AI did not create the data-center power problem, but it made the problem impossible to ignore. Danielle Fong has been building toward advanced high-density power and the market is finally running toward her.
In part three, we get into customers, constraints, roadmap choices, and the personal decision to leave a whole class of academic work on the table.
Transcript
Pablos: This thing presumably you guys are gonna build it. I guess first market’s gonna be data centers.
Danielle: Data center is small and large.
Pablos: So that’ll be the megawatt scale, 10 megawatt and beyond. And what they’ll need is a source of something that burns, whatever it is.
Danielle: That’s right. Which can include if you’ve got a lot of land you get solar as cheap as you can get it. You get the cheap form of the hydrogen electrolysis. That you can just clip in and you can put as much power into that as you want. And even though it’s a bit less efficient, very inexpensive on both the solar and the hydrogen side. And then you end up with an excess of hydrogen as a buffer. And it’s like the first buffer that you want to use is your batteries. And the next is you want to use your hydrogen. And then after that is your fuel, which you could produce methane, like in a chamber for storage.
Pablos: A data center wants this because they need power anyway. They need electrons.
Danielle: If what they’ve got is LNG or something else as a fuel supply, this is an easy way for them to get to electrons that’s baseload.
Pablos: Because even though you said you want to solve energy storage
Danielle: They want dispatchable power that can actually follow with a fair amount of capacitance, actually.
And they actually want it DC. They don’t necessarily want it to be AC.
Pablos: AC is a stopgap. We converted all to DC at the end anyway.
Danielle: People are moving to 800 volts DC.
Pablos: I think data center of the future is going to be all DC.
Danielle: Which is what the solar cells are putting out.
Pablos: How would this play out? I guess in that case, you’ll build something like a shipping container scale thing that has gas going on one end and electricity coming out the other end. And then what’s the, what’s it look like? What other applications does this make sense for? Does it make sense to make planes fly off of this?
Danielle: So I think that there’s some interesting, quite a lot of interesting propulsion. From eVTOL at smaller scale to, it’s basically you have an electric turbo fan, you operate this, and you can have a lot of them along the wing. Potentially MHD type propulsion potentially space
Pablos: Magneto hydrodynamic.
But what let’s start with eVTOL simple things so right now I load it up with batteries I get to fly for nine minutes and then I have to land and charge.
What you’re saying is we would basically get the same like range calculation as a Cessna or something in an EV tall yeah because I could just fill it up
Danielle: Vastly more energy density
Pablos: Now we can use electric motors for propulsion if we want we don’t need to deal with the maintenance and the manufacturing difficulty of jets and those things
Danielle: They make the best wildfire protection drones. They carry 100 pounds of water and they squirt it. Those guys would love.
Pablos: 1,000 pounds of water.
Danielle: And they use batteries, and that sets the range. There’s slightly more water than there is battery, and that means that the range is shorter. If you try to solve the problem with engines, which people do, the amount of bespoke wacky engineering associated with, you now have this vibrating loud thing of a certain size, also heavy and it’s hot. What you want is almost as good as batteries in terms of like usability. But higher energy density and comparable power density. You still want to probably rely on power batteries for your sort of capacitance short-term stuff. But it doesn’t have to be an hour. A little bit of chemical battery if you need it
Pablos: Also don’t have the charge cycle.
Danielle: Andre Carpathy talks about Waymo and that when you’re engineering to like five nine six nines a really high level of reliability.
Every nine takes about the same amount of work. And so we think that basically because we can use any fuel this, once we get it really great, means that it’s well-suited to being all the way from that deep backup all the way to base power instead of just having to stack batteries until they’re too expensive or too heavy.
Pablos: Okay, so hopefully people have a sense of what you’re trying to build. What’s the hard part here? Why isn’t it done? Is it really going to work? Is this a fringe science project where are we at with this thing have you cracked it like is it working
Danielle: I think we’re hot on the trail we think we’ve cracked a bunch of the hard problems but they’re still probably like many hard problems will reward cleverness in order to actually get to hey we have a product that’s being made in large quantities
Pablos: Hard problems on what scale? Problems to make it better cheaper smaller.
Danielle: It’s interesting so why haven’t other people i ask myself every single day why hasn’t people thought about this
I spoke with richard swanson founder of sunpower he’s one of the original sort of solar greats. What I did not know is that he started out with photovoltaics and using a tungsten emitter, He achieved 26% efficiency at 500 suns in the 90s. That’s much better than I need. Good. Anyway, I asked him, Hey, why didn’t anybody think of this? And he said that I did think about using sodium as an atomic emitter into the flame. People knew about that. But A, the challenge of recovering the sodium was left as an exercise to the student, commonly done. And B, he thought that the brightness of a flame that was seeded with something like this would be capped by the standard Boltzmann emission spectrum, the kinetic temperature of that flame.
And since it was so narrow band, it wouldn’t be very bright. And he was doing broadband. And that’s what the typical theory would expect this. And this is not true. There’s a lot of interesting physics here. But that’s one part. And then the other part is to get to really high temperatures so that you have these really good brightnesses and so you can start to push these efficiencies into the really sort of unique physics regimes. Nobody would recuperate up past 800 degrees, basically the temperature, the metals could tolerate. Even at 800, you’re starting to need all this high nickel alloy and stuff. People had not solved using ceramics. There are partial answers. But people hadn’t deftly solved it.
Pablos: You guys are well into that now.
Danielle: We have. We think we’re the best in the world. We think we really, really know how to get this. Of course, now we want to get yield rates into the 90s. But already it’s like things are working in our favor. And basically the trick is because the ceramics are so hot and so hard and because the thermal expansion and contraction across the thermal gradients and the cycles means that you have this thermal expansion plus the hardness means that there’s a huge amount of stress. But you can kind of invert it and if you have just a small amount of relaxation, a small amount of strain relief, that relieves a huge amount of stress.
The problem is where you have sharp corners instead of curves. You basically make a unit cell system, a whole system out of structures that are curved everywhere. All curves. Quite a lot like nature.
Pablos: And so that, I’ve observed that around here. Maybe we’ll take some pictures. But the thing, there’s no part of this that you haven’t been able to demonstrate at some scale. There’s no part of this that you can’t do. Now you have the engineering job of putting it together and it’s something that can be productized, but we know it’s going to work. And we don’t really know for sure what the maximum efficiency you can get is, but we know it’s going to be good.
You’re off to the races here. And what’s the plan for the sodium? Is the sodium liquid?
Danielle: We use sodium chloride, so regular table salt. And when it heats up, at first it becomes liquid and like a gas, like any other material, like water. You can think of it as it’s like boiling. But the way that it transports itself in the system is that it wicks itself.
There’s all of these labyrinthine passages in the heat exchanger, and that’s where it deposits. And if it’s hot, then it forms a surface of liquid that’s actually transparent. And it moves by surface tension, it’s moved by gravity, and it’s moved towards the area that’s hottest, which is where it’s evaporating. This allows us to passively drive the transport of the salt throughout the system. And this is what was left to the students by previous evidence. And this was not our initial idea. It was sort of like we noticed that this was happening in our experimentation and decided, hey, it seems to be wicking. Can we design an entire system around this where in the heat exchanger you’re actually capturing the energy, the latent heat of the salt as it condenses and as it evaporates? This sort of started because, just because of the amount of experimentation that we would do, but also experimentation around, we were trying to make lightsabers for Halloween and things like this, real lightsabers. And the most durable materials were these ceramic that even when you didn’t directly salt them, they were like emitting sodium. And I’m like, how is this working? And so it turns out the molten salt wets alumina and goes into all the pores and actually makes it stronger. It densifies it. It seals it. It transports through the system. The whole thing works sort of like a sponge, like waterboarding.
Pablos: When are the lightsabers going to go on?
Danielle: Not sure exactly how to market that. As soon as I have one that’s really good, I’ll definitely want to show it off. But do I want to put this in the hands of people? Maybe.
Pablos: Boring company.
Danielle: I think I figured out probably the critical thing to make quite a long blade. But then what everyone wants to know is, hey, does it make the whooshing sound and can you duel with it? And I’m like, no, not yet. We don’t know how to make that work.
Pablos: That sounds a little PG-13, but okay.
Danielle: Exactly. It’s a side quest.
Pablos: Well, in a previous life, I helped start a sword fighting school.
Danielle: Really?
Pablos: And so I have a lot to bring to that particular problem. When the time comes, let me know.
Danielle: Oh, good. Here are the elements. You can stabilize a flame by a laminar flow element, and you can stabilize a flame by swirling it. With our 3D printing, you can make a nozzle that does both. It swirls and is a laminar flow element. This is extremely stable, and it allows you to put energy into the blade. That keeps it stable. In addition, they drive currents, and with the salt seeding, it’s a plasma. You can actually have a significant amount of flux repulsion.
Pablos: Wow, really?
Danielle: So I don’t know how far this can be done, but it might be possible to do it quite far. But you need a significant energy source. You won’t be able to block bullets and you won’t be able to block blaster shots that don’t work.
Pablos: But you could another blade that also is made the same way.
Danielle: What’s the deal here?
Pablos: Oh my god
Danielle: It’s because you’re obsessing about this since the lightsaber is so primal
Pablos: Just that’s what made Star Wars so successful is that any kid can just immediately internalize what it means to beat someone over the head with a stick and cooler if it’s like hot and bright
Danielle: And able to slice through steel doors.
Pablos: I’ll slice through some steel doors. That would be rad.
At some point you had to decide, I’m going to leave academia
Danielle: You can kind of come back
Pablos: Well, I guess you could come back.
Danielle: Basically, I just reach out to the professors. I just work with them directly. Which I would recommend, by the way.
Pablos: That is amazing insight. I’ve hired a bunch of PhDs
And it takes a year to deprogram them from the cult indoctrination of academia.
Danielle: Go to the top professors who are over the status thing.
And it works. You can still go work with them.
Pablos: A lot of people are indoctrinated into that, and I think that’s what keeps them from
Danielle: But they know something is wrong. A lot of things have gotten too much into the status game, so that means that there’s not really communication going on.
Then you meet with people who really know their stuff. They make it understandable. This is a big signal. Don’t go to the second players. Go to the real players. And then things are a lot simpler.
When we were tapping into this, pick something technologically that has a huge amount of headroom. I say, energy density. I’ve got 70x over the competition. Pick something I can do with a minimum amount of money so that I can get past this necessity of these existential financing rounds—heavily dilutive. And then do it as visibly as possible and make it as tangible as possible. We could work for one group or another group, but really figure out who needs it the most. Which is two things in the Venn diagram. They need it the most and they have the most money.
Pablos: What lessons have you learned?
Danielle: Find people who are not only the richest in the world, but they are desperate for what you are providing.
Pablos: Very good advice.
Danielle: That basically became the AI people.
Pablos: For you, AI is certainly the demand driver it’s the demand driver forever.
People get off the rails because they think this is making rich people richer. But you gotta subsidize the development of every new technology with rich people
I think our lot in life as Americans we get basically born rich
And your job is to waste as much money as possible in your lifetime on cool gadgets and shit
Trying to figure out what works so that we can get economies of scale up get the cost down and then make those things possible for the rest of the world where they didn’t get born rich. And I think it’s a reasonable way to think about
Danielle: No, I think so, yeah.
Pablos: About inequality and about what’s happening in the world.
Danielle: I look at it somewhat differently. There’s the capitalist, noblesse oblige, great. That motivates some. I kind of just look at it as like the previous system that one would hope would be sufficient to be a fertile ecosystem in which energy innovations, which are necessary, could emerge and sustain themselves. Simply was not true for a wide variety of things. Everyone agreed would be great, but the system was fucky in some way. And as a result of compounding unseriousness in timelines and scale and consistency of funding and focus, essentially has a 0 % success rate.
Whereas if you have the most highly capitalized companies in the world, desperately needing power infrastructure, you can build on top of that. You can say, hey, you need something. We need something. Suddenly this flips everything on its head because now there are private companies. They’re very well capitalized. They need as much power as possible in one place. But they might not need all of the power and all of the backup that they need. They might still want to be connected to the grid. These are positive sum exchanges that can easily be made.
Pablos: I think it’s the hyperscalers that are going to save us.
Danielle: We couldn’t build these things before when the oil industry was the biggest industry on earth.
No, they were too conflicted.
Pablos: I think Congress was staffed by Shell and Chevron.
Danielle: People tried. We had investments from Total Energy and they were super excited. Then Christophe de Marjorie died in Moscow, and the guy who was head of New Energy lost out to the guy who shut down refining capability in Europe. I’m like, oh no. And then the price of oil plummeted. It was like, all right, oil majors are not going to be the solution here.
But hyperscalers, neoclouds, this is actually incredibly fertile and better organizing, more efficient than a bunch of different government programs where it’s not really clear what the outcome is intended to be.
Pablos: Super cool.
Links
- Lightcell: Danielle Fong’s company turning heat into light and power
- Danielle Fong: Danielle’s writing and projects
- Danielle Fong on X: Danielle’s X account