
Meta is escalating its AI push with an aggressive campaign to acquire top-tier talent, build cutting-edge infrastructure, and secure its place in the global race for AI dominance.
In its latest move, the company has hired Trapit Bansal, a prominent researcher from OpenAI credited with helping launch the company’s first AI reasoning model, o1, and contributing to reinforcement learning efforts alongside co-founder Ilya Sutskever.
Bansal’s defection to Meta’s newly formed AI superintelligence team marks a significant win for CEO Mark Zuckerberg, who has made no secret of his ambition to catch up with — and potentially outpace — rivals like OpenAI, Google DeepMind, and emerging players like DeepSeek. Bansal will be joining a growing list of elite scientists Meta has lured away in recent months.
Register for Tekedia Mini-MBA edition 18 (Sep 15 – Dec 6, 2025) today for early bird discounts. Do annual for access to Blucera.com.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment.
Among them are Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai, all former OpenAI researchers, as well as Jack Rae from Google DeepMind and Johan Schalkwyk, who previously led machine learning at Sesame AI. Meta’s willingness to offer compensation packages rumored to reach as high as $100 million reflects just how serious the company is about claiming a leading position in AI.
The team’s goal is to develop cutting-edge AI reasoning models — the kind that can perform complex tasks, analyze and solve problems step by step, and power autonomous AI agents across Meta’s vast ecosystem, from social platforms to enterprise tools.
Meta currently lacks a publicly available reasoning model on par with OpenAI’s o3 or DeepSeek’s R1.
But Meta’s approach isn’t limited to hiring alone. It has also made strategic financial moves, including a 49% stake in Scale AI — a $14.3 billion investment that gave the conglomerate access to one of the world’s largest data-labeling platforms. Scale’s founder, Alexandr Wang, has joined Meta’s AI superintelligence team and is expected to play a central role in its infrastructure buildout.
Additionally, Meta explored acquisition talks with a string of influential AI startups, including Safe Superintelligence, co-founded by Sutskever; Mira Murati’s Thinking Machines Labs; and Perplexity AI, the search startup. While none of those deals materialized, they reveal a clear pattern of Meta targeting research labs and startups at the cutting edge of AI development.
All of this comes as AI reasoning models — which allow systems to “think through” problems before responding — become the next frontier in the race to build smarter, more adaptable systems. With OpenAI’s GPT-4o, Google’s Gemini 1.5, and DeepSeek’s R1 already raising the bar, and Meta has so far lagged in releasing an equivalent model, Bansal’s expertise may prove key to closing that gap.
Sam Altman, OpenAI’s CEO, recently acknowledged that Meta has attempted to poach his company’s talent with the $100 million offer, although he insisted that “none of our best people have decided to take him up on that.” Still, the departures paint a different picture — and Meta’s momentum is hard to ignore.
Zuckerberg sees Meta’s superintelligence lab as more than just an R&D unit. Like Google’s DeepMind, it is expected to power a wide range of AI agents across Meta’s platforms, from WhatsApp and Instagram to business-facing tools being developed under former Salesforce AI chief Clara Shih. The company has already committed over $70 billion in AI capital expenditure through 2025.
With Bansal onboard and its infrastructure expanding rapidly, Meta is positioning itself as a central player in the next phase of AI evolution — one that will be shaped by talents and the ability to retain the brightest minds.