In a bold move to secure a dominant position in the global AI race, Meta officially launched “Meta Compute” on January 12, 2026. Described by CEO Mark Zuckerberg as a “top-level initiative,” Meta Compute is tasked with designing, building, and financing the massive physical infrastructure required to achieve artificial superintelligence.
The initiative maps out a decade-long plan to move beyond traditional data centers and build power capacity on a scale never before seen in the private sector.
The Vision: Tens to Hundreds of Gigawatts
Meta Compute isn’t just about adding more servers; it is about securing the energy and hardware foundation for the next generation of AI models, including the upcoming Llama 5 and beyond.
- Massive Power Targets: Meta plans to build tens of gigawatts of capacity this decade, with a long-term goal of reaching hundreds of gigawatts.
- Strategic Advantage: Zuckerberg emphasized that how the company “engineers, invests, and partners” on this infrastructure will become a primary competitive differentiator against rivals like Google and Microsoft.
- Capital Investment: The launch follows a staggering $72 billion capital expenditure in 2025, with much of that funding directed toward AI-ready power and cooling systems.
Leadership: The “Superintelligence” Dream Team
Meta Compute will be led by a trio of high-profile executives, combining technical engineering, strategic planning, and global diplomacy.
| Executive | Role in Meta Compute | Background |
| Santosh Janardhan | Technical Architecture & Silicon | Meta’s Head of Global Infrastructure; former Google exec. |
| Daniel Gross | Long-term Strategy & Partnerships | Co-founder of Safe Superintelligence (SSI); former Apple AI lead. |
| Dina Powell McCormick | Sovereign & Govt Partnerships | Former Trump National Security Advisor; Goldman Sachs exec. |
Dina Powell McCormick, newly appointed as President and Vice Chair, will specifically lead efforts to partner with sovereign wealth funds and governments to finance these multi-billion dollar gigawatt-scale projects.
Fueling the Machine: The Nuclear Push
To power the massive Prometheus (1GW) and Hyperion (5GW) superclusters due to come online throughout 2026, Meta has aggressively secured “baseload” clean energy.
- 6.6 GW Nuclear Deal: On the eve of the Meta Compute launch, the company inked agreements with three nuclear energy firms: Vistra, TerraPower (Bill Gates-backed), and Oklo (Sam Altman-backed).
- 24/7 Reliability: Unlike solar or wind, nuclear provides the constant “always-on” power required by frontier AI training runs that can last for months at a time.
Strategic Acquisitions & Industry Context
The launch of Meta Compute comes alongside several other aggressive moves by the company in early 2026:
- Manus Acquisition: Meta recently acquired Manus, a high-growth AI agent startup, for over $2 billion to accelerate its consumer-facing “agentic” capabilities.
- Scale AI Investment: Meta holds a significant 49% stake in Scale AI, ensuring a steady stream of high-quality labeled data for its training clusters.
- Reality Labs Pivot: Reports indicate Meta is cutting roughly 1,500 jobs (10%) from its Reality Labs (Metaverse) division to further consolidate resources into the Meta Compute and AI infrastructure teams.
What This Means for the Future of AI
By establishing Meta Compute as a standalone initiative reporting directly to the CEO, Meta is signaling that the bottleneck for AI progress is no longer just code—it is power and land.
While Google and Amazon leverage their existing cloud businesses, Meta is building a sovereign compute empire designed to support its vision of open-source superintelligence that can be deployed across its 3 billion+ users on Facebook, Instagram, and WhatsApp.
