Meta has appointed former OpenAI researcher Shengjia Zhao as the new Chief Scientist for its Superintelligence Lab, as CEO Mark Zuckerberg announced Friday. The move is the latest and most senior hire in Meta’s aggressive, multi-billion-dollar campaign to build an elite AI team by poaching top talent from rivals.
This strategy, personally driven by Zuckerberg, aims to accelerate Meta’s quest to build artificial superintelligence after internal setbacks delayed its own models. Zhao, who worked on the original ChatGPT, will now help set the scientific direction for Meta’s ambitious goal, reporting to Scale AI founder and foAlexandr Wang, the
In a Threads post, Zuckerberg confirmed the high-level appointment, stating, “in this role, Shengjia will set the research agenda and scientific direction for our new lab working directly with me and Alex.” The hire solidifies the leadership of the new Meta Superintelligence Labs (MSL), a division formed on July 1 to consolidate the company’s AI efforts.
View on Threads
Zhao is a heavyweight in the AI research community, widely recognized as a co-creator of ChatGPT and a key architect of OpenAI’s pioneering reasoning models. During his tenure at the now rival lab from 2022 to 2024, he was instrumental in developing not only the world-changing ChatGPT and its powerful successor, GPT-4, but also the o1 reasoning model. That model significantly advanced the “chain-of-thought” paradigm, a crucial step toward enabling more sophisticated, multi-step AI cognition.
His work also included contributions to GPT-4.1 and leading initiatives in synthetic data generation, a technique now considered critical for training and refining next-generation foundation models.
His appointment is backed by a formidable academic and research pedigree. Zhao holds a PhD in Computer Science from Stanford University, where he studied under the influential AI professor Stefano Ermon, and a bachelor’s degree from the prestigious Tsinghua University in Beijing.
His work has been recognized with numerous accolades, including an ICLR 2022 Outstanding Paper Award, a JP Morgan PhD Fellowship, and a Qualcomm Innovation Fellowship. With over 14,000 academic citations and an H-index of 23 for his influential papers on topics like variational autoencoders and AI calibration, Zhao brings a deep well of scientific credibility to Meta’s commercial ambitions.
For Meta, Zhao is more than just a star hire; he is a strategic necessity. His expertise in advanced AI reasoning directly addresses the capability gaps that reportedly contributed to the underperformance and subsequent delay of the company’s own Llama 4 “Behemoth” model.
He now completes a new leadership triumvirate for the Superintelligence Lab, reporting to Chief AI Officer Alexandr Wang and working directly with Zuckerberg to set the scientific agenda. The move to bring in a researcher of Zhao’s caliber is the clearest signal yet of the urgency driving Meta’s new direction—a strategy forged directly in the crucible of its recent internal setbacks.
A Strategy Forged in Crisis
Meta’s aggressive “buy or poach” playbook was forged in the fires of its own internal crises. The company’s pivot came after the significant development failure of its ambitious Llama 4 “Behemoth” model, which was postponed in May 2025 after underperforming on key benchmarks.
With its internal development faltering and having lost most of the original Llama research team, Meta turned its vast financial resources outward. The internal turmoil reportedly fostered a “panic mode” among engineers, creating immense pressure to acquire talent after being spurned in takeover bids for startups like Runway and Safe Superintelligence (SSI).
The Great AI Talent War
This pivot kicked off an escalating talent war across Silicon Valley. Meta has systematically poached top minds from OpenAI, Apple, and Google, including Jason Wei, Hyung Won Chung, and Ruoming Pang. The company also hired three engineers from Google’s Gemini team after their recent success.
The hires were surgical. The poaching of specific researchers from OpenAI, for example, secured expertise in AI reasoning, a known capability gap for Meta. This “mafioso” style poaching has put the entire industry on the defensive.
The campaign has been marked by staggering compensation packages. Reports suggest offers reaching hundreds of millions, with one rejected offer to a top researcher valued at an astonishing $1.25 billion, highlighting the extreme financials of this conflict.
The relentless poaching sparked a full-blown crisis at chief rival OpenAI. In a leaked internal memo, Chief Research Officer Mark Chen expressed a raw sense of corporate violation, telling his staff, “i feel a visceral feeling right now, as if someone has broken into our home and stolen something.”
In a clear defensive move, OpenAI has since ramped up its stock-based compensation to over $4.4 billion to prevent a further staff exodus. The conflict has become a public spectacle, with CEOs trading barbs over the enormous signing bonuses.
The collateral damage from Meta’s strategy has been significant. Its $14.3 billion investment in Scale AI, designed to secure Alexandr Wang, led to a client exodus over neutrality fears. The fallout forced Scale AI to lay off 14% of its workforce in July and face a major data security breach.
Billions in Compute: The New Currency of Power
Underpinning this talent acquisition blitz is a colossal investment in infrastructure. Zuckerberg has publicly committed to spending hundreds of billions on the compute power necessary to attract and empower this new team. In a July 14 post, he declared, “we’re also going to invest hundreds of billions of dollars into compute to build superintelligence. We have the capital from our business to do this.”
This new infrastructure strategy is radical. Reportedly inspired by xAI’s rapid deployments, Meta is embracing “tent” style data centers that use prefabricated modules and forgo traditional backup generators. In Ohio, the company is even building its own on-site natural gas plants to ensure steady power.
This investment includes the construction of multiple gigawatt-scale “titan clusters,” with the first, named “Prometheus,” set to come online in 2026. This focus on raw processing power is central to Meta’s recruiting pitch.
Zuckerberg explained the new calculus of the AI talent war in a recent interview. He noted that while executives once asked about management scope, today, “here, people say, ‘I want the fewest number of people reporting to me and the most GPUs.'” This sentiment was echoed by Perplexity CEO Aravind Srinivas, who recalled a recruitment attempt being shut down with the line, “come back to me when you have 10,000 H100 GPUs.”
This brute-force approach of combining nearly limitless capital with a relentless talent campaign is Meta’s high-stakes answer to its recent struggles. Analyst Gil Luria noted the uniqueness of the Scale AI deal, which was an investment “…not to even buy a whole company but just to have the head of a company head up your AI effort.” For Zuckerberg, the goal is to buy his way to the top.
