Powering AI: Why Natural Gas is Emerging as the Pragmatic Solution
The boom in artificial intelligence and supercomputing is sparking a wave of new data-centre construction across the globe. These hyperscale facilities, each running tens of thousands of servers around the clock, need enormous amounts of electricity. In the U.S. alone, data centres are estimated to account for 6.7%–12% of national power demand by 2028 — up from just 4% in 2022.
But building and operating these facilities hinges on access to clean water and reliable power. And right now, power is the critical pinch point. More than 134 GW of proposed U.S. data-centre load is sitting in interconnection queues, up sharply from last year and well beyond what the grid can realistically deliver in the near term.
Our new report, AI’s Power Problem: Grid Strain, Gas Solutions, and Investor Strategies, takes a closer look at these challenges and highlights how natural gas and innovative investment structures are emerging as the most pragmatic near-term solutions.
The Power Gap
On the supply side, solar, battery, and gas projects are all being lined up to meet this surge in demand. Yet the timelines for most renewable and transmission projects are much longer than those for data-center buildouts. Developers cannot afford to wait, which is why natural gas is moving to the forefront.
Gas-fired generation offers the characteristics data-centers need most:
• Reliability for constant baseload demand, with flexibility to ramp up and down quickly.
• Speed — plants can often be built within 2–4 years, compared with a decade or more for nuclear or major transmission upgrades.
• Economics — in many regions, combined-cycle gas turbines remain cheaper than renewables paired with storage.
• Supply depth — with only 7% of U.S. reserves drilled today, the country has decades of gas available to support growth.
Of course, gas comes with its own considerations. Economics depend on pipeline access and regional pricing, projects often carry higher upfront costs, and emissions remain under scrutiny. But as a bridge to the future, gas remains the most bankable and scalable solution for powering AI today.
Where Investors Fit In
The scale of investment required is staggering. McKinsey estimates that $1.3 trillion will flow into “energizers” for AI infrastructure by 2030, covering power, transmission, cooling, and equipment. And the deals are already taking shape:
• Brownfield redevelopments turning retired coal sites into gas-anchored energy campuses.
• Mobile “power-as-a-service” deployments offering flexible, fast-response solutions.
• Greenfield “power islands” designed to self-supply entire hyperscale campuses.
• Hybrid utility partnerships spreading CAPEX and risk.
• Behind-the-meter plants (modular turbines, CHP, fuel cells) giving operators direct control over their power supply.
Each of these strategies comes with different return targets and risk profiles — but together they point to the same conclusion: AI’s growth is reshaping energy investment, and natural gas is central to that story.
Our new report, Powering AI Data Centers: Grid Bottlenecks, Natural Gas Solutions, and Investor Strategies, dives into:
• The scale of AI-driven electricity demand and grid strain.
• Why natural gas is the most pragmatic near-term solution.
• Real-world investment strategies and case studies.
• Risk-return dynamics across different deal structures.
