Today, the generative AI competition has shifted from algorithmic showdowns to a race for computing power. Meta is now making an unprecedented bet on its infrastructure. Following last year's commitment that "AI infrastructure will be a core competitive advantage," the company recently launched a global AI infrastructure initiative called "Meta Compute." CEO Mark Zuckerberg announced on Threads that Meta will build dozens of gigawatts (GW) of dedicated energy and computing capacity this decade, and plan for hundreds of gigawatts or even more in the long-term infrastructure system.

As a reference, 1 gigawatt of electricity can power about 750,000 U.S. households. According to industry forecasts, the total power consumption of U.S. AI data centers is expected to surge from around 5 gigawatts today to 50 gigawatts in the 2030s. Meta's move means it will directly participate in this unprecedented battle for energy and computing power, integrating electricity, chips, data centers, and network architecture into its strategic vision.

To ensure the realization of this grand project, Zuckerberg personally appointed three key executives to form a "triumvirate":

- Santosh Janardhan (Meta's global infrastructure head) will lead the technical architecture, self-developed chips (Silicon Project), software stack, and the construction and operation of global data centers and networks;

- Daniel Gross (co-founder of Safe Superintelligence, joined Meta in 2024) will be responsible for long-term capacity strategy, supply chain partnerships, industry analysis, and business modeling;

- Dina Powell McCormick (former government official, current President and Vice Chairman of Meta) will focus on coordinating with governments worldwide to promote policy support, investment, and financing for infrastructure.

This setup clearly indicates that Meta no longer wants to rely on cloud services or external suppliers but aims to build a fully integrated, self-controlled AI infrastructure ecosystem. This move also reflects the industry trend: Microsoft is actively partnering with AI infrastructure providers, while Alphabet, Google's parent company, acquired data center company Intersect in December 2025. Tech giants are increasingly viewing "computing sovereignty" as the vital point of competition over the next decade.

With the launch of the Meta Compute plan, the battlefield of the AI competition has extended from laboratories to power plants, chip factories, and government negotiation tables. Whoever controls the fundamental aspects of energy and computing power may define the form and boundaries of the next generation of AI.