Good afternoon we are here to see all right so you know it's great to see you again. I just want to orient the audience a little bit because fuel cells can be a little bit wonky. So Bloom, Bloom makes fuel cells that run a fuel. So in Bloom's case usually it's natural gas. It runs across a stack of catalyst to chemically make the electricity and doesn't involve combustion. And Bloom has been developing and selling this technology for 15, 20 years. But all of a sudden you know it's a hot topic. And why are all the data center operators suddenly interested in Bloom and fuel cells? Tell us, K.
Good afternoon everyone and Katie and AIO, thank you for having me here. Look, the stage is about 500 square feet. Okay, if you look at typical cabinets of CPU servers, those are the servers that are in all your computers that you're used to. Okay, it'll roughly consume about 130 kilowatts of power. That 130 kilowatts is the equivalent power just for the stage worth of servers in a data center for 100 homes, 100 average US homes. The new GPU chip, Blackwell chip that's coming out, is 13 times more than that. So instead of 100 homes, this little space of data center will require the same amount of power as 1,300 US homes. And if you listen to, you know, Jensen from Nvidia speak, 3 years from now the Rubin plus chip that's six times that, so close to 7,000 homes worth of electric power is what it'll take to power just the computers in a room the size of the stage. And what do they want? Onsite power. That kind of power to be able to provide in your neighborhood is not going to be easy because power is usually produced far away in a large central power plant and it's brought by highways and surface streets called transmission and distribution to where you need to get it. And building that transmission distribution takes years. And we as a country, none of us want it in our backyard. So the amount of litigation that goes on, the amount of stuff that goes on before anything gets done, the chances of getting another 100,000 miles of transmission built in the next 6 years? Why 6 years? By then the world would have decided if America is going to lead in artificial intelligence or some other country that's not as democratic is going to lead. And that's not a choice. We have to win this. For us to win this, we need the power. For us to need the power, time to power becomes important. And you cannot, if you cannot depend on your transmission distribution, on your highways to be open, then you better be where you need to be instead of having to travel, which means the power station comes right next to where your data center is. So onsite power becomes very important. Bloom was founded 24 years ago with the notion that this moment will come. We built the company for this moment. Onsite power becomes extremely important. So what do we do? Like you said, we take natural gas or hydrogen in the future, any fuel, we are agnostic to it as long as it has the hydrogen. Without burning, without combustion, we convert that chemical energy in the fuel to electricity. No air pollution, no noise, so you can put it in a neighborhood. There's no smokestacks coming out of it. And it's highly reliable power. That's why data centers are interested in our solution. And the policy environment obviously has changed a lot in Washington around energy companies. How does this affect Bloom? So we have always believed that there are three things that are extremely important when it comes to electricity and energy: it needs to be affordable, it needs to be available and accessible, and it needs to be clean, because at the end of the day we need to breathe the air around where things happen. If you can provide all those three things reliably, there will be a market no matter what the policies are. Everybody needs electricity. Right now, electrical abundance in this country is a national security issue. It's not just an energy issue, it's not just a cost issue.
So to build your own power plant for the data center, these are companies that make 60 to 80% margins. They can afford to pay to build the power plant rather than making the ratepayer pay for it. That's another argument politically for the power plant to be paid for by the person who wants that much power in the small space. Yeah, and I know we talked about this before. The emergence of DeepSeek and more efficient AI was this kind of pivotal moment in terms of, you know, perception for the industry. Do you feel in your position of, you know, talking to data center companies, do you feel like the AI infrastructure buildout has gotten over its skis?
I would say we are going to look back very similar to the 2001.com bubble. The internet, do we use more of it or less of it? Same thing's going to happen with AI. I'm not going to talk to stock prices of certain companies, but if you just look at AI, we can't overstate how important it is going to be in our lives. So I'll give you a very simple example to demonstrate. Right, a million square foot warehouse, you know, taking care of your fulfillment needs of whatever you order online, it used to be about 1.5 megawatts, mostly manual forklifts going and taking out pallets, unpacking pallets, everything. The modern one within 2 years that they're building right now and talking to us for power, instead of 1.5 megawatts is going to be 6 megawatts. Three of those megawatts are for inference AI to do all this work more smartly, with greater accuracy and with lower cost, so you can enjoy the benefit. Every aspect of our life is going to be touched in some profound way in terms of productivity and gains by AI. So if anything, if DeepSeek-like things make training easier, inference is going to happen a lot faster, and inference is where 90% of the demand is going to be, both for the data centers as well as for the power. So if anything, it only speeds up the time to power issue and not slows it down. So the AI infrastructure buildout is on schedule and just going to keep growing. There's kind of no bubble. I think Jensen said that, right? The CEO of Nvidia, like last week, the only restriction to AI's growth is going to be power. Is that enough? It is about power.
And K, you've been in the clean tech sector for over two decades. You've seen lots of ups and downs. You know, where are we right now in this kind of phase of the climate tech kind of boom and cooling off period?
I think energy transition got off on the wrong foot. How so? We didn't do all of the above until we got the clean technologies to get to maturity. The way I describe it is we were flying an airplane at 30,000 feet called modern life that requires electricity. Single engine plane. And we were replacing the engine, new engine, and yanking out the old engine before the new engine got fully running. That's a bad idea. We need the new engine, but those things take time. So we need that transition to be smooth, and we need all of the above through that process. And it's about optimizing the amount of carbon dioxide in the atmosphere, which is what the climate issue is. Not a fight between fossil and renewable. It's about reducing the carbon footprint. So I think all of the above is where it's going to go going forward. And you know, transitions sometimes take some missteps before they go. But at the end of it all, this distributed world with abundance of energy is going to create a better world for our children and our grandchildren. I'm an optimist. I believe that there'll be a better future.
