25 Comments
User's avatar
Matt's avatar

Great analysis. You didn't mention behind the grid power generation as a major source of short-term energy supply. Specifically, utility scale natural gas power plants that are only providing power to data center and that are not necessarily connected to the grid. Many of the players you mentioned I'm sure will benefit from AI data center growth but what are your thoughts about energy producers playing a bigger role in the short term? It seems that the tech players are focusing on natural gas to supply the AI data center power load as the most cost effective and quickest way to scale. This was missing from your analysis and I think is worth a second look.

Expand full comment
Djoker Nole's avatar

Great article on the topic of natural gas fueling hyperscaler DC's is on the A16Z substack, called "Gas-Fired Intelligence". Highly recommend

Expand full comment
Market Sentiment's avatar

Interesting. Are there any big public players that are doing this?

I specifically ignored oil and gas companies as they are excessively affected by the commodity pricing and also we won't be able to accurately judge how much of their reveue is coming from data center deals.

Expand full comment
John Stewart's avatar

GEV and BE seem to be serving the behind-the-meter and Bring Your Own Power (BYOP).

Expand full comment
Jerry Keefer's avatar

When considering natural gas you should also consider the possibility of being curtailed due to gas pipelines serving residential heating before industrial gas fired generation. Take a good look at what happened in 2022 with Winter Storm Elliot. There are so many parameters to consider.

https://www.naes.com/winter-storm-elliott-2022-the-impact-on-the-u-s-power-grid-and-recommendations-for-resilience/#:~:text=The%20winter%20storm%20of%20December%202022%2C%20known,challenges%20for%20both%20electric%20and%20gas%20utilities.

Expand full comment
Joshua S's avatar

Matt, this was my thought as well. The only way I see the anticipated data center/AI build out working is if they include behind the grid power generation in their data center build out. Essentially building power plants, attached to natural gas pipelines, on their data center grounds. I can’t imagine the US grid being able to handle the anticipated demand any time close to data center build timelines.

Expand full comment
William's avatar

Yes I like your ideas. Pick and shovel players make sense to me thanks!

Expand full comment
Market Sentiment's avatar

Thanks William.

Expand full comment
Bunny's avatar

In your model there's "% of Portfolio" only with no absolute portfolio value based on QTY of holdings - however, one can still suss out the median Dividend Yield to come in at ~1.12% This ups your overall blended return. The reason I bring this up is if you add in your big gas distribution networks and producers (i.e. ENB, XOM) with their long term, contracted take-or-pay revenue streams then your total portfolio would reflect larger returns with lower risk based on a larger ecosystem. Great work BTW.

Expand full comment
Shivaram Y.S's avatar

Great. This research documents and reinforces with what solar / wind and any renewable needs to be prioritized and resonates with growing data storage requirements. ❤️

Expand full comment
Market Sentiment's avatar

Thanks.

Expand full comment
Paul Watkins's avatar

This is amazingly useful. Congrats

Expand full comment
Market Sentiment's avatar

Thanks Paul.

Expand full comment
AT's avatar

Yes, very well done. Would love to be able to track your portfolio in real time - is that possible?

Expand full comment
Ryan Hunt's avatar

This is excellent research! Fantastic article

Expand full comment
Market Sentiment's avatar

Thank you!

Expand full comment
SE Nguyen's avatar

Do we have a good ETF for energy industry, similar to this portfolio?

Expand full comment
Disciplines collide's avatar

Love the thorough analysis and insights! Super helpful

Expand full comment
Ang's avatar

Great article. Amazing how quickly net zero fascination seems to be disappearing 😂 I’m liking a PM mining and power combo.

Expand full comment
Ik's avatar

Great post, but I quibble with statement about “operating leverage.”

“AI models are expensive to build and train, but once they’re ready, serving additional users costs very little.”

Of course, creating the models are a significant expense. But, inference is no trivial matter either. In fact, many retail customers (not on the API)—even the small percent that are paying for a monthly subscription does not cover the cost generated by their inference use.

Expand full comment
Craig A Williamson's avatar

Agree with the energy thesis. Surprised the nuclear sector was mostly left out of your discussion and portfolio. Given nuclear is compelling as a long term solution why wasn’t it included?

Cameco (CCJ), NexGen (NXE) and sector ETF’s URA and URNM are examples worth exploring

Expand full comment
Bunny's avatar

Re: spreadsheet - Nuscale Power Corp. SMR

Expand full comment
Market Sentiment's avatar

??

Expand full comment
Bunny's avatar

SMR is Nuscale's symbol. You had FLNC.

Expand full comment