When you think about the constraints facing artificial intelligence, you probably picture a shortage of advanced chips or maybe talent. But Nvidia Corp. (NVDA) CEO Jensen Huang has a different answer: electricity. On Wednesday's episode of The Joe Rogan Experience, Huang laid out a future where the biggest tech companies don't just buy power from the grid but generate it themselves using small nuclear reactors sitting right next to their data centers.

The Bottleneck Isn't What You Think

Rogan asked Huang directly whether energy has become AI's biggest obstacle. Huang's response was unambiguous. "It's the bottleneck," he said. Not GPUs. Not computing power. The limiting factor for how fast AI can scale is now the availability of electricity itself.

This is a pretty significant shift in the narrative. For years, the story has been about chip shortages and manufacturing capacity. Nvidia has been racing to produce enough advanced processors to meet demand. But according to Huang, we're entering a new phase where the infrastructure challenge isn't silicon, it's watts.

Rogan mentioned that Alphabet Inc. (GOOG) (GOOGL) is already building nuclear facilities to power AI operations. Huang said he hadn't heard that specific detail but wasn't surprised. In fact, he predicted that over the next six or seven years, "you're going to see a whole bunch of small nuclear reactors" pop up across the industry.

Compact Reactors Next to Data Centers

Huang isn't talking about traditional massive nuclear plants. He's describing something far more modular: compact reactors generating hundreds of megawatts, installed on-site where companies need them. The idea is that tech firms would essentially become their own power generators, reducing their dependence on the grid and ensuring a reliable, continuous supply of electricity for their increasingly power-hungry AI workloads.

"We'll all be power generators," Huang told Rogan, drawing a comparison to farms producing their own resources. He explained that generating power locally doesn't just solve the supply problem, it also reduces strain on the broader electrical grid. And when these companies produce more power than they need, they can feed the surplus back into surrounding communities.

Rogan called the approach "the smartest way to do it," and Huang agreed. As AI workloads continue to grow exponentially, building the capacity companies need and contributing excess power when possible will become essential.

Google Is Already Making Moves

Huang's vision isn't purely hypothetical. Google announced plans in 2024 to purchase power from Kairos Power, a developer of small modular reactors. The company is targeting its first advanced reactor to come online by 2030. A separate 2025 agreement between Kairos and the Tennessee Valley Authority marked the first time a U.S. utility committed to buying electricity from next-generation reactors.

This isn't a niche trend. Analysts are projecting that data center electricity consumption will jump 175% by 2030. Goldman Sachs has compared this surge to adding a new top 10 energy-consuming country to the global grid. That's not incremental growth, that's a fundamental shift in how much power the world needs to produce.

U.S. electricity demand overall is expected to grow 2.6% annually through 2030, a jump of 1.2 percentage points driven largely by data centers. To put that in context, the country has rarely seen power demand rise above 2% at any point over the past 20 years. AI isn't just changing how we compute, it's changing how much energy we consume.

What This Means for Hardware

For Huang, the takeaway is straightforward: AI's next frontier won't be defined by hardware breakthroughs alone. It will be defined by who can produce enough power to keep the machines running. You can have the most advanced chips in the world, but if you can't power them reliably, they're not much use.

This reframes the competitive landscape. It's no longer just about who can design the best processors or train the largest models. It's also about who can secure or generate the electricity to support those operations at scale. Energy infrastructure is becoming just as critical as computing infrastructure.

Market Response: Energy and Nuclear Stocks

The market is already pricing in this shift. Energy and nuclear-related stocks have posted strong gains year-to-date, reflecting growing investor confidence that power generation will be a major beneficiary of AI expansion. Here's how key stocks in the sector are moving in after-hours trading, along with their year-to-date performance: