Kathryn Huff
Appearances
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
They grow and grow every time you look at the news. And so we are in a position where the existing clean energy infrastructure has to expand to support those data centers, especially data centers that need 24-7, reliable, always on power. If you own a multi-billion dollar data center, you don't want it to be running at 2% capacity because the winds stop blowing.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
You need 100% power 24-7 regardless of weather.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, you're absolutely right. Transmission, especially building new high voltage power lines to move gigawatts of power from a generator to a consumer is expensive. It cuts through land that usually needs permits. Sometimes it can be very slow. So co-locating data centers with smaller modular reactor builds has an advantage.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
The more modularity in these builds is certainly also supposed to contribute to the speed and reliability with which we can deploy them. The idea being that it takes a really long time to build a gigawatt scale nuclear power plant. But if you build a 30%, you know, sized 300 megawatt plant, reactor, then maybe you can build a few more of them, get some lessons learned.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
They maybe move a little faster. You might be paying slightly more per kilowatt hour, but you should be able to deploy them quicker and learn faster, thereby coming down the cost curve of those construction learnings. That's a whole future there that has not yet been realized.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, I think we have technologies at all sizes in the advanced reactor space. The most recent two builds in the United States were the Vogel Unit 3 and 4. Those are AP-1000s from Westinghouse. They're big, like conventional nuclear reactors, but they incorporate some passive safety. And so they're kind of a Generation 3-plus design at the sort of big gigawatt scale.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
They were built pretty over schedule and over budget. And so there's a lot of trend towards shrinking that kind of design, a light water reactor that's just smaller. You've got designs from Holtec and GE Hitachi. Westinghouse has a shrunken version of the AP1000 called the AP300. Yeah, you've got NuScale is one of those small modular light water reactor designs. But then in addition to that...
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
You can also build small modular advanced reactor designs. And there are a number of companies pursuing commercialization of that with two deployments already happening with the support of the Department of Energy. These would involve more advanced coolants and fuels like sodium, which is a liquid metal. Is that the molten salt? Yeah.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So sodium has a couple of cool features as a coolant. It's highly conductive. And so it's extremely performant to move the heat from the fuel to where you need it in the turbine through an exchange process eventually. But that conduction is really useful, but it's also very helpful for the neutronics control of that criticality that I talked about earlier.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, so it keeps the neutrons fast by not moderating them. If you combine that with a metal fuel, sodium-cooled fast reactors have an opportunity to be quite passively safe using reactor physics and negative feedbacks from the expansion of the fuel and the expansion of the coolant and sort of neutronic behavior of the coolant to drive power down if power goes high, keeping things balanced.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So, critically, sodium in this case is a liquid metal. When it's combined with something else, it becomes a salt. There are salt reactors. We can talk about that. But, yeah, sodium itself is also somewhat corrosive. You can't see through it because it's liquid metal, and it is pyrophoric. So, when it gets wet, it tends to burst into flames.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yes. It is almost exactly like a coal plant. You're just boiling the water in a different way.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So, star on Earth would be pretty hard with fission. We await our fusion colleagues, for that matter. But the fission reactors that are being deployed with some of these advanced coolants will get much, much hotter safely than conventional reactors. We're talking...
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
800 degrees Celsius, really high temperatures, which is quite a bit more than the 300 degrees C we would usually see from conventional light water reactors.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, absolutely. And if you're going to use that heat directly, then it's very useful for industrial applications like, you know, reducing steel, reducing iron for the steelmaking process.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Right, and rather than burning natural gas or coal to make direct heat, the first application of the X-Energy high-temperature gas reactor will be at a Dow chemical plant where they'll use both the direct heat and the electricity that they convert. Some of that heat straight from the reactor won't be converted at all into electricity.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Not exactly the way we convert heat into electricity, but it is certainly the case. It is the case that high temperature heat is higher quality because it is easier to convert. You have a little bit less loss. Yeah, very good.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, one would hope, right? And then you can use that energy for other things that will advance human prosperity, right? We can also use it to displace the kinds of fossil energy that we still rely on and contributes to the climate crisis.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It remains to be seen precisely how much each design will cost. But we're looking at a scenario, I would refer you to the liftoff report that Dewey put out, but it estimates that new nuclear power, you know, is going to be in the order between $120 per megawatt hour for the first of a kind, all the way down to nth of a kind, maybe in the $60 per megawatt hour range.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That's a big range, but it's very similar to the kinds of ranges we're seeing when we look at, you know, renewables plus grid-scale storage, which is the only comparable reliable 24-7 clean energy that involves renewables, right? Alternatives would be natural gas with carbon capture. Cheaper, certainly, maybe between a 100 at the high end and $60 per megawatt hour at the low end.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Geothermal has a very, very big range, but very similar to nuclear. You know, 130 at the high end, 57 at the low end. Hydropower is always cheap.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It's a great question. DOE has estimated that if we want to hit net zero by 2050...
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah. We will have to build at least 550 to 770 new gigawatts of firm, clean power. Some of that will be hydro or, you know, geothermal and things like that, battery storage. But at least about 200 gigawatts of that will need to be nuclear. And we're not the only ones that made that calculation. Across the world, dozens of countries have committed to tripling nuclear energy in their countries.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
I had to get that in. Absolutely, yeah. Around the world, dozens of countries have committed to tripling nuclear power.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Some of them could. You know, I mentioned sodium-cooled fast reactors earlier, which have metal fuel. And while TerraPower isn't currently planning to recycle in the United States, it is an amenable technology for the kinds of recycling that other nations do.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
France, for example, recycles a great deal of their spent nuclear fuel, resulting in lower volumes, lower masses, and much shorter lifetimes of recycling. long-term radioactivity by putting the longest lived and most useful isotopes back in the reactor. We could do that, but in the United States, we don't currently have the infrastructure to do that.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So it would take a real government effort to move forward on recycling. But when I was a DOE, this was definitely something that we were continuing to do research on. And there was a great deal of interest from the commercial side in seeing recycling be back on the table in terms of options. Molten salt reactors also were mentioned earlier.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
I should note, in the SMR-MSR universe, molten salt reactors are also very amenable to recycling.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, we are still mining uranium. Some of the best uranium in the world comes from mines in Canada, but it exists in a lot of places, including the United States, Australia, Kazakhstan. How about Greenland? Oh gosh, too soon. So yeah, I think the reality is Unless you do a great deal of recycling, you're going to continue to mine uranium.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So, yeah, recycling would reduce our need for new, fresh uranium.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
The biggest bottleneck for that uranium fuel cycle is that, you know, the mining of uranium, there's lots of sources of it, but then the process at conversion and then enrichment of that fuel, where you increase the number of isotopes of uranium-235 per kilogram of total uranium, that enrichment process and the fuel fabrication process all happen at a much smaller number of facilities internationally.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And so, yeah, you know, Quite to your point, international collaboration has been necessary to ensure that, you know, if Russia, who dominated historically conversion and enrichment capabilities in the last 20 years or so, if they decided not to sell to the United States, we needed to have more capabilities in the U.S. and among our allies, France, the U.K., etc. And so that has been underway.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
In fact, right behind me is the law where we banned Russian uranium from Russia, you know, imports in the near term so that we could protects some of our ability to invest in new enrichment capability.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So there's two different ways, but basically the spent fuel that you start with is a mixture of uranium atoms and split fission products. So the two parts that the uranium atom splits into, this might be iodine and technetium, practically, you know, half of the isotopes in the periodic table or in the chart of the nuclides.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That mixture needs to be separated so that those fission products are removed from the total. the already split atoms.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It is. It's quite a bit more complicated and it's extremely useful for nuclear engineers because we end up producing a lot of those isotopes inside reactors during the fission process.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That's an interesting idea. You know, some of them fall into chemistry groups that can be useful for industrial applications and things like that.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So one of the interesting possibilities for recycling, which I'll get to how you do it in a second, but one of the interesting possibilities for recycling is that some of those products that aren't useful in the reactor and are otherwise waste, some of those products are hard to generate otherwise, but can be useful for, you know, medical reasons, imaging, other kinds of radioactive materials.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Absolutely. Technetium-99 metastable is routinely used for things like thyroid imaging and whatnot.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That's right. A very specialized, high-precision radioactive dumpster.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
No. But generally speaking, what you end up doing is you take out those fission products and you still have a great deal of enriched material left. And during the fission process, you have been breeding a little bit of plutonium.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Some of the uranium-238 atoms have absorbed a neutron and then another neutron, and they move their way up the chart of the nuclides into plutonium-239, which is also fissile. So you can put both the uranium and plutonium back into the reactor usefully. along with some of the other transuranic elements, and that's recycling. You can do it with aqueous chemistry or electrochemistry.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So first of all, the volume is pretty small to start with. But yeah, there's been a lot of work in fuel utilization so that the amount of fuel that you put in is used to the maximum extent practical. A lot of that has resulted in designs that will leverage a higher initial enrichment of uranium. You'll see this in a lot of small modular reactor designs because it allows the reactor to be smaller.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And in a lot of cases, it'll allow for a higher fuel efficiency or fuel utilization. What kind of radioactive waste do you get out of fast reactors? It's a metal, uranium, plutonium fission product object, right? And it's never sort of, when it sort of comes out of a reactor, it's going to be in the same form it is.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
You know, you get a lot of Simpsons kind of images where it looks like a glowing green goo. Spent nuclear fuel is actually a Uranium oxide coming out of conventional reactors, so it's a type of ceramic. It's quite heavy, and it's in little pellets, but it's a solid and quite dull. In the case of a sodium-cooled fast reactor, it would be a solid metal cylinder.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
In the case of a molten salt reactor, where the uranium is dissolved into a liquid salt in the reactor, these don't really exist commercially yet, but they've been proposed. That would be a liquid, but that would be the only kind of reactor that has sort of a liquid waste reactor.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, so it's a rock ore when it comes out. And then when you put it back in, it's a solid of uranium oxide typically. And it's contained in a canister that's steel and then concrete and layers upon layers of shielding.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, so it's fairly straightforward to maintain a fission reaction at a stable, steady state at this point with our development of technology. But there are feedbacks that could drive that reaction into super criticality. And you have to spend a lot of time balancing the reaction in the reactor. Whereas a fusion reaction, its tendency is to end. This is why it's so hard to make a fusion reactor.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It's a very hard reaction to maintain. Because it's hard to maintain and because there's not this sort of opportunity for easy supercriticality, then you don't end up in this situation where you might become overpowered, create a lot of extra heat you weren't ready for, melt things down, etc.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And so fundamentally, since the stable state of a fusion reaction is to not be happening, it's a little bit easier to keep them safe.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So in fusion, there are activated byproducts, but not the kind of high-level fission products that are very radioactive for a very long time that we'll have to manage as a spent nuclear fuel in fission.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That's right. And multiple nations are making real progress on that. Finland is just about to start operating their final repository. Canada has just selected a site for their final repository. Sweden has gotten permission to begin constructing their final repository.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
If I have time to bring that person to a nuclear power plant, I would and show them the containment structure. But if I don't, then I would show them a video of Sandia experimenting with the structures that protect us from nuclear power plant accidents. So Sandia, is it FFRDC? Yeah, Sandia National Laboratory. In New Mexico? Are they in New Mexico, I think? Yeah, that's correct, in Albuquerque.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
There's some experiments to show that you could ram a jet engine right into the side of a containment building, a sort of standard steel and concrete structure, the dome over the reactor, and it survives just fine.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Nuclear waste and Chick-fil-A bring you— Power plants like that do produce a lot of jobs and tax revenues that communities love. Maybe they use them to build the Chick-fil-A.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
We're seeing it already. We've seen requests for proposals and power purchase agreements signed to ensure that new reactors are coming online. Amazon has invested in Xenergy. Microsoft has invested in restarting the Three Mile Island unit. We have... very clear demonstration that the money is already going towards those new deployments and restarts.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And I think it's helpful to have those deep pockets working towards reducing the risk for end-of-the-kind builds. But yeah, it's hopefully going to result in more power even than those tech companies are going to need because that's what we need for our country.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That probably will be overkill for especially the regulatory and nuclear nonproliferation community. You want to make sure that certain nuclear material is kept extremely extremely well controlled and observed, counted, you know, tracked. Because there's a universe in which if you could put it in a suitcase, a briefcase, then there's a potential to make a dirty bomb out of some of the spent fuel.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, the first couple are in the very early stages of construction, and we're expecting them to be completed. Some of them hope for a five-year construction timeline, but it'll probably be more like 10 years. And then if we don't have orders on the books for more and more of them in the next year or two, Then you have to wait another 10 years for the next one.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And so what we're seeing from data centers and other kinds of companies, utilities like Dominion thinking about small modular reactors, then you see those orders. That's an opportunity to have those first deployments five, 10 years from now that are connected to the grid, followed by the next deployment.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So I run a research group that writes software for modeling and simulating advanced reactors and their fuel cycles. a lot of what we've been talking about today. I write multi-physics software with my graduate students.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, it's when you combine different physics, especially when you're combining physics at different scales. In my case, it's neutronics on the very small sort of angstroms and 10 to the negative 14 seconds kind of time scales with thermohydraulics, which is more like seconds and meters.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
If those two things affect each other, which they do when you're talking about reactor feedbacks and reactor accidents...
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Multiscale multiphysics, if you want to sound highfalutin.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That's right. That's right. So I studied advanced reactors like molten salt reactors and high-temperature gas reactors, sodium-cooled fast reactors, and... Recycling strategies. This is something really important.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
I do. I absolutely do. It's my favorite thing to do in the whole wide world. I'm sure you understand.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Only if I get to go to the party with the reactor in the basement.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah. So in fission, which is conventional nuclear power here in the United States and around the world, you separate, you break apart a heavy atom like uranium-235 or plutonium-239, whereas fusion gains energy from the binding reaction between two light particles fusing. And these would be isotopes of helium or hydrogen, typically. And so you're talking about completely different ends of the
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
periodic table, the forces that hold a nucleus together, the binding energy of that nucleus, you can achieve a net increase, a net sort of output in energy by coming down the isotopic curve by splitting a big atom or by coming up it by fusing two light atoms. And that's because of the shape of the binding energy curve
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, absolutely. I think a lot of it's around numerics, right? Talking to people about the metrics that matter. If you're worried about safety, the metric that matters is deaths per terawatt hour generated.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, it's only interesting as a metric if you're using it to compare energy sources. And so a number of different data sets have informed a number of deaths per terawatt hour from solar, wind, coal, et cetera. And nuclear is way, way, way down at the bottom, near, you know, slightly below solar and wind, actually.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Even if you include Fukushima and Chernobyl, interestingly, no deaths at Three Mile Island. And you have... You know, geothermal, I think, is one of the only ones that is lower than nuclear in terms of deaths per terawatt hour. And it's construction projects. There's serious issues maintaining wind turbines, like putting things on roofs and installing.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
That sort of thing doesn't happen on a nuclear construction site because of the incredible amount of regulation. And so you really are left with, like, accidents. And so there's plenty of lifecycle analyses of deaths, but also, you know, carbon per terawatt hour. And the reality is nuclear... in terms of generation, is higher, and so the denominator is huge.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So interestingly, renewables like wind and solar and geothermal and nuclear, they're all in the same category, which is just magnitudes and magnitudes lower than fossils. And, you know, nuclear ends up winning because historically nuclear has been the largest source of carbon-free, emissions-free power, right?
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And it's those emissions from the fossils and, you know, that really takes them to a completely different order of magnitude in terms of deaths per terawatt hour. And so I would say that renewables are safe. Nuclear is safe. Geothermal is safe.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It's really the fossils you need to be worried about when you're talking about safety, because they impact human health in a really demonstrable, clear way.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It's a great question, right? The WHO has estimated all kinds of things for full, total cost and death of the fossil industry and separates it into particulate-related premature deaths and whatnot. I don't know the exact number for people dying mining coal, but it's not zero.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
It's pretty high, right? Like you would lose people in collapses.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And Paul's right that, you know, it hasn't been a huge impact on humans' day-to-day lives because it's an incredibly small volume of spent fuel from fission reactors. We have 60 years of spent nuclear fuel. You know, it has been producing close to 20% of our electric power in the United States.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
And it's produced enough material to, if you sort of ignore the packaging and focus on the uranium, fill a football field. Not very tall, you know, a A few meters high.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
I'm a fission girl. I'm a sort of classic nuclear energy fission person. I have great optimism for fusion, but it will take quite a long time. And interestingly, you probably know this, that NIF test is an inertial confinement fusion experiment.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
A lot of the commercial proposals that are attempting to commercialize fusion in the near term are more like the ITER device in France that is magnetic confinement fusion experiment. It's a slightly different approach and seeing real breakthroughs in the ITER device may be quite a few years from now.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So, you know, we await some breakthroughs in things like the first wall protection, things like that. But for me, I'm here about kind of conventional nuclear energy and advanced fission energy that's sort of on the horizon right now.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
So there are no operating fusion reactors. There, like, could be someday.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, I'm not a fusion scientist. I, like, work in a nuclear class.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Every light switch has a little don't forget to turn me off, like, sticker.
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Yeah, it's totally data centers and things like that, but it's also increased electrification like for EVs, but all kinds of other things as well. And an increased sort of revival of industrial manufacturing that requires not just electricity, but heat. We see carbon-free steel companies wanting to start up in the United States and they don't have enough heat. Where do you get heat?
StarTalk Radio
Solving AI’s Energy Problem with Kathryn Huff
Usually you burn fossil fuels. Very few clean energy sources are available to provide that direct heat. Nuclear's there for it. So yeah, we have a huge amount of demand. Are we even factoring it in? The projections for how much new gigawatt worth, many tens of gigawatts worth of capacity are going to be needed to support data centers in the coming decades to support these kinds of endeavors.