Dan Dreyfus 在Sohn 2025推介 Mirion Technologies

Thank you. I'd like to begin with an observation. If you put them on truth serum, the deepest pockets in the world are all in panic mode: big tech, manufacturers, and governments. Looking out over the next 5 years or so, as they're making plans to build more semiconductor fabs, more gigafactories, more steel mills, and more data centers, most of them have no idea where they're going to get their next gigawatt of power from.

To throw fuel on the fire, autonomous-vehicle fleets and robotaxis are starting to grow exponentially, and they're all going to be electric. Then you've got commercial buildings swapping out their old gas boilers and replacing them with electric heat pumps. All of this, in aggregate, is going to consume incomprehensible amounts of electricity.

1. AI Accelerates Power Demand

I'm sure we all remember DeepSeek Monday back in January. Ever since then, there's been a debate on AI demand, both for compute and for power demand. I'm personally in the camp that that's going to grow for a really long time. But let's just say that's completely wrong—and it's not. If you ran a conservative stress test on power balances and completely removed AI demand from your demand forecasts, power markets would still be in deficit, because the bottom line is that the world is going electric.

We're going to be measuring human progress by how much electricity we consume. We run the risk of falling behind because we're going to be short on power. So nuclear, which has basically gone from zero to hero from a perception standpoint, is going to play a critical role in bridging this supply-demand gap.

2. Nuclear Becomes Strategic

As you can see on one of our slides, we have a long way to catch up to China, which has been building a state-of-the-art grid for the last 20 years, while here in the U.S. we've been basically doing nothing. But the good news is that's about to change. We got some big news this past Sunday that the Trump administration is considering an executive order to quadruple the size of America's nuclear fleet to 400 gigawatts by 2050.

Now, it's going to cost a lot of money, probably $3 trillion to $5 trillion, but strategically, this is mandatory for the country. We need a fleet of modern, clean baseload power if we want to stay competitive.

3. Mirion Makes Nuclear Safer

Now, let me be unequivocal: there's no nuclear power cycle without Mirion Technologies. In nuclear, there's one rule above all else: safety first, safety last, and safety always. Mirion dominates the nuclear radiation-detection space. Radiation detection is the backbone of safety for all nuclear power plants. It's heavily regulated, and it's mandatory.

After finally recovering from Three Mile Island and Fukushima and getting public perception to embrace nuclear again, the nuclear industry simply cannot afford to have another accident. So, again, rule number one in nuclear is safety, and rule number two is don't forget rule number one.

4. Recurring Revenue Drives Resilience

The 2 things I love most about Mirion's business model are, one, its multidecade growth profile, and two, its resiliency. 70% of the revenues are recurring and usually last for the life of the whole nuclear plant, which is around 80 to 100 years. So you have massive duration to your cash flows. These revenues are super predictable, nondiscretionary, and effectively recession-proof.

5. Three Engines Power Growth

To simplify, Mirion makes money in 3 ways. First, as existing nukes run harder, they get paid more. These nukes today are running as hard as they've ever run. Higher utilization means more demand.

Second, they're cashing in on plants getting life extensions and restarts. Every reactor that's getting a second lease on life needs Mirion. We all saw Microsoft firing up Three Mile Island again. 5 years ago, if you told me that Microsoft would be doing a deal to reopen Three Mile Island, I would have said, "Not a chance." But here we are: Microsoft Azure, brought to you by Three Mile Island.

Finally, a big nuclear new-build cycle is coming to a theater near you. New reactors are going up, and Mirion's going in. Based on the backlog of new builds, this creates a powerful earnings-growth algorithm that should allow them to grow for decades.

6. Mirion Targets Decades Of Growth

So, quick history. Mirion went public in 2021. It came public through a Goldman SPAC. Goldman did 2 SPACs, which actually came out of the group that I used to work for back in the day. The first SPAC was Vertiv. That's been a 9-bagger. The other SPAC is Mirion. I'll tell you, I think over time, Mirion's got what it takes to give Vertiv a run for its money.

In conclusion, if you do the work, you'll probably come to a similar answer as we did. Mirion should do around a buck and a quarter of earnings in 2029, which from there should really continue to compound at potentially an accelerated rate into the 2030s, because that's when the new-build cycle really starts to heat up.

Industrial growth companies with high barriers to entry and big growth trade at 25 to 30 times earnings. At the low end of that range, putting a 25 multiple on it puts Mirion in the low 30s by the end of 2028, up from about 17 today. More importantly, this is a compounding engine that's just going to keep going into the 2030s and the 2040s. So it's one of these buy-and-never-sell investments.

Thank you very much.

Thanks, Dan. I think most of us would agree in this room that nuclear has certainly caught a bid, if you see the stocks, the uranium stocks as well. The question is: You talked about Trump and the support for the nuclear program, but as we know, the political winds can shift very quickly, and let's say they do.

I remember when Joe Biden came into office, the first thing he did on the first day in office was completely kill the Keystone pipeline, which was a multibillion-dollar, multistate program. How do you quantify, mitigate, and think about that potential risk in the next year and a half if Congress changes power, or in the next 3.5 years if the White House changes power?

7. Nuclear Answers The AI Power Crunch

So, a typical Google search: If you take a regular 100-watt light bulb, that would be like turning that light bulb on for 3 seconds. Then you do an AI search, and that would be consuming the amount of power of keeping that light bulb on for 2 to 3 minutes. Then you do an agentic conversation with AI. That takes up the amount of power of keeping that light bulb on for 8 to 9 minutes.

The only way we're going to scale AI, semiconductor fabs, life sciences facilities, and gigafactories without emitting an obscene amount of greenhouse gases into the atmosphere is through nuclear. There's no other way. The rest of the solutions are basically semifraudulent if they're telling you that they can do it.

I thought it was encouraging that even the Sierra Club changed its tune after 50 years on nuclear. In 2024, it came out and said nuclear is great. So I think there's been a structural and permanent perception shift where both sides of the aisle are going to be leaning into nuclear in a big way.

Thank you, Dan.