TL;DR
Meta has created a new unit called Meta Compute to coordinate planning, deployment and operations of a large AI datacenter buildout. The group is led by Santosh Janardhan and Daniel Gross and will work with new President and Vice Chairman Dina Powell McCormick on government and financing partnerships.
What happened
Meta announced a new initiative, Meta Compute, to centralize the planning, deployment and operation of the company’s expanding AI datacenter footprint. CEO Mark Zuckerberg framed the effort around an ambition to add tens of gigawatts of capacity this decade and scale to hundreds of gigawatts over time. Santosh Janardhan, currently head of global infrastructure, will continue managing technical architecture, software, silicon, developer productivity and datacenter and network operations. Daniel Gross, who joined Meta’s Superintelligence team in mid‑2025, will lead long‑term capacity strategy, supplier partnerships, industry analysis, planning and business modeling. Both executives will coordinate closely with Dina Powell McCormick, recently appointed President and Vice Chairman, who will focus on partnering with governments and sovereigns to build, invest in and finance the infrastructure. The announcement accompanies large capital plans and recent long‑term nuclear energy contracts tied to Meta’s power needs.
Why it matters
- A multi‑gigawatt datacenter program raises Meta’s electricity needs and influences its energy procurement and local planning.
- Centralizing infrastructure strategy could become a competitive advantage if Meta executes on scale and cost control.
- Close engagement with governments and sovereigns implies major public‑private partnerships and potential regulatory scrutiny.
- Long‑term power contracts, including nuclear deals, show Meta is securing energy sources to match projected compute demand.
Key facts
- Meta formed a new unit named Meta Compute to oversee AI datacenter planning, deployment and operations.
- Mark Zuckerberg said Meta plans tens of gigawatts this decade and hundreds of gigawatts or more over time.
- Meta Compute will be led operationally by Santosh Janardhan and strategically by Daniel Gross.
- Dina Powell McCormick joined Meta as President and Vice Chairman and will focus on government and financing partnerships.
- Meta forecast $72 billion in capital expenditure for fiscal 2025 and expects higher spending the following year.
- Meta is working on gigawatt‑scale datacenter projects in Ohio, Louisiana and Texas, among other locations.
- The company has signed long‑term nuclear energy contracts with TerraPower, Oklo and Vistra; combined with existing deals, Meta has contracted roughly 6.6 gigawatts of nuclear power.
- Source reports say Meta’s Llama 4 met a lukewarm reception and the company faces competition for talent and uncertainty around next‑gen foundation models.
- Zuckerberg has reportedly shifted from open Llama models toward proprietary projects codenamed Avocado and Mango while still releasing some models like the Segment Anything Series.
What to watch next
- Progress and timelines for the gigawatt‑scale datacenter projects in Ohio, Louisiana and Texas (not confirmed in the source).
- Delivery and integration of the contracted nuclear power to meet Meta’s projected compute needs (delivery schedules not confirmed in the source).
- Status and public availability of Meta’s reportedly proprietary codenamed models Avocado and Mango (not confirmed in the source).
Quick glossary
- Datacenter: A facility that houses computer systems and associated components such as storage, networking and power infrastructure.
- Gigawatt: A unit of power equal to one billion watts, commonly used to describe large‑scale energy capacity.
- Capital expenditure (capex): Money spent by a company to acquire, maintain or upgrade physical assets like buildings, servers or networks.
- Foundation model: A large AI model trained on broad data that can be adapted for many downstream tasks.
- Long‑term power contract: A multi‑year agreement to purchase electricity or capacity from an energy provider, often used to secure supply and price stability.
Reader FAQ
What is Meta Compute?
A newly created Meta unit responsible for planning, deploying and operating the company’s AI datacenter infrastructure.
Who will lead Meta Compute?
Santosh Janardhan will handle technical and operational architecture; Daniel Gross will manage long‑term capacity strategy and business modeling. Both will work with Dina Powell McCormick.
How much power does Meta plan to add?
Zuckerberg stated plans for tens of gigawatts this decade and hundreds of gigawatts over time.
Are Meta’s next models open source?
The source reports a reported shift toward proprietary models codenamed Avocado and Mango, while noting Meta still releases some models like the Segment Anything Series.
When will the new datacenters come online?
not confirmed in the source
OFF-PREM Zuck forms Meta Compute to pave the planet with 'hundreds of gigawatts' of AI datacenters No wonder he's going nuclear Tobias Mann Mon 12 Jan 2026 // 22:21 UTC Meta has formed a…
Sources
- Zuck forms Meta Compute to pave the planet with 'hundreds of gigawatts' of AI datacenters
- Meta Launches Meta Compute to Boost AI Data Centers …
- Mark Zuckerberg says Meta is launching its own AI …
- Meta launches new "Meta Compute" initiative to build AI …
Related posts
- Meta to cut hundreds of Reality Labs metaverse roles as AI focus shifts
- Mark Zuckerberg launches Meta Compute to scale the company’s AI infrastructure
- Nvidia and Eli Lilly to open Bay Area AI drug lab with Vera Rubin GPUs