Meta Platforms has entered a critical phase in its artificial intelligence roadmap by formalizing the leadership of its most ambitious AI initiative to date: Meta Superintelligence Labs (MSL). The recent appointment of Shengjia Zhao, a former OpenAI researcher instrumental in developing ChatGPT, GPT‑4, and OpenAI’s AI reasoning model o1, marks a pivotal turn in Meta’s competitive push to lead in frontier AI. Zhao assumes the role of Chief Scientist, working under the recently appointed Chief AI Officer Alexandr Wang, previously CEO of Scale AI.
Zhao, a co-founder of MSL, had already been quietly leading research operations within the lab. Now with his leadership formalized, Meta signals that it is no longer just participating in the AI race—it is preparing to define the next lap.
Timeline of Developments Leading to Zhao’s Appointment
The establishment and evolution of MSL follow a methodical build-up over the past few months. Meta began consolidating its AI units in late June 2025, merging its FAIR (Fundamental AI Research) lab, Generative AI teams, and LLaMA foundation model developers under a single umbrella. This move laid the groundwork for a unified superintelligence strategy.
Around the same time, Meta invested heavily in both talent acquisition and infrastructure scaling. The company recruited Alexandr Wang to lead MSL—despite his non-research background—a decision initially seen as unconventional. Wang’s appointment, however, was soon balanced by Zhao’s scientific credibility. By early July, Meta publicly announced the structure of MSL, positioning it as a hub for developing “human-level AI” systems.
Zuckerberg’s official announcement naming Zhao as Chief Scientist came on July 25, although Zhao had already begun forming the lab’s intellectual backbone. Supporting hires quickly followed, including researchers from OpenAI, DeepMind, Anthropic, Apple, and Safe Superintelligence. Many of them had previously worked with Zhao on AI reasoning and multimodality.
Looking ahead, by 2026, MSL will have access to Meta’s 1 gigawatt AI supercomputing cluster Prometheus in Ohio. Prometheus will support massive-scale training runs essential for frontier models, with plans already underway for a 5 gigawatt expansion under a new cluster named Hyperion.
Implications for Meta’s AI Architecture
Meta now operates three main AI arms:
-
FAIR, led by Yann LeCun, continues to focus on long-term, blue-sky research into future techniques and theoretical models.
-
MSL, under Zhao and Wang, has a more aggressive, near-term focus on building general-purpose AI systems that can reason, plan, and interact like humans.
-
Product AI Teams, which integrate advances into Meta’s apps (Instagram, WhatsApp, Threads, and Facebook) through LLaMA-based models and Meta AI tools.
This structural segmentation allows Meta to pursue foundational innovation while also competing directly with current market leaders like OpenAI and Google DeepMind. The result is a bifurcated strategy: one eye on the 10-year horizon, the other on beating GPT-class models in the next 12–18 months.
Zhao’s Role and the AI Talent War
Zhao’s appointment is not just a formality—it’s a strategic message. At OpenAI, he was known for driving several core innovations, particularly in AI reasoning. His work on the o1 model introduced a “new scaling paradigm” for reasoning capabilities, pushing beyond language generation into domains like decision-making, planning, and symbolic logic.
By bringing Zhao on board, Meta is directly targeting a weakness in its portfolio: the absence of a model to rival o1. Zhao is expected to build Meta’s own reasoning model, using techniques refined at OpenAI and adapted for Meta’s infrastructure scale.
To support this, Meta has aggressively recruited top AI minds—sometimes offering eight- and nine-figure packages, reportedly with tight deadlines known as “exploding offers.” This tactic has ignited debate across the AI ecosystem. Critics argue that such high-stakes poaching destabilizes the talent pipeline, while supporters view it as necessary for achieving breakthroughs.
The current composition of MSL reflects Meta’s global ambition. The team includes researchers from OpenAI’s Zurich office, top Chinese universities, and a handful of elite AI institutions. Among the 44 confirmed team members, only two are of Indian origin—a statistic that has sparked commentary given India’s broader contribution to global AI.
Infrastructure at Scale: The Role of Prometheus
Meta’s AI aspirations hinge not just on talent but also on compute capacity. The Prometheus cluster, projected to come online in 2026, will operate at 1 gigawatt—enough to power 750,000 homes. This level of scale is rare outside of state-level AI projects. Once fully operational, Prometheus will give MSL the firepower needed to train trillion-parameter models with real-time iteration and feedback cycles.
Plans for Hyperion, a 5 gigawatt cluster, are already underway, indicating Meta’s intent to own the infrastructure race alongside its model development strategy.
The Road Ahead: Coordination, Competition, and Credibility
The creation of MSL is the culmination of a months-long strategic pivot. From structural unification to elite hiring and infrastructure investment, Meta is betting that it can transform its decentralized AI efforts into a cohesive, superintelligent engine.
Yet challenges remain:
-
How will MSL integrate with FAIR without redundancy?
-
Can Zhao’s reasoning model match or exceed o1 in terms of capability?
-
Will Meta’s AI systems reflect responsible safety practices or fall into the race-for-scale trap?
The answers will define Meta’s place in the AI future. With Zhao now formally at the helm and compute resources scaling rapidly, Meta is no longer chasing the AI leaders—it is building its own path.
