Sources: Paradox bought commitments for 1.7GW of gas-fired generators from equipment supplier OnePWR for $3.4B, the largest known such purchase by an AI lab (Martha Muir/Financial Times)

Paradox’s commitment covers 1.7 gigawatts of gas-fired generating equipment from OnePWR, for $3.4 billion. The purchase is notable because it is reportedly the largest known such commitment by an AI laboratory. It shows how urgently AI companies are seeking electricity supplies.

OnePWR is identified in the source as an equipment supplier. The deal is not described as a purchase of electricity itself. Instead, Paradox is securing generator equipment capable of producing power, using natural gas or another gas fuel in combustion engines or turbines.

The source presents the agreement as a way to bypass traditional suppliers while energy demand rises. It does not state where the generators will be installed, when they will operate, or whether the $3.4 billion covers construction and fuel. Those details would be needed to assess the project’s full cost and schedule.

A gas-fired generator is a machine that uses a gas fuel, usually natural gas, to produce electricity. Gas is burned in an engine or turbine. The resulting hot, expanding gases create mechanical motion, which drives a generator. The generator then converts that motion into electrical current.

For example, a turbine can spin at high speed while connected to an electrical generator. The turbine’s rotation turns the generator’s magnetic components, producing electricity that can flow to a data center or the wider grid. Some plants also capture heat for additional efficiency, though the source does not describe OnePWR’s equipment.

Gas generators can provide power when demand rises, which helps explain their appeal for energy-intensive data centers. They also require fuel deliveries and emit carbon dioxide and other pollutants. The source establishes Paradox’s planned capacity and supplier, but it does not specify the exact generator technology or fuel arrangements.

One gigawatt equals 1,000 megawatts, so 1.7 gigawatts equals 1,700 megawatts or 1.7 million kilowatts of maximum generating capacity. That is a very large block of power for one company to secure. Actual electricity production would depend on how often the generators run.

The arithmetic gives the deal a headline price of about $2,000 per kilowatt: $3.4 billion divided by 1.7 million kilowatts. This figure describes the commitment’s average price per unit of capacity. It does not necessarily represent the complete cost of a finished power plant, transmission links, fuel systems, or long-term operation.

The source calls the purchase the largest known such deal by an AI lab, but it provides no industry price benchmark. Therefore, the scale is clear, while judging whether the equipment is expensive or cheap requires more information about what the $3.4 billion includes.

AI companies need large, dependable electricity supplies for data centers. The source says energy demand is surging and describes generator equipment as scarce. That combination can make ordinary power arrangements too slow or uncertain for a company expanding its computing capacity.

A direct equipment commitment gives Paradox a claim on 1.7 gigawatts of gas-fired generation from OnePWR. The key mechanism is supply-chain control: instead of waiting for a utility to build or procure enough capacity, Paradox secures equipment itself. It may still need sites, permits, fuel, grid connections, and construction.

This approach can shorten a critical bottleneck, but it does not eliminate every delay. The source does not say whether Paradox will own, operate, or finance the generators. It also does not identify the data-center locations or explain how the equipment will connect to electricity networks.

Large gas generators built for AI data centers would add new electricity-producing capacity. If they operate, they would consume more gas and produce carbon dioxide, along with other combustion emissions. They could also reduce pressure on local grids by supplying power close to the data centers.

The mechanism is a trade-off. More generation can ease shortages and sometimes lower prices during tight periods. But a huge new AI load can raise demand, while fuel costs and limited pipeline or grid capacity can push prices upward. If generators run frequently, their emissions and gas consumption will be greater than if they operate only during peaks.

The source gives no forecast for electricity prices, gas demand, or emissions. It only identifies Paradox’s planned 1.7 gigawatts of gas-fired equipment. The eventual effects will depend on operating schedules, local infrastructure, fuel markets, and whether the generators replace or supplement grid power.

Other power sources could include electricity from the existing grid, solar and wind farms, nuclear plants, hydropower, batteries, or other gas and fuel technologies. A data center could also combine several sources. The best mix would depend on location, reliability needs, cost, and available infrastructure.

Renewables can provide low-carbon electricity, but their output varies with weather and time. Batteries can shift power for limited periods, while nuclear and large transmission projects generally take years to develop. Grid supplies may be constrained if utilities cannot add capacity fast enough. These are general energy considerations, not details provided about Paradox’s plans.

The source’s central point is urgency: energy demand is surging, and equipment is scarce. Directly committing to OnePWR’s gas-fired generators may therefore be faster than waiting for new alternatives. The source does not say whether Paradox is pursuing renewables, nuclear power, batteries, or utility contracts as well.

AI training repeatedly processes enormous datasets through specialized computer chips. Running a model, often called inference, performs calculations whenever the system answers a request. Both activities use electricity, and demand rises with larger models, more users, and longer operating hours.

Inside a data center, electricity powers servers containing processors, memory, and networking equipment. The processors perform mathematical operations, memory stores data, and networks move information between machines. Cooling systems remove the heat produced by this work. Together, these systems turn electrical power into usable computing capacity.

The source does not explain Paradox’s models, data centers, or exact electricity needs. It does report a commitment for 1.7 gigawatts of gas-fired generators, linking the company’s expansion to the broader surge in energy demand from AI. Securing generation may help address that demand, but it also creates fuel and emissions consequences.