From Capital Expenditure to Strategic Imperative

When Meta disclosed last year that it expected to spend as much as $72 billion on capital expenditures to support its AI ambitions, the message was clear: the company intended to compete at the highest level of the generative AI race, regardless of cost. At the time, CFO Susan Li framed the spending as foundational rather than optional, calling advanced AI infrastructure a “core advantage” in developing the best models and products.
On Monday, Meta took the next logical step. CEO Mark Zuckerberg announced the launch of Meta Compute, a new internal initiative focused on dramatically expanding the company’s AI infrastructure and, just as critically, its access to energy. The announcement marks a shift from abstract investment promises to a concrete, long-term buildout plan that could reshape both the tech industry and the power grid.

An Energy Footprint Measured in Gigawatts

Zuckerberg’s most striking claim was not about servers or chips, but electricity.
“Meta is planning to build tens of gigawatts this decade, and hundreds of gigawatts or more over time,” he wrote, emphasizing that infrastructure engineering, investment, and partnerships would become a lasting strategic advantage.
To put that in perspective, a single gigawatt is enough to power hundreds of thousands of homes. Industry analysts already warn that AI could push U.S. electricity demand from roughly 5 gigawatts today to more than 50 gigawatts by 2030. Meta’s ambitions alone would represent a significant portion of that projected growth.
This framing reflects a growing consensus across Big Tech: the future of AI competition may be constrained less by algorithms and more by power availability, land, permitting, and grid access.

Meta Compute’s Leadership Triumvirate

Zuckerberg also outlined a leadership structure that underscores how multidimensional the initiative will be spanning engineering, supply chains, geopolitics, and finance.

Santosh Janardhan: Building the Machine

Janardhan, Meta’s longtime head of global infrastructure, will oversee the technical backbone of Meta Compute. His remit includes data center architecture, software systems, custom silicon, developer productivity, and the operation of Meta’s global data center fleet and network. In effect, Janardhan will be responsible for ensuring that Meta’s AI ambitions are physically possible.

Daniel Gross: Planning for Scarcity

Daniel Gross, who joined Meta last year after co-founding Safe Superintelligence with former OpenAI chief scientist Ilya Sutskever, will lead a long-term capacity strategy. His group will handle supplier relationships, industry analysis, and business modeling areas that have become increasingly critical as AI hardware, energy contracts, and construction timelines grow more constrained and competitive.

Dina Powell McCormick: Navigating Governments

Dina Powell McCormick, Meta’s recently appointed president and vice chairman, will work with governments to help build, deploy, finance, and approve infrastructure. Her role highlights a reality tech companies can no longer ignore: AI infrastructure is now a public-policy issue, touching energy markets, national security, and regional economic development.

Meta Isn’t Alone but It’s Doubling Down

Meta’s announcement comes amid an industry-wide infrastructure arms race. Microsoft has aggressively partnered with third-party AI infrastructure providers. Alphabet has moved to acquire energy-focused data center firms to bypass grid bottlenecks. Apple, while comparatively restrained, is under pressure to clarify its own AI infrastructure strategy.
What distinguishes Meta is the scale and explicitness of its commitment. Rather than relying primarily on cloud partners, Meta is signaling a willingness to vertically integrate energy, compute, and AI development a model more reminiscent of industrial-era giants than modern software companies.

The High-Stakes Bet on Infrastructure as Destiny

Meta Compute reflects a deeper shift in how AI leadership is defined. Model quality still matters. Talent still matters. But increasingly, the decisive factor may be who can secure the physical resources required to train and run ever-larger systems.
By tying AI success directly to power generation, data center construction, and geopolitical coordination, Meta is acknowledging that the generative AI era will be won not just in research labs, but in negotiations with utilities, regulators, and suppliers.

Summary: AI’s Future Runs on Power, Not Just Code

Meta Compute is more than an internal reorganization. It is a declaration that the next phase of AI competition will be fought on an industrial scale. In committing to tens and eventually hundreds of gigawatts, Meta is betting that infrastructure scarcity, not innovation scarcity, will define the winners and losers of the AI era.
If Zuckerberg is right, the companies that dominate AI will look less like traditional software firms and more like energy-intensive industrial powerhouses. The real race, then, is no longer just for smarter models, it’s for the electricity to run them.