heels of Joelle Pineau’s departure as Meta VP and head of its Elementary AI Analysis (FAIR) division, a unit Meta folded into its bigger AI efforts after beforehand describing it as “one of many solely teams on the planet with all of the conditions for delivering true breakthroughs with a number of the brightest minds within the {industry}.” Some noticed FAIR’s restructuring as an indication that Meta was prioritizing merchandise over analysis, an industry-wide concern for some AI security consultants.
For Meta, nonetheless, there are additionally pragmatic questions on its future in AI. In conversations with The Verge, {industry} insiders questioned whether or not Wang is the fitting selection to go the brand new lab, since Scale AI doesn’t construct frontier fashions and Wang himself doesn’t have an AI analysis background. And even Meta executives admit that catching up will likely be a problem. Throughout its most up-to-date earnings name in April, Zuckerberg — who referred to as one in all Meta’s focuses for 2025 “making Meta AI the main private AI” — touched on the competitors he was up towards. “The tempo of progress throughout the {industry} and the alternatives forward for us are staggering. I wish to be sure that we’re working aggressively and effectively, and I additionally wish to be sure that we’re constructing out the main infrastructure and groups we have to obtain our objectives,” he stated.
Simply two months later, that crew grew to become Meta’s superintelligence lab. However the AI engineer who steered clear doesn’t remorse his determination.
“You’re anticipated to present just about your entire self to Meta AI,” he says. The cash merely wasn’t adequate for that.