Nvidia CEO Jensen Huang flags AI data-center power crunch as chip demand outpaces supply

Nvidia is warning that the next constraint on artificial-intelligence expansion may not be chips, but electricity. In recent remarks highlighted by Yahoo Finance, Nvidia CEO Jensen Huang said that getting power for AI data centers can take years, implying that even strong demand for Nvidia’s processors could be slowed by the pace of power-delivery projects.

Huang’s comments come as demand for Nvidia’s AI hardware continues to exceed what the company can supply, according to the report. That framing places the supply chain and production ramp of semiconductors alongside a slower-moving bottleneck on the infrastructure side, namely the ability to connect large computing facilities to sufficient grid capacity.

The same coverage also notes that Evercore ISI sees a potential investment angle from the power bottleneck narrative. Specifically, it suggested that Bloom Energy, a provider of power generation technology often associated with distributed energy systems, could benefit if data centers and other large users increase spending to secure and scale electricity faster.

Bloom Energy is not described in the coverage as a substitute for Nvidia’s chips, but rather as part of the broader effort to make more power available where and when AI compute is deployed. The implication is that as companies compete for grid capacity, they may place more emphasis on solutions that can be built and brought online more quickly than traditional grid upgrades, though the report does not provide detailed program specifics.

For Nvidia, the message matters because it affects planning across customers’ entire AI buildouts, not just semiconductor purchases. Nvidia’s business is closely tied to how quickly hyperscale and enterprise customers can deploy AI infrastructure, including the data centers that host training and inference workloads. If power becomes the rate limiter, the timing of new deployments, GPU installation schedules, and overall demand visibility could all become harder to predict.

More broadly, the AI hardware boom has increasingly drawn attention to energy as a strategic resource. Large clusters consume significant electricity and require robust, reliable power delivery. If power availability lags behind compute demand, companies may prioritize phased expansions, refurbish or repurpose sites that can reach capacity sooner, or seek infrastructure approaches that reduce dependence on long lead-time grid connections.

What Nvidia did not disclose in the reported remarks is any quantification of the time horizon in specific markets, the expected duration of the power constraint, or how much of its near-term AI revenue growth could be affected. The coverage also does not provide concrete details linking specific customer projects to Bloom Energy, such as named customers, contract sizes, or deployment timelines.

The next thing to watch is whether Nvidia’s customers and data-center operators translate the power-risk discussion into procurement and construction decisions that become visible in industry spending. Investors will likely look for additional commentary from Nvidia on demand drivers beyond chip supply, and for evidence that power infrastructure spend is accelerating in ways that could translate into measurable results for power-equipment and energy-system providers.