AI Companies Race for Gigawatt-Scale Energy Amid “Bragawatt” Skepticism

The rapid expansion of artificial-intelligence models has created a parallel demand for electricity. Training large language models requires data centers that consume gigawatts of power, straining grids and financing for new infrastructure. Google, Meta, OpenAI and Anthropic are pursuing contracts and projects to secure the megawatt-hour capacity needed for next-generation AI workloads.

New York Times

The AI surge has outpaced energy growth. Limited supply and high capital costs for new power plants have made electricity procurement fiercely competitive. The competition extends to securing long-term power purchase agreements with utilities. Executives now publicly announce the energy they aim to secure, framing such pledges as strategic differentiators.

New York Times

The primary firms in this race are Google, Meta, OpenAI and Anthropic, each seeking power agreements, grid upgrades, or dedicated generation to meet AI compute needs. Waldemar Szlezak, head of digital infrastructure at private-equity firm KKR, has highlighted the term “bragawatt” in a Financial Times column, urging investors to focus on grid realities.

New York Times

AI model training currently requires multiple gigawatts of continuous power. Companies claim the ability to generate or secure future gigawatt-scale capacity, though project timelines remain undisclosed. Critics note that the term “bragawatt” is used to describe such speculative capacity announcements.

New York Times

Representatives from the four leading AI firms stress that electricity is central to long-term strategy, citing infrastructure investments to address grid realities. KKR’s Szlezak, in a Financial Times commentary, urged investors to “look past the A.I. hype and focus on the reality of today’s power grid,” highlighting the gap between announced ambitions and current capacity.

New York Times

Industry analysts and environmental groups label many announced projects as “bragawatts,” suggesting the claimed gigawatt-scale capacity may be more rhetorical than operational. Skepticism focuses on the speed of new generation assets, financing needs, and the risk of over-promising to attract capital.

New York Times

• “hype and focus on the reality of today’s power grid.” — Waldemar Szlezak, Head of Digital Infrastructure, KKR

New York Times

The AI boom’s reliance on power-hungry data centers could strain existing grids, limit the pace of model training, and force firms to allocate large portions of capital to energy procurement. Critics argue that inflated gigawatt claims may distort investor expectations and exacerbate uncertainty about future electricity demand.

New York Times