The next AI arbitrage: Betting on energy, not on algorithms

The next AI arbitrage

In my earlier articles, I discussed the growing fizziness in the AI sector and how capital is increasingly chasing narratives rather than fundamentals. We drew parallels with the dot-com era, emphasizing that while technological breakthroughs are real, they don’t always equate to sustainable shareholder returns.

Whether today’s Artificial Intelligence (AI) enthusiasm eventually becomes a bubble or cements itself as a genuine long-term growth story is something only time will reveal. But there is one way to benefit from this frenzy without betting directly on AI stocks, that is, by investing in the enablers powering AI’s explosive rise. And at the foundation of this entire ecosystem lies one critical input: electricity.

The AI ecosystem rests on many layers - GPUs, software, data, cloud infrastructure, and talent. But the final, invisible layer is what holds everything together: a continuous, reliable power supply. AI data centres are energy-hungry, requiring uninterrupted power to operate high-density compute clusters and sophisticated cooling systems. Their electricity needs can be met through three broad sources: fossil fuels (60%), nuclear (18.6%), and renewables (21.4%).

Fossil fuels like Natural gas (43%) and coal (16%) power a bulk of the US Electricity needs, where the maximum AI data centres of the world are located.

If you want to keep the data centres running, then you need a cheap and reliable source of energy. Renewables are neither cheap nor reliable. Nuclear is cheap and reliable, but takes time to add capacity. This leaves us with fossil fuels, which are cheap and reliable. With the kind of AI data centres coming up in the US, the electricity needs are going to shoot through the roof. And natural gas is something that will have to fire up the power plants till the time nuclear capacities are ramped up across the United States. This leaves us with a great opportunity that the market participants have missed so far.

The first chart clearly shows that the energy sector’s weight in the S&P 500 has collapsed to around 2.6%, far below the 15% levels seen in 2008 and well under its peak during the last super-commodity cycle of the 1980s-1990s. In contrast, technology has expanded to dominate the index. This structural under-representation suggests that ownership in energy remains depressed relative to historical norms, even as AI’s power needs surge globally.

The S&P 500 Energy Index versus the S&P 500 Information Technology Index highlights that energy relative to technology is near its 2021 historical lows. Given the massive rally in tech-especially AI-linked names, this ratio sits at an extreme. Any rotation or even partial normalization could lead to energy outperforming technology in the coming cycle.

Natural gas futures are approaching levels last seen during the December 2022 spike. The recent upward momentum hints at a potential inflection point, often seen during early phases of broader commodity bull runs. Combined with strengthening demand from global LNG markets and rising consumption from data centres, natural gas futures appear to be pricing in a favourable demand outlook.

AI may or may not fulfil its most ambitious promises, but the infrastructure powering AI is set for an unavoidable and sustained boom. As global data-centre loads double, the world will need more electricity.

Yet energy remains one of the most under-owned and undervalued segments of global equity indices. This creates a rare opportunity to invest in the power behind AI and not just the AI narratives.

Even if AI enthusiasm cools, global electricity demand will keep rising through electrification, cloud expansion, and industrial transitions. That makes the energy sector a uniquely positioned, low-risk way to benefit from the AI era-whether or not the hype holds.