TL;DR
- Hardware Push: Meta Superintelligence Labs is building a dedicated hardware team, hiring Rui Xu from acqui-hired startup Dreamer to lead the effort.
- Device Vision: Zuckerberg envisions AI-powered smart glasses as the primary computing device for delivering personal superintelligence to users.
- Rapid Reorganization: Meta cut 600 FAIR researchers and restructured its entire AI division around MSL under Alexandr Wang over the past year.
- Leadership Fallout: Yann LeCun, Meta’s longtime chief AI scientist, departed in March 2026 after clashing with Wang and has since launched AMI Labs.
- Financial Scale: Meta spent $72 billion on capital expenditure in fiscal 2025, mostly on AI infrastructure, with 2026 spending expected to exceed that figure.
Meta Superintelligence Labs is assembling a dedicated hardware team, tapping Rui Xu from the recently acqui-hired AI startup Dreamer to lead an effort that extends the lab’s ambitions beyond models and into physical devices.
The hardware hire signals MSL’s ambition to move beyond AI models into physical devices, following a year of aggressive reorganizations. Meta cut 600 legacy AI researchers, absorbed metaverse leadership, and spent more than $14 billion acquiring Scale AI talent, all to build a vertically integrated superintelligence operation. With Yann LeCun’s recent departure after clashing with MSL head Alexandr Wang, the lab is now consolidating control over both the software and hardware dimensions of Meta’s AI future.
MSL’s Device Ambitions Take Shape
Xu previously worked on smart devices at ByteDance, leading a lab that shipped millions of units in China, and brings management experience from Xiaomi, Lenovo, and Tencent. Before joining Dreamer, he served as COO of K-Scale, a robotics startup that shut down last year. Nat Friedman, who leads products and applied research at MSL, had invested in K-Scale through AI Grant, the program he co-founded, creating a prior connection between the two.
MSL’s hardware push fits squarely into Meta’s stated goal of delivering personal superintelligence to its users. On a February podcast, Wang described a future where AI agents would be “always on, see what you see, hear what you hear” across multiple form factors throughout the day. That device-first framing has direct implications for the hardware team Xu now leads.
CEO Mark Zuckerberg laid out the company’s hardware ambitions in a July 2025 vision statement that positioned wearable devices at the center of Meta’s AI strategy.
“Personal devices like glasses that understand our context because they can see what we see, hear what we hear, and interact with us throughout the day will become our primary computing devices.”
Mark Zuckerberg, CEO of Meta (via Meta)
Zuckerberg’s framing suggests Meta views smart glasses, not phones or headsets, as the primary delivery mechanism for its AI models. Meta already sells its Ray-Ban smart glasses and has been iterating on the form factor for several years. Some Reality Labs engineers have already moved to MSL to prototype AI software on Reality Labs hardware, with the two divisions working closely together.
Meanwhile, Meta signed a multi-billion dollar deal in February 2026 to rent AI chips from Google for model development, underscoring the scale of the company’s AI infrastructure buildout.
Building on this, Xu’s background in shipping consumer hardware at volume positions him to bridge the gap between MSL’s AI research and the manufacturing realities of wearable devices. For Meta, the challenge is not just building models but embedding them into hardware that consumers will wear daily, a task requiring supply chain expertise alongside AI engineering talent.
OpenAI has charted a five-device hardware roadmap including earbuds slated for September 2026, and other tech giants are also racing to build an AI-native personal device beyond the smartphone. Meta’s existing footprint in wearable hardware gives it a potential head start, though translating a smart glasses line into a full AI device ecosystem presents engineering challenges that remain largely unsolved across the industry.
How Meta Rebuilt Its AI Division Around MSL
Xu’s hardware team is the latest addition to an organization reshaped through a series of rapid restructurings. In October 2025, Meta laid off 600 in an AI shake-up at its Fundamental AI Research (FAIR) unit, while simultaneously ramping up hiring for what was then called TBD Lab, its newly formed superintelligence team tasked with achieving artificial general intelligence. FAIR leader Joelle Pineau had departed the organization as the unit lost influence throughout 2024.
By August 2025, Wang stated that FAIR’s research would be integrated and scaled into TBD Lab’s larger model development efforts. He justified the October cuts in an internal memo obtained by Axios.
“By reducing the size of our team, fewer conversations will be required to make a decision, and each person will be more load-bearing and have more scope and impact.”
Alexandr Wang, Head of Meta Superintelligence Labs (via Axios)
From there, the reorganization accelerated. Shah took over AI products at MSL after four years running Reality Labs, reporting to Nat Friedman. Shah now oversees AI integrations across both Meta’s Family of Apps and Reality Labs divisions, while Gabriel Aul replaced Shah as head of the Metaverse Product Group.
Furthermore, Meta brought in Daniel Gross and Nat Friedman, both of whom now hold senior roles in the lab. Jason Rubin, Samantha Ryan, and Thamara Sekhar now report to Aul. Ryan Cairns continues to lead Horizon OS, elevated to an org-level product group reporting directly to CTO Andrew Bosworth.
According to internal memos obtained by Business Insider, Bosworth maintained that the metaverse remained a company priority despite the leadership reshuffle. However, the resource shift toward MSL told a different story, with Reality Labs losing its senior product leader to the AI lab and engineers transferred to prototype MSL’s software on Reality Labs hardware.
The lab’s rapid build-out of talent extended well beyond internal reshuffling. Wang arrived at Meta through the $14B Scale AI deal in June 2025. By July, OpenAI veteran Shengjia Zhao had been named chief scientist for the new lab.
Moreover, by August the lab had been split into four groups and was aggressively recruiting top AI researchers from rivals including OpenAI and Thinking Machines Lab. In October, the 600 layoffs fueled internal conflict over the lab’s direction.
In late 2025, Meta acquired agentic AI startup Manus, and in March 2026 acqui-hired the Dreamer AI team. In less than a year, Meta systematically dismantled its legacy AI infrastructure to rebuild around a single lab under Wang’s leadership.
That consolidation came with visible costs. Yann LeCun, Meta’s longtime chief AI scientist and a foundational figure in deep learning research, quit Meta to launch his own AI venture after clashing with Wang.
LeCun described Wang as “young and inexperienced” and accused Zuckerberg of pushing aside the former AI team after the disappointing Llama 4 AI model release. LeCun has since launched AMI Labs, and his departure illustrates how the rapid reorganization has alienated veteran researchers who built the foundations Wang’s team now builds upon.
What Comes Next for MSL
Building a dedicated hardware team positions MSL as a vertically integrated AI operation, controlling everything from model development to the devices that deliver those models to users. MSL’s stated vision distinguishes it from competitors focused on centralized AI automation: Zuckerberg has framed personal superintelligence as something that should augment individual users rather than replace human labor wholesale.
In contrast, that approach mirrors the strategy Apple has long used with its hardware-software ecosystem, though MSL is attempting it at a far earlier stage of product maturity, before any of its devices have established a mass-market presence.
Wang has promised “incredible velocity” from MSL in the coming months. Whether the lab can deliver on Zuckerberg’s vision for personal superintelligence while managing the internal tensions left by LeCun’s departure and the gutting of FAIR remains an open question.
Meta’s earnings filings show the company spent $72 billion on capital expenditure during fiscal 2025, the vast majority directed at AI data centers and related compute infrastructure, with analysts anticipating 2026 spending could exceed that figure.
With that financial runway and an expanding hardware team, MSL now has the resources and organizational mandate to pursue Zuckerberg’s vision at scale. Whether the lab’s aggressive consolidation of talent and hardware capabilities can outpace competitors building their own AI device ecosystems will determine whether Meta’s bet on personal superintelligence pays off.