AI is no longer only a software race. For Nvidia and Amazon, the next bottleneck is electricity, and the scale of the response is moving the industry deeper into power infrastructure.
Power is becoming part of the AI supply chain
Nvidia is investing up to $3 billion in Lancium, according to The Information. Lancium is the power infrastructure developer behind the OpenAI and Oracle data center in Texas.
The reported deal shows how closely AI growth is now tied to access to large blocks of energy. Chips, cloud services and data centers still matter, but they cannot function without enough electricity to run them continuously.
A $2 billion stake would give Nvidia roughly 20 percent of Lancium. The company would be valued at around $10 billion under that scenario.
Lancium already has four gigawatts of power under contract in Texas. It is also developing sites for up to 15 more gigawatts, which signals that the company is positioning itself around AI-scale energy demand rather than ordinary computing growth.
Amazon is backing a major Texas power plant
Amazon, meanwhile, is backing a massive gas-fired power plant in Pecos County, Texas. The facility is designed to supply an Amazon data center.
According to the New York Times, the plant could emit up to 33 million tons of CO₂ per year. That would make it the dirtiest power plant in the country.
The project includes 35 gas turbines designed to generate up to 7.65 gigawatts of electricity. That figure places the plant in the same conversation as the enormous energy footprints now associated with advanced AI infrastructure.
Amazon spokesperson Margaret Callahan said the company's climate goals still stand. She also admitted that AI data centers could make hitting them harder.
That tension is central to the moment. Companies want to expand AI capacity quickly, but the power needed to support that expansion can create direct pressure on climate commitments.
The climate issue is no longer abstract
Environmental groups are raising alarms about health and pollution risks tied to the Amazon-backed power plant. Their concern is not simply that AI uses energy, but that the infrastructure built to feed AI systems can carry local and atmospheric costs.
The reported emissions figure gives those concerns a concrete scale. Up to 33 million tons of CO₂ per year is not a background detail; it is central to how the project will be judged.
The same dynamic applies across the wider AI sector. If companies need gigawatts of electricity for data centers, the source of that electricity becomes part of the technology story.
AI data centers are often discussed in terms of models, chips and cloud platforms. The Texas projects show another layer: land, turbines, power contracts and emissions are now part of the competitive landscape.
Agentic AI adds another layer of demand
Climate scientist Zeke Hausfather estimates that a heavy user of agentic AI burns about as much energy per year as a clothes dryer, and that number is climbing.
That comparison matters because agentic AI points toward systems that do more work on a user's behalf. If usage rises and each heavy user requires more energy, the pressure on data center operators increases.
This does not mean every AI interaction has the same footprint. The source only gives one estimate for a heavy user of agentic AI. But it does indicate why major companies are looking beyond ordinary data center planning.
The energy question is becoming practical. Who can secure power? Where can it be generated? How much pollution will it create? And how will companies balance expansion with public climate promises?
What this means for the AI race
The moves by Nvidia and Amazon point to a broader shift. AI infrastructure is becoming physical infrastructure.
For Nvidia, the Lancium investment would connect the chip leader more directly to the power systems that allow large AI data centers to operate. For Amazon, backing a gas-fired plant shows how cloud and AI growth can require dedicated electricity generation at enormous scale.
The pattern is clear from the facts in the source: AI data centers are demanding gigawatts, and companies are responding with billion-dollar investments, large power contracts and new generation projects.
That creates a more complicated future for AI. Performance and availability will still matter, but energy access, emissions and local health concerns will increasingly shape where AI systems are built and how they are judged.