Is Rolls-Royce a Secret AI and Crypto Play?

Everyone’s talking about chips, data centers, and the cloud.

But there’s a bigger problem brewing underneath: Where will all the power come from?

Every time you ask ChatGPT to summarize a meeting, or Grok what a post on X means, you’re triggering what’s known as an AI inference. This is the moment when a trained model runs calculations to deliver an output—text, images, code, you name it. Multiply that by billions of prompts a day across consumer and enterprise use cases, and the power demand starts to spike.

Inference workloads like these are now one of the fastest-growing drivers of electricity use in the digital economy.

In 2023, AI workloads consumed about 4.3 gigawatts of power—roughly 8% of total global data center demand—and inference accounted for 80% of that. Industry analysts project AI energy usage could grow 26% to 36% annually and may surpass the entire national power consumption of countries like Iceland by 2028.

Crypto mining is adding to power demand, too.

Bitcoin mining now consumes between 120 and 175 terawatt-hours of electricity per year—similar to the annual usage of countries like the Netherlands or Poland. After Ethereum moved to Proof-of-Stake in 2022, miners shifted to other proof-of-work networks like Ethereum Classic and Kaspa. These rigs run continuously and draw heavy power, often from the same regions where AI data centers are expanding. In states like Texas, crypto operations have become a key factor in rising grid demand and have been linked to multiple grid-balancing events over the last several years.

U.S. data center electricity demand is expected to grow sharply—conservative estimates suggest it could double by 2028, while more aggressive forecasts point to nearly triple the current levels.

The chart below from McKinsey & Company visualizes this trajectory, showing total data center power consumption rising from about 12 gigawatts today to more than 30 gigawatts by 2030.

Just last year, 60 data centers in Virginia’s “Data Center Alley” abruptly dropped off the grid after a voltage irregularity, switching to backup generators. That one move created a surge of excess electricity so severe that PJM and Dominion Energy had to cut power plant output in real time to prevent a regional blackout. Federal regulators now say incidents like this are becoming more common—and that large-scale disconnections by AI and crypto operations are creating new systemic risks the grid was never designed to handle.

Rolls-Royce Holdings plc (RYCEY) has a solution that doesn’t rely on sun, wind, or good weather.

Its Small Modular Reactor (SMR) program is building factory-assembled nuclear plants designed to go live faster and with lower up-front costs than traditional reactors. Each unit delivers 470 megawatts of clean, always-on baseload power—enough for over a million homes, or the kind of clustered computing zones Big Tech is deploying at hyperscale.

In 2025, progress is accelerating:

• In February, Siemens Energy partnered with Rolls-Royce to supply the turbines and generators for its SMRs. This gives Rolls-Royce a complete nuclear power package ready for global deployment.

• The Czech utility ČEZ acquired a 20% stake in Rolls-Royce SMR. Its goal: deploy 3 gigawatts of capacity across Europe. That’s about the same output as three full-scale nuclear plants—but delivered in modular chunks, not 10-year megaprojects.

• Rolls-Royce is one of four finalists in the UK’s SMR competition and has also been shortlisted in Sweden. Selection decisions are expected later this year.

This isn’t Rolls-Royce Motor Cars. That company is owned by BMW. This is Rolls-Royce Holdings plc, the engineering firm behind nuclear submarine reactors and military-grade propulsion systems. It trades primarily on the London Stock Exchange under ticker RR.L, and U.S. investors can access shares through the over-the-counter (OTC) market under ticker RYCEY.

If AI and crypto are going to reshape the economy, Rolls-Royce is one of the few companies building the reactors to keep it powered.