(MEMO) Memo - The Bull Thesis On Natural Gas, AI-Powered

By: Wilson

Artificial intelligence is the new buzzword in business, but while the tourists are fascinated by the marvels of the new world-breaking technology, the people who are creating AI are fixated on an increasingly difficult issue - power.

Sam Altman, CEO of OpenAI, recently asked the Biden administration to help fund and build 5 gigawatt data centers. To give you an idea of how much power 1 gigawatt is, it's the equivalent of 100 million LED bulbs. This page from energy.gov is especially helpful.

What's more fascinating about the AI arms race versus the traditional data center arms race is that AI requires a data center that's capable of handling massive amounts of power demand needs in one location. So while engineers try and figure out ways to make AI more power efficient, the hyper scalers (Facebook, Google, Amazon, and Microsoft) will look for ways to build a larger data center that uses a cheap enough energy source and is economically sensible. Perhaps that's why both Google and Microsoft are looking at nuclear power as an alternative source of energy.

But in the meantime, the boom we are seeing in AI won't stop anytime soon, and power burn needs will likely vastly surprise to the upside. This is why I have a hard time grasping what S&P is trying to predict with its drop in natural gas power burn demand assumption by 2029.

The contrast between S&P's forecast of -2.7 Bcf/d and EQT's +10 Bcf/d forecast is probably the largest I've ever seen in any commodity.