In The Race To Power Data Centers, Which Fuel Will Win Is The Wrong Question

Forbes

13-06-2025

Greg Robinson is the cofounder and CEO of Aston , a company building a clean, firm power network to serve the world's largest energy users. getty
Having spent my career in clean energy, I have had the opportunity to talk to a range of energy experts. From leaders at utilities and independent power producers to regulatory decision makers, we nerd out on topics ranging from whether the future of power is on-grid or off-grid to how we'll meet surging energy demand and what the optimal fuel mix is to do this.
Lately, the thing I'm asked about most is which fuel I think will win the race to power the data centers that power AI. It's a fair question, but in a shifting energy landscape, it's one with a complicated answer.
The U.S. power market is a very complex business model, and the grid itself is a convoluted web of old tech like coal plants and new tech like solar and wind, making and transferring power to keep the proverbial lights on for individuals and businesses alike.
In my view, the idea that one fuel will 'win' is the wrong way of looking at it. The current landscape consists of a shrinking coal footprint; growing natural gas and renewables like wind, solar and batteries; and blossoming nuclear. Below, I share the context that gets us closer to understanding what the future holds. The Flaws And Faults Of Our Options
At the risk of stating the obvious, the energy industry can only consider the technologies that are available today. For this reason, there's often a 'use what you have' mentality. For example, there's a lot of talk about the potential of nuclear power to solve the surging energy demand, and small modular reactors (SMRs) will play a large role, but experts say that technology is still years away.
As a result, utilities and data center builders have been looking to gas, but a bottleneck in the production of turbines is slowing procurement until at least 2029. That, combined with the rising cost of building new plants, upgrading the pipeline network to provide enough gas is making gas less available and affordable. Where in 2022, it costs roughly $800/kWh to build a gas plant, projections are now that it will cost $2,400/kWh.
But even if building more gas plants were a timely and affordable solution, it is still a dangerous premise. Assets like gas and coal plants are 30-plus-year assets. So building new gas plants now means that once viable nuclear power arrives and the cost of solar and batteries continues to drop, we run the risk of doing to gas plants what we have been doing to coal plants: turning them off because we overbuilt the network and there are better options available. The Rush To Zero Marginal Costs
Businesses across various industries try to innovate their way toward zero marginal cost. They would rather pay larger capital costs upfront one time and then use that asset for a long time. AWS is a good example. During the rise of cloud computing, AWS had the innovative idea that rather than build and sell servers to companies, it would develop servers in the cloud and then rent that space to those who need it. Netflix is similar. Pay for the content once and then rent it out to millions of users.
Fossil fuel-based energy will never reach zero marginal cost because you need to process resources to use them. Nuclear also has a fuel but seldom needs to be refueled. Solar and wind backed by batteries are pushing the closest to zero marginal cost. The price of solar and batteries has come down nearly 100% and will keep getting less expensive.
This brings me to the real factor that will determine which fuel will win: It's the fuel with the most contracts. Which Fuel Will Win?
The energy market is contract-driven. Specifically, contracts that are known as 'take or pay.' This means when suppliers (like a natural gas producer) make a deal with buyers (like a utility or power plant), the contracts are long-term (usually 15 years at minimum), and the buyer agrees to take a predetermined amount of energy at a fixed price. If they don't take the energy, the buyer pays anyway. So it's really 'pay or pay' contacts.
This structure leaves little incentive for buyers locked into contracts to make any moves until the contract ends. If we build more gas plants, that will effectively make those operational long after less expensive modes of energy become available.
How we structure contracts needs to change. Long-term contracts made sense during the decades of relatively steady power demand, but they don't work in this era of surging power demand. Leaders in the energy industry and the large customers who depend on their power need to reconsider a new approach to energy contracts, or we'll find ourselves trapped in the past, which will ultimately impact our future.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?