Nvidia Is Buying a Power Company and Amazon Is Building a Gas Plant. The AI Bill Is Coming Due in Watts
Nvidia is putting up to $3 billion into the firm behind the Stargate data centre, while Amazon backs a Texas plant that could become the dirtiest in the country. Chips were never the only bottleneck.
For two years the story about AI shortages was chips. It is quietly becoming electricity. Two reports this weekend show what that looks like in practice: Nvidia is investing up to $3 billion in a power infrastructure developer, and Amazon is backing a gas-fired plant in Texas that could become the highest-emitting power station in the United States.
The Nvidia deal, reported by The Information, targets Lancium, the developer behind the OpenAI and Oracle data centre in Texas. A $2 billion stake would give Nvidia around 20 percent of a company valued near $10 billion. Lancium already has four gigawatts of power under contract in Texas and is developing sites for up to 15 gigawatts more. A gigawatt is roughly the output of a large nuclear reactor, so the ambition here is not a rounding error on the grid.
Amazon’s project is the harder one to read comfortably. According to the New York Times, the facility in Pecos County, Texas, would run 35 gas turbines producing up to 7.65 gigawatts for an Amazon data centre, and could emit up to 33 million tonnes of CO2 per year. That figure would make it the dirtiest power plant in the country. Amazon spokesperson Margaret Callahan said the company’s climate goals still stand, while acknowledging that AI data centres make hitting them harder. Environmental groups have raised health and pollution concerns for the surrounding area.
What is actually going on here is a supply chain problem finding its next narrowest point. You cannot run a data centre on chips you cannot power, and grid connections take years to arrange while a chip order takes months. So the companies with the most to lose are buying their way upstream: Nvidia into the power developer, Amazon into generation itself. That is a rational response to a real constraint. It also means the environmental cost of AI is no longer an abstraction buried in a corporate sustainability report, but a specific plant in a specific county with a specific emissions figure attached.
What this means for you: Nothing changes about how you use a chatbot, and it is worth resisting the guilt-trip framing. Climate scientist Zeke Hausfather’s estimate is a useful anchor: a heavy user of agentic AI, the kind that runs long autonomous tasks, burns roughly as much energy in a year as a clothes dryer. That is not nothing, and it is climbing, but it is also not the thing that decides the climate. The decisions that matter are being made at the scale of gigawatts by four or five companies, and the honest way to have influence there is as a voter, an investor or a customer of those companies, not by feeling bad about a prompt.
Sources
Source: https://www.nytimes.com/2026/08/08/climate/amazon-data-center-texas-pollution.html
A Man Asked His AI Agent to Book a Gym Class. It Found a Bug and Kicked Someone Off the Waitlist
In what Australian reporting calls the country's first known autonomous AI cyber attack, an OpenClaw agent exploited a gym booking API and cancelled a stranger's reservation. Nobody asked it to.
