Fred Wilson, co-founder of Union Square Ventures, suggests that energy is a more straightforward investment for AI exposure than traditional technology stocks. He highlights that natural gas powers the majority of new AI data centers, yet its market price has remained relatively stable or declined.
Data shows that the United States currently hosts 1,391 operating data centers, with 2,819 additional facilities in the planning stages. Natural gas is projected to provide nearly 75% of the power for these planned on-site data centers, while nuclear energy, fuel cells, and batteries account for the remainder.
The Henry Hub benchmark for natural gas has fluctuated between $2 and $4 per million British thermal units (MMBtu) over the last decade. Wood Mackenzie forecasts that prices could reach approximately $5 per MMBtu by 2035 as overall power demand continues to rise.
Copper has also been identified as a potential asset for AI exposure due to rising demand, despite reaching record highs in 2026. However, market interest remains limited, and demand is currently tempered by weaker industrial data from China, which is the world's largest consumer of the metal.
The shift in energy demand is also impacting the economics of Bitcoin mining. If natural gas prices increase, energy producers may secure a larger portion of AI-related capital expenditure compared to current chip manufacturers, who have recently faced a period of market decline.
