Nvidia and Amazon are making massive infrastructure bets to feed artificial intelligence's voracious electricity demands. Nvidia is investing up to $3 billion in Lancium, a Texas-based power developer that already manages four gigawatts of contracted capacity. Amazon is building a gas-fired power plant in Texas with capacity up to 7.65 gigawatts, which would produce an estimated 33 million tons of CO₂ annually, positioning it as the country's dirtiest power plant.
These moves reflect a hard reality: training and running large AI models consumes staggering amounts of electricity. Data center power consumption has become a bottleneck for AI expansion. Rather than wait for utility companies to build infrastructure, tech giants are taking direct action to secure dedicated power supplies for their operations.
Lancium operates an established model, developing power generation alongside data centers in Texas, where deregulated electricity markets and available land make the state attractive. Nvidia's investment signals confidence in this approach while giving the chipmaker priority access to power for its AI infrastructure customers.
Amazon's gas plant represents a more aggressive strategy. A 7.65 gigawatt facility rivals major utility-scale power stations. The carbon footprint raises environmental concerns as tech companies simultaneously commit to net-zero emissions targets. This contradiction highlights the tension between climate pledges and AI growth.
The investments underscore that AI infrastructure is becoming as much about energy as computation. Without reliable power, silicon sits idle. Both companies recognize this constraint and are betting billions to remove it from the equation.
Texas dominates this buildout because its deregulated grid allows companies to develop private power generation without utility middlemen, and abundant land accommodates sprawling facilities. Other states with regulated grids present higher barriers to direct power development by tech firms.
These deals will likely accelerate as AI workloads grow. Other hyperscalers including Google