Free tools Windows power users keep installed
One-click scans. No signup required.
Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Meta recruited several prominent OpenAI researchers in 2025 as it built Meta Superintelligence Labs (MSL), part of a broader effort to compete in frontier AI. Sam Altman said Meta made offers involving roughly $100 million signing bonuses; WIRED later reported packages worth as much as $300 million over four years. Those figures describe reported offers, not verified cash payments to every recruit. The more important question is whether Meta can turn expensive hires into a durable research organization and better products.
What happened in Meta’s 2025 hiring push
In June and July 2025, Mark Zuckerberg accelerated Meta’s recruitment of researchers and engineers from OpenAI and other AI companies. The effort formed part of a larger reorganization around Meta Superintelligence Labs, which brought together new recruits and existing Meta AI work. Meta was not starting from scratch: its FAIR research group had a long history in AI. The goal was to combine that base with people experienced in building and scaling modern frontier models.
The campaign extended well beyond OpenAI. Reporting described hires or recruiting efforts involving Google, Anthropic, Apple, and other organizations. So “Meta poached OpenAI researchers” captures one conspicuous part of the story, not the full composition of MSL. WIRED’s account of the team details its varied backgrounds.
Meta also made a roughly $14.3 billion investment in Scale AI, whose CEO Alexandr Wang became a central leader in Meta’s AI reorganization. That investment had strategic and commercial dimensions; it should not be reduced to a simple purchase of one executive’s services. Wang’s arrival and the new lab nevertheless illustrated how aggressively Meta was consolidating leadership, talent, and resources around AI.
#1 Best Overall
Who moved from OpenAI?
Public reporting confirms several prominent OpenAI-to-Meta moves, but it does not justify saying that most of OpenAI’s best researchers left. Here are notable examples supported by the supplied reporting:
| Person or group | Prior affiliation | What reporting says |
|---|---|---|
| Shengjia Zhao | OpenAI | Reported as joining Meta and taking a significant research or leadership role in the new organization. Coverage has described him as an OpenAI co-founder or founding researcher; those labels should not be taken to mean he alone created any particular model. |
| Yang Song | OpenAI | Reported by WIRED as joining Meta as a research principal. |
| Four researchers in the initial wave | OpenAI | WIRED reported four OpenAI researchers leaving for Meta in June 2025. The departures show a real recruiting success, but not a mass transfer of OpenAI’s research staff. |
| Alexandr Wang | Scale AI | Moved into a central Meta AI leadership role after Meta’s investment in Scale AI; he was not an OpenAI defector. |
Sources: WIRED on the initial OpenAI departures; WIRED on Yang Song; WIRED on MSL’s broader roster.
Being approached is not the same as accepting a job, and an employment move does not disclose a person’s pay, exact responsibilities, or contribution to any past model. Nor should every recruit to MSL be labeled an OpenAI defector: the group drew from multiple firms and backgrounds.
Recommended Free Tools
What the headline compensation figures do—and do not—mean
In June 2025, Altman said Meta was making offers that included signing bonuses of about $100 million, with annual compensation on top. TechCrunch reported his comments. WIRED separately reported that packages for some leading candidates could reach about $300 million over four years, with more than $100 million in first-year compensation. Those figures were reported package values, not publicly disclosed employment contracts.
A package’s headline value is not necessarily guaranteed cash. It may combine salary, a signing bonus, stock or other equity, and performance incentives. Equity can depend on a company’s share price and vesting schedule; bonuses or other components may be conditional. A four-year maximum also is not the same as money paid on day one. Meta disputed at least one specific compensation report, calling it inaccurate and ridiculous. Contracts are private, so the exact value and terms for individual hires are difficult to audit.
Rank #2
The largest reported numbers appear to concern a very small number of elite candidates, not a standard rate for AI employees. The figures still signal how valuable major labs believe certain combinations of research, engineering, and leadership experience can be.
Why Meta would pay so much
Frontier AI requires more than a clever idea. A lab needs people who can scale training, build reliable infrastructure, develop reasoning and multimodal systems, set evaluations, and coordinate research with product teams. Experience at a leading lab can include tacit knowledge—how teams organize experiments, diagnose failures, and move from a promising result to a working system—that is difficult to acquire by hiring a large group of inexperienced staff.
Recruiting experienced leaders can therefore shorten the time needed to build a competitive lab. It can also help attract further hires, signal executive commitment, and make a research effort credible inside and outside the company. But expertise is complementary: a researcher’s impact depends on compute, data, engineering support, sound management, and colleagues who can work together.
Meta brings formidable advantages of its own: the capacity to fund large compute projects and distribution through Facebook, Instagram, WhatsApp, and Meta AI. Its challenge was not merely to hire smart people, but to connect longstanding research strengths with the newer model-building and product capabilities that had made OpenAI and Google prominent competitors. Zuckerberg’s direct involvement and MSL’s “superintelligence” ambition made the shift unusually high-profile. WIRED’s reporting on the organization and The Atlantic’s analysis offer context on the strategy.
OpenAI’s response: retention, morale, and rivalry
Altman criticized Meta’s recruiting effort publicly. OpenAI research chief Mark Chen reportedly sent an internal memo promising a strong response on talent, while Altman invoked a contrast between “missionaries” and “mercenaries”—researchers motivated by a shared mission versus those attracted by large paydays. These were arguments from OpenAI’s leadership during a competitive recruiting campaign, not objective tests of why individual people changed jobs.
OpenAI’s response reflected more than concern about filling vacant roles. Departures can interrupt projects, take institutional memory with them, and unsettle colleagues who wonder whether their company can retain key people. Recruiting pressure can also raise compensation expectations across the industry. At the same time, a person leaving does not prove a company is failing, and turnover can give an organization a reason to clarify priorities or reshape teams. WIRED covered the leadership response; its reporting on Altman’s remarks supplies further context.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Did Meta win the recruiting fight?
On the narrow question of whether Meta recruited prominent OpenAI people, yes: multiple reported moves made the campaign consequential. Meta also assembled a high-profile group quickly and recruited across rival companies. But that is not the same as winning the broader AI race. Some candidates reportedly declined offers, public compensation terms remain opaque, and recruiting a star does not automatically produce a cohesive team or a breakthrough model.
Early signs of instability matter, too. WIRED reported in August 2025 that at least three researchers had left MSL within roughly two months of its launch. That does not establish why they left or prove that the lab failed, but it complicates the idea that extraordinary offers create lasting loyalty. WIRED’s report on the departures is a reminder to judge retention as well as recruitment.
MSL also faced an integration problem. Combining established Meta groups with incoming teams can bring useful range, but it can also produce overlapping mandates, unclear reporting lines, competing research cultures, and friction over who controls compute or sets technical direction. Paying for reputations is no substitute for making those arrangements work. Conversely, a highly paid hire may contribute through team-building, infrastructure, or research direction rather than a single public breakthrough.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The talent war is broader than Meta versus OpenAI
Google DeepMind, Anthropic, xAI, Microsoft, Apple, and new research startups all compete for scarce expertise. Some researchers move between established labs; others start companies. Former OpenAI leaders’ ventures, including Safe Superintelligence and Thinking Machines Lab, illustrate how talent can leave not just for a rival employer but to build a new institution. Startup equity can offer its own uncertain upside: unlike a reported corporate package, its eventual value depends on whether a young company succeeds and on the terms of its ownership.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteThis market rewards a narrow set of people with unusual leverage: researchers who have led major programs, scaled training runs, developed influential systems, or built teams; research engineers who can make large experiments run; infrastructure specialists who keep training and inference systems working; and product leaders who can translate capabilities into useful services. “Top researcher” is not a single job category, and public fame alone is a poor measure of someone’s current contribution.
That concentration creates a sharp labor-market divide. A handful of specialists may attract compensation packages measured in tens or hundreds of millions of dollars, while most AI practitioners do not have comparable bargaining power. The spectacle of star bidding can also encourage startups to form around researchers who want more autonomy, a different mission, or a share in future equity upside. Axios has reported on continued AI talent churn.
How to judge whether the hiring paid off
Hiring totals and compensation headlines are inputs, not outcomes. A better scorecard asks whether MSL can retain the people it recruited, coordinate its teams, and turn research into useful, competitive systems. Relevant evidence includes:
- Models and evaluations: What does Meta release, and how do independent evaluations assess capability, reliability, and limitations? A launch announcement is not an independent benchmark.
- Practical performance: How do quality, latency, and cost compare for real uses, rather than only on a headline score?
- Research pace: Does the organization produce sustained technical progress and move from research to deployment?
- Team durability: Do recruits stay and build groups, or does turnover continue? Are responsibilities and research priorities clear?
- Product adoption: Does Meta integrate its work into Meta AI and its platforms in ways that people actually use?
- Access and openness: What model weights, tools, or technical details does Meta make available? Openness can encourage developer adoption, but should be assessed from what the company actually releases and the conditions attached.
Meta announced Muse Spark in April 2026 as the first model in a new series built by MSL. That is a concrete post-recruitment milestone and evidence that the organization produced a launch. It does not by itself prove that the hires caused the result, that the model leads competitors, or that the organization has solved its retention and coordination challenges. See Meta’s Muse Spark announcement for the company’s description of the model.
The durable lesson of the 2025 campaign is that frontier-AI talent had become strategically valuable enough for Meta to use extraordinary offers, executive attention, and organizational restructuring to pursue it. The campaign demonstrated Meta’s willingness to spend and its ability to attract notable researchers. Whether it built a lasting competitive edge depends on what those people can do together—and on the models, products, and sustained research that follow.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
