Countries and companies in the AI race are increasingly realizing they are in an energy race. On some days, it seems as if AI is driving energy, with AI data centers representing the most significant new power demand growth in decades. On other days, it seems as if energy is driving AI, with the largest technology companies scrambling to secure power and even vertically integrating to avoid external energy constraints. Whether AI or energy was in the driver’s seat was one of the questions at the SAIT AI Summit this past week, in Calgary, Alberta, the energy hub of Canada, as executives from major energy companies spoke about using AI tools while also working to supply the energy this technology requires.
“Energy companies are in the unique position of deploying artificial intelligence internally to enhance operations, while also providing energy to data centers powering this transformational shift in technology. Across our events and programs, we are seeing more integration and implementation of artificial intelligence at the executive level within the energy sector,” said Lora Bucsis at SAIT, Director, Product and Learner Success.
The increasing overlap between tech and energy is a welcome change for the energy sector, with the largest companies in the world, from Alphabet to Meta, now deploying all their cash, and then some, into the physical economy. Commodities have not benefited from a true capex boom in decades, but this is now changing. Goldman estimates that capex spending amongst the S&P 500 will accelerate by 17% this coming year, driven primarily by data centers and associated power needs.
Returns on AI have so far not kept up with the level of investment, but unique dynamics may sustain capital spending regardless of immediate returns. Governments view AI advantages as existential, so there is nothing stopping governments from printing money into AI-related capex to accelerate these potential future outcomes, however unproven. There is a long history of countries spending without any defined return when the investment provides a fundamental shift in economic or military capabilities. AI fits this pattern. In case there was any uncertainty the White House even released a briefing stating that “to remain the leading economic and military power, the United States must win the AI race.” Companies have a similar view where if AI is the last big platform shift, or if it provides a permanent corporate advantage, the risks to underinvesting are much higher than overinvesting. Alphabet CEO Sundar Pichai has said that “the opportunity with AI is as big as it gets.” The spending will continue.
Natural gas companies are now the ones looking like they will benefit the most from this spending, and it doesn’t hurt that major gas hubs in Alberta and Texas have actually had negative gas pricing at the exact time that these projects are looking for low cost and reliable supply. The delay for grid connectivity has also made on-site gas-fired generation appealing, as well as conversions.
“PJM and ERCOT are a perfect example of the natural gas bull story,” said Wish Bakshi, founder of AQ Energy, a consultancy that helps unlock the power of data and AI for commodities trading, power plants, renewables and LNG operations.
“PJM, home to Virgina (data center capital of the world), is fast tracking 10GW of new generation. 70% of this new generation is natural gas focused. Pennsylvania has positioned itself aggressively to capture AI data center investment, leveraging its massive Marcellus shale gas reserves to offer reliable, cost-effective energy supplies.”
“In ERCOT, quick build permits and cheap Waha prices (a major natural gas trading hub in Texas) is attracting massive AI Data Center investment. The Stargate project in West Texas, a $500 billion initiative from OpenAI and Oracle, incorporates natural gas-fired power generation as its primary energy source. Elon Musk’s xAI facilities, Colossus 1 and 2 in Memphis, Tennessee, utilize on-site gas turbines for power generation,” Wish continued.
Other regions are focusing on smaller scale gas projects that they believe balance the needs of investors and regional customers. This has also helped project owners deal with the backlog for larger turbines, by buying smaller turbines that are more readily available. S&P Global Commodity Insights has tracked this uptick in orders for smaller equipment, with orders for small gas turbines for North American data centers totaling about 1.9 GW over the past year. Including reciprocating engines and fuel cells, that number jumps to 3.2 GW. Manufacturers are investing in equipment capacity to meet these needs, with Mitsubishi announcing that they are doubling their gas turbine production.
Gas has the speed and scale to provide a natural solution to the rapid AI driven demand growth. Onsite turbines can be permitted and installed to meet project timelines, and the gas resource is proven across North America at a variety of prices. All of this is coming at a time when LNG exports across North America are coming to fruition, further accelerating gas demand. In a world of shiny AI investment opportunities, natural gas may be the winner over the next decade as energy executives increasingly find themselves on stage with technology executives.