Daniel Romero

Daniel Romero

AI Small Cap Energy Supplier

xAI has contracted this $3B small cap for the largest data center in the world

Daniel Romero's avatar
Daniel Romero
Sep 24, 2025
∙ Paid

This article will be part of a series on data center suppliers.

The subscription pays for itself. Start improving your returns.



AI capex is on track to reach $1 trillion per year by the end of the decade.

Image

These estimates from Goldman Sachs date back to June:

Image

Since then we’ve seen Meta raise its capex outlook, Oracle announce a $300B deal with OpenAI, and NVIDIA commit $100B to provide OpenAI with capital to buy GPUs.


We’re still in the early innings of the AI supercycle, which is why it makes sense to revisit some data center suppliers.

I believe the industry remains tremendously undervalued. Many investors hesitate because these names have already run up, but that mindset is a mistake. It’s the anchoring fallacy, and it has stopped people from making money.

Why do I say this industry still trades cheap? Let’s look at the valuations of some key suppliers for data center buildouts (EV/EBITDA):

  • Vertiv 25x

  • nVent 20x

  • Modine Manufacturing 17x

  • AMD 28x

  • NVIDIA 28x

  • Micron 7x

  • Dell 9x

  • Super Micro Computer 13x


All of these companies are core suppliers to large-scale AI infrastructure deployments. The same industry is seeing massive capex plans, yet the multiples don’t reflect this wave of spending. The market still isn’t valuing the sector for what it is: the foundation of AGI. A supercycle of investment to replace human labor with AI, both intellectual and physical.

Even if you aren’t hyper-bullish on AI, the reality is clear: these capex plans are locked in, and growth will be explosive, even if the investments don’t end up being accretive. The pick-and-shovel suppliers profit when gold is being chased, not only when it’s found. Earnings will come one way or another. With valuations still reasonable, and this being a decade-long opportunity, studying these companies for investment makes a lot of sense.


On the deployment side, the biggest non-compute bottleneck is energy. This shortage is already pushing up electricity prices across the US, straining both companies and households.

The image is a line graph showing the average electricity price in US cities, measured in US$/kWh, from 2016 to 2026. The y-axis ranges from 0.13 to 0.19, and the x-axis spans years, with a sharp upward trend starting around 2022. The graph, sourced from the Federal Reserve Bank of St. Louis, includes a watermark from BCZ Research 2025. Peter Berezin\'s post text suggests concern about data center capital expenditure hitting an "energy wall," implying rising electricity costs could impact technology infrastructure investments.

Of course, much of this comes from the war in Ukraine, but data center energy demand as a share of total consumption is expected to double by 2030.

The image is a line graph depicting the projected growth of US data center energy consumption as a share of total US power demand, from 5% in 2023 to over 10% by 2030. The graph, sourced from Apollo, shows a steadily rising green line against a timeline from 2023 to 2030, with percentage values on the y-axis (3% to 12%) and years on the x-axis. The post text from The Kobeissi Letter highlights this increase, driven by digitalization and AI, with McKinsey estimating the share will more than double in five years, and data center load potentially accounting for 40% of new demand by 2030. An "Apollo" watermark is present, indicating the source.

The demand for electricity is so high that data center developers are desperate for energy approvals, yet securing them can take years, a timeline many cannot accept. This is why valuations of ex-Bitcoin miners have surged.


Bitcoin mining is no longer as profitable as it once was, but these companies already hold capacity approvals that allow them to draw hundreds of megawatts from the grid. They were sitting on a diamond mine without realizing it, and now they have the opportunity either to be acquired or to use that capacity directly for AI workloads.

Beyond this, companies that know how to position themselves as electricity providers for data center operators will be some of the biggest winners of this cycle. That’s exactly the case with the company I’ll be covering today.

I first came across it in SemiAnalysis’ latest article on the xAI Colossus 2 buildout. I already knew Elon Musk planned to use gas turbines to speed up energization, but I didn’t realize the supplier was a $2.5B Texas-based company.


Given the level of interest in opportunities like this, I’ll break down what the company does and how large the opportunity could be.

Subscribe and gain full access

Keep reading with a 7-day free trial

Subscribe to Daniel Romero to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2025 Daniel Romero · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture