An energy dream team of behemoth Chevron Corp., Engine No. 1 and GE Vernova, tasked to create a multi gigawatt-scale co-located power plant and data center, is the latest power grab to produce energy sources for artificial intelligence (AI).
The first projects, called “power foundries,” are designed to integrate lower-carbon technologies such as carbon capture and storage capable of capturing more than 90% of CO2 emissions from the turbines, along with renewable energy options.
The joint development initially plans to deliver four gigawatts – enough to power to 3 million to 3.5 million U.S. homes – via co-located data centers in the Southeast, Midwest and West, by late 2027.
“Energy is the key to America’s AI dominance. By using abundant domestic natural gas to generate electricity directly connected to data centers, we can secure AI leadership, drive productivity gains across our economy, and restore America’s standing as an industrial superpower,” Chris James, founder and chief investment officer of investment firm, Engine No. 1, said in a statement. “This partnership with Chevron and GE Vernova addresses the biggest energy challenge we face.”
The challenge is illustrated in what is essentially a land grab by nearly every major tech company, governments, investment firms, and energy conglomerates for a place in the booming AI ecosystem. Exxon Mobile Corp., like Chevron, continues to advance behind-the-meter power-generation projects and discussions with hyperscaler tech firms desperate for reliable supply to meet vast energy demands for AI.
On the heels of DeepSeek’s recent emergence as a likely catalyst in high-volume AI use at a modest price, energy demands are expected to soar in the U.S. and abroad as more enterprises and individuals flock to take advantage of the technology.
“The DeepSeek model’s innovative engineering techniques enabled its developers to use relatively cheap hardware, improve compute and energy efficiency, and offer performance comparable to existing LLMs,” S&P Global Ratings analyst Miriam Fernandez said in the report. “That suggests that AI remains prone to disruption.”
At the same time, President Donald Trump’s executive order last month to remove government policies he considers “barriers to American AI innovation” is likely to spur what is already breakneck AI development. Trump’s order, in fact, preceded a $500 billion infrastructure joint venture between OpenAI, Oracle Corp., and SoftBank called Stargate that is intended to mass produce data centers. [Amazon.com Inc. alone has vowed to spend $150 billion on data centers over the next 15 years.]
Demand for data centers worldwide could grow by 240GW (some 240 billion gigawatts) between 2023 and 2030, with the U.S. accounting for 12% of all power generated, according to management consultant McKinsey. Most data centers used to train AI models currently under construction are in the 100MW and 250MW (100 million to 250 million watts) range, and are much much larger than anything built before, clean energy entrepreneur Michael Liebreich recently shared in a BloombergNEF piece.
