Cummins Signed a Data Center Power Deal. The Market Still Prices It Like an Engine Company.

Every AI infrastructure name that actually works is already expensive.

GE Vernova. Constellation Energy. TransDigm. Real businesses, real growth, priced years ahead of themselves. Run the numbers forward on any of them and the market has already paid for the good news before it arrives.

That is the trap in this trade right now. The story is true. AI needs power. Data centers need generators, batteries, and grid equipment nobody was building at this scale five years ago. But "the story is true" and "the stock is cheap" stopped being the same sentence a while back.

So we went looking for the name the story has not found yet.

A 105-year-old engine maker, not a chip company

Picture the machine underneath this story: the same engines that kick on when a hospital's power drops are now doing the same job for a data center, plus batteries built to carry the load in between. That is Cummins. Diesel and natural gas engines, over a century of them. Trucks, construction equipment, standby power. Not a headline name in this trade. Not a hot IPO.

In its most recent quarter, Cummins disclosed something specific: it was selected to supply battery energy storage systems for a large US data center project. Management said it plainly, on the earnings call: "robust customer orders for standby power for data centers" as a direct driver of record results. That is not an analyst's guess about future demand. That is a signed contract.

What that costs you today

At $587.68 a share, here is what you are buying, and it is worth digesting before you look at anything else in this trade. Revenue of $34.7 billion over the trailing year. Free cash flow of $3.4 billion, more than double where it stood a year ago. Return on invested capital of 15.7%. That ROIC does not clear the 25% bar we hold a true compounder to. It is a real number for a real, growing industrial business, not a story stock.

Now put it next to the rest of the trade. Bloom Energy, the purest fuel-cell bet on the same data-center-power story, trades at 142 times EBITDA. GE Vernova, the highest-quality name in the group, trades near 63 times EBITDA. You do not have to pay 142 times EBITDA for this story. Not anymore. Cummins trades at 27.8 times normalized after-tax operating profit, and it already has the contract the other two are still promising. Of every name we screened in the AI-power trade this month, this is the cheapest way in. It is not close.

Five scenarios, three years out

We do not do single-number predictions here. Every name gets five scenarios, because the future is a range, not a point.

| Scenario | 3-Year Price | Annualized Return | What Has to Happen | |---|---|---|---| | Super Bull | ~$1,498 | +36.6%/yr | Data-center power becomes its own disclosed, fast-growing segment. The market re-rates Cummins like an AI infrastructure name, not an industrial one. | | Bull | ~$1,052 | +21.5%/yr | The order book converts faster than guided. The multiple expands modestly as the story gets real. | | Base Case | ~$763 | +9.1%/yr | Current growth holds. The multiple stays roughly where it sits today. | | Bear | ~$501 | -5.2%/yr | The power story stalls out. The multiple compresses back toward a normal industrial range. | | Super Bear | ~$260 | -24.3%/yr | A truck-cycle downturn hits at the same time the power story fails to deliver. Both engines stall together. |

Weighted by how likely each one is, expected return lands around 10% a year. Below the 12% bar we need to call something a clean buy. Above every other name we found in this trade.

What breaks the thesis

Cummins still makes most of its money from trucks. The Class 8 truck cycle has turned before without much warning, and it will again. If that cycle rolls over while the data-center story is still small, this stock has real downside from here, not just disappointment.

What we are watching for next: whether Cummins starts breaking out power-systems and data-center revenue as its own reported line, backed by a multi-year order book. That is the moment this stops being "an engine company having a good quarter" and starts being a re-rating. Edges like this one go stale fast once the crowd catches on.

The verdict

This is not a buy signal today. Base case says the price is roughly fair, maybe a touch cheap. But of everything in the AI-power trade we have screened this month, this is the first name where the price has not already run past the story. Most of the crowd is still bidding up the names everyone already recognizes. This one is not on their list yet.

Worth watching. Worth owning if that segment ever earns its own line on the income statement.

The same discipline that found this in a stock screen finds the same edge in a sales pipeline: the cheapest real signal beats the loudest story, every time.

This is our own research process, shared for education. It is not financial advice and not a recommendation to buy or sell anything. Do your own work before you put money behind any of it.

Start your free 7-day trial: https://www.strategysprints.com Ready to accelerate now? Book a Discovery Call: https://calendly.com/strategysprint/discovery-call

Happy hunting. Simon & The Sprinters 🐬⚡️🐆