Transcription
It's been a big year for Bloom Energy, a fuel cell manufacturer who has been winning large data center deals. Our guest today on data center richness is Jeff Barber, the vice president of data center sales for Bloom Energy. Jeff discusses why Bloom is providing on-site power for large data centers and their utility providers and how fuel cells can address some of the community concerns about noise and emissions from data center projects. Let's get started.
Welcome to data center richness. I'm Rich Miller and I've spent 25 years telling the story of data centers, cloud computing, and AI infrastructure. In this podcast, I'm sharing conversations with the innovators building our digital future. Now, here's our show.
Jeff Barber, welcome to Data Center Richness.
>> Hey, thank you, Rich. Great to see you. It's been a little while. Thank you for the invite. Appreciate it.
>> You're most welcome. We've talked a number of times over the years about uh different elements of the data center industry.
>> Maybe you could uh uh share how your journey has evolved and uh uh and maybe a little bit about Bloom Energy in your current position.
>> Yeah, absolutely. So, um, yeah, my journey here, I would say Rich, it was circuitous is probably the best word for it. Um, I spent many years filling up data centers, right? So, I be actually I began my career in operations in the early 90s because I'm old. Uh, working for the California State Lottery, raised floor, mainframes, deck, vax systems, ancient technology, this brand new company called Cisco. I got to work with all of that stuff um, in my early 20s. And from there I transitioned into a little bit more hardcore technology with companies like Oracle and HP and and a very long time at EMC which was acquired by Dell of course. So spent a lot of years filling them up and then um went to work as a partner with a data center developer. So then spent a lot of years on the site selection construction tenant hunt if you will negotiating with utilities which was much easier a few years ago even. Um and uh, you know, I was intrigued by what I was seeing with uh with some utilities starting to pull you know will serve letters or or ESAs uh from developers which as you know can be tremendously impactful especially if you've signed a tenant and you've raised money and you've done all the things you need to do to build a multi-hundred million dollar facility. Um and Bloom had been been calling for a little bit. So I said, "hm, on-site power generation, that's really interesting." Started checking around a little bit. They had some great deployments with companies like Equinix and others uh and uh came over to stand up the data center vertical. So Bloom um just about three years ago, two and a half years ago, decided to really focus on data center, focused on developers, hyperscalers uh which we've doubled down on here recently at Bloom. So we now have um uh a vertical team with everyone from the industries industry companies like Digital and Equinix and um, you know, UPS vendors they all work on this team now so we speak the language and our uh our results are starting to show you know the fruits of that labor.
Yeah, it the earnings looked good yesterday and uh the data center sector was sort of spotlighted as as one of the areas where where Bloom is really making progress. I Bloom Energy has had such an interesting journey. I I started covering the company back in 2012. uh went out to the the HQ, I guess it was at Sunnyvale at the time, and uh and talked with Peter Gross and some of the folks there about uh outreach to the data center sector. Fuel cells were kind of a new concept for data centers at that time. It was kind of viewed as maybe a little bit of a niche for for regions where you had low prices for natural gas, maybe high utility costs, but so much has changed. It seemed part of this is you know that disruption sometimes moves slowly in the power chain but also the energy landscape seems to be a big part of that story. Uh maybe tell me a little bit about how Bloom sees some of that the evolution and changes in energy and what kind of opportunities this has created.
>> That's a great question. I'll I'll start with, you know, quick little Bloom history and fuel cell history. Bloom's been around over 20 years. I often I often hear when I'm speaking to customers that this is a new technology or it's or it's risky or what have you. We've been deploying globally for over 13 years now with with with a a very stable uh product. Of course, fuel cells were patented over 150 years ago. So I want to put a nail in that coffin that this is not a new technology. It's an electrochemical process that's very well understood used on all the moon missions as an example. Um since we're we're talking about Peter I'll uh one of my favorite quotes from him is the data center industry loves innovation as long as it's 10 years old. And that's very very true. Right. So back then uh when you were when Peter was working with the company, he's still heavily involved with the company, you know, we we had PayPal on Bloom for a decade under Dean Nelson um back in the day in Utah, right? That's a decade of no gens and primary base load power for a critical facility. Um, so that I think that really started that was the the catalyst to start the journey with data centers and from there you know the product has matured uh dramatically but the you know the bottom line there is Bloom Bloom has perfected the manufactured scale of fuel cells and Bloom has perfected that manufacturing process with no precious metals no rare earth metals US-based manufacturing you know supply chain risk is nil um so That that's some of the unique aspects of Bloom. The you know bringing this to tens or hundreds of megawatt scales to critical facilities is not easy. There's all kinds of thermodynamics and all kinds of manufacturing and all kinds of installation challenges that Bloom has overcome and the you know the results are are are showing that they're proving. So um yeah, it is a a perfect solution for distributed generation. Um, and the the other second half of your question, you know, Rich, I really see it as a the the company was built truly for this moment. um, you know, a little bit ahead of the curve I would say but if you think of the confluence we're seeing with uh generation coming offline and the the the glacial speed of new transmission right we simply cannot serve this demand as as a country and this you know obviously winning the AI AI race is critical so the concept of on-site generation for these critical facilities is now becoming the norm I mean we recently released our our our power report and we we the surveys demonstrated that data centers 38% of them are now planning on-site generation that to you and I in the industry we're like oh yeah on-site of course everybody's thinking about it but just a few months prior to that six months prior it was only 13%. we're considering, right? So, there's a critical mass. There was a tipping point and we're certainly seeing that. I know we'll get into the a relationship later and we could talk about some recent transactions such as Oracle in the news, but uh it's really that confluence of generation coming offline, transmission not moving quickly enough. Um I will speculate that that future transmission will be much more expensive than uh than Bloom certainly but much more expensive than the market is accustomed to. Right? Copper pricing is not going down. Steel pricing is not going down. Right? Those things will show up in the rates and there there is in many markets Bloom is already cheaper than the utility. Uh I believe that that trend is going to continue and be be much more macro. it will become the norm that Bloom will be the most affordable uh form of electron on the market.
>> Yeah, that that will there's I think always been this perception that fuel cells were perhaps a premium solution. I I think the last couple of years with the power constraints in many key data center markets has really changed the the willingness and really the the requirement uh to think about on-site power because the the customers in this case the hyperscalers um they have schedules to meet they're all competing with one another for AI supremacy what whatever there's there's some very ambitious customers who see the speed to market for their data center capacity as a differentiator. Yeah. With that backdrop, uh how does Bloom fit into the the new wider universe of options that's available and that data center operators are looking at as they think about rather than just taking the utility feed, doing something with on-site power. Where do you guys fit?
>> No, I love that question, Rich. Yeah, so Bloom um Bloom solid oxide fuel cells are intended to be primary base load power. I'm not I'm not a backup gen and I'm certainly not a battery. Something I hear quite often. Um I mean they kind of look like mini batteries, I guess, but I'm not a battery. So primary base load power with or without the utility, that's a critical point. um to your earlier statement, yes, there's a very real ROI for both the developer and the tenant in coming to market years ahead of of uh what what they're going to see from the utility or if utility is there um how about densifying the data halls, rack densification, right? Nvidia is talking 150, 200, one megawatt racks. Don't ask me how we're going to cool that, but the industry the industry is is creative. um, but basically what that's saying is, okay, if I have a 50 megawatt connection from the utility, but I need 300. Okay, deploy 250 of Bloom, deploy 50 of the utility. We're simply supplementing those electrons. um, you hit on another point in your statement as well. You know, Bloom does have, I would say, premium electrons in the sense that there's not there's no combustion. There's no NOx or SOx with Bloom, right? No pollutants coming out of our box like like combustion would have. We're giving you triple converted power. So, I'm giving you UPS quality power, a perfect wave into your facility. um, and that that is premium, right? These are designer electrons outside the building. um, it's important to note that inside the building or even if I'm consolidating through MV or transformation into the building, nothing changes. Those electrons are no different than anyone else's electrons. Your your ATS infrastructure, UPS, PDUs, busway, you know, tap cans, nothing changes. Uh, your facility simply believes that it has a 300 megawatt utility connection in this example. um, another the other side of that coin is if utility is simply not available um whether long lead items new transmission needs to be built substations need to be constructed Bloom does not require any of that. So we can be completely islanded our grid former forming inverters are Bloom designed and manufactured. They're incredibly robust. This is not like solar and wind grid forming inverters. These are mission critical multiple nine type of of products. So with or without the utility um we can switch if utility does show up in the future we can integrate at that time as well. So where does it fit? Time to power much cleaner solution you know very real ROI for the for the tenants and for the developers by coming to market so much sooner and no impact to the residents. That's very important, Rich. Right. The the the the solution is extremely quiet, 65 dB. um it does not require new transmission. It does not require substations and permitting and easements and all of the bureaucracy that comes along with that. Most importantly, it's the developer andor the tenant solving their own power problem. The these these electrons, these rates are not being passed to the residents and to the other buildings, right? or other commercial customers.
>> That seems to be a a an issue that's rising in visibility. Obviously, the the idea of community resistance to data centers is not new. I was covering this back in 2014, but but I think some of the media coverage of uh resource usage by data centers has in fact prompted greater attention uh from residents who are concerned about these these issues. Do does that come up in the conversations for folks who are considering energy options and and uh and looking at Bloom?
>> Absolutely. Every single time. uh it comes up, right? It it it's it's uh it's the economy stupid, meaning it hits their it hits the residents wallets, right? Their rates go up because a large developer requests 500 megawatts. The utility services that well that transmission and that load uh expense is typically passed to the other businesses and to the residents. Not always. Now you have you do have forward-thinking utilities like AE that are solving this problem with distributed generation right it is a our industry as you know better than anyone Rich does have I think a a forthcoming image problem right now it it contributes a tremendous amount of of employment um, you know, people forget that a large campus uh, you know, a gigawatt scale campus will employ people for a decade most likely 8 to 10 years at least. The tax revenue is significant. But the perception matters, right? So I I think that distributed generation is a very important way to ease that perception to improve that perception, especially when it's quiet and it's not it's not producing any pollutants.
>> We've all been sort of a little bit amazed at the speed with which uh data center scale has accelerated. At least I am and I've been watching this industry for a long time. Uh some of the numbers we see on scale when we're talking about gigawatts, it is a different ballgame in terms of being able to have supply chains keep up. Uh and within the on-site power world, uh there's a lot of discussion about what uh the availability and timelines are for generators, gas engines. Uh when it comes to being able to scale up manufacturing, um how is Bloom situated to be able to deal with the kind of scale that we're seeing nowadays?
>> Very very well. the company, you know, obviously what we do is power as a company and the company saw um this confluence that I mentioned earlier of transmission and generation coming offline and the company began to prepare for it. That's what good companies do. So, we you mentioned uh in the before we started recording, you mentioned that you'd been to our Sunnyvale uh manufacturing facility that has has moved to Fremont, California in a massive uh gigawatt scale plus manufacturing line. That's where we produce the wafers. We'll be at 2 gigawatts of production by the end of '26 and that can be doubled again. um, a very big part of what we did there was standardize the entire manufacturing process and scale it up. You know, knowing that this confluence is happening and would only grow. um, the machine learning AI revolution, if you will, only increase that exponentially, right? So, yeah, we're ready for it. um, you're you know you're seeing that again in with partnerships such as AE where they're you know they've contracted for up to a gigawatt primarily focused on developers and hyperscalers. Bloom can meet that demand and deploy in months versus years.
>> Yeah, I want to talk more about AE and some of your other uh recent customer wins in a minute. Uh one of the things that comes along with scale is there's a lot of discussion about about different kind of workloads. AI is a little bit uh different from cloud in terms of both the heat that it's throwing off obviously which as you mentioned the the the cooling departments in many data centers are are scrambling to sort out solutions for on the power side uh the uh jobs are large and can spin up and down uh fairly quickly. uh you know does that present any any particular challenges or or opportunities for Bloom?
>> Yeah, great great question Rich. Yeah, the the oscillations in workloads the fluctuations in the micro millisecond one second one minute range depends on what you're doing. As you know they're wreaking havoc within the facility and even with some even with some utilities. UPS vendors are seeing problems obviously uh, you know, breakers and switches and all of that. um, this situation is absolutely perfect for Bloom and I'll tell you why. When the company started deploying microgrids in other words the ability to to to form the grid to load follow to step in loads 12 years ago. We we have deployed ultra capacitors as part of our product for well over a decade. It it turns out that that very fast discharge recharge capability is perfect for these types of workloads. I do hear a lot in the industry, mostly competitive FUD, but I do hear a lot that, oh, fuel cells are not good with oscillating workloads. Anybody listening or watching right now, please reach out to me and let me squash that in about five seconds. um, you know, if we look at the solution that Bloom had, I'd love to say that we predicted this. It wasn't a prediction. Ultra capacitors were just the best the best uh energy storage to instantaneously respond to to load profile changes, whatever the type of building, whether it's a factory or retail or a health care system. um, when this AI revolution hit the market, Bloom was sitting here with the perfect solution that's being validated, right? The GPU vendors are now requiring power shelves for instance that is in the rack. That is capacitance. UPS vendors are now deploying capacitors previous to the batteries. Right? So we have UPS systems and and and shelf systems handling the micro and millisecond loads. um, if it gets to the on-site generation from Bloom, we have ultra and super capacitors as a a matter of routine for over a decade now. So uh, you know, that we have released a lot of information on this. This is not just just Jeff talking. There is a tremendous amount of empirical data like I said. So please reach out to me and let me let me kill that one really quick. So no, it's the it presents a massive opportunity for Bloom. You know, these are these are Bloom design systems um that happen to be perfect for this type of workload.
Well, one of the examples of that is one of the the sort of big wins, the early big wins I saw in this uh recent sort of uh series of uh customer announcements from Bloom involved Coreweave, which is one of the leading players in AI and certainly knows a lot about AI and and their uh, you know, workload requirements and and power. Uh, you're working with them on a project in Illinois. How did that come about and how is Bloom able to to help that project in that specific geography?
>> That the developer there is Carissa Cororeweave is the tenant and the Bloom customer um as is public. That's a scenario where Bloom um is able to densify um a brownfield or you know might be a new build or a new fit out. They were not getting enough from the utility. Bloom is able to interconnect with the utility and simply supplement that feed. It it really is that simple. um, you know, they look any tenant or or developer will look around and say, "Oh, geez, I don't have enough power, right? I need I need 100KW per rack. I'm designed for 40. What am I going to do? I need more from the utility." Well, you can't get it. So, let's do on-site. Let's integrate with the utility. We interconnect with the utility. Again, once those electrons are in the building, there is no operational impact whatsoever and no design impact whatsoever as far as the facility knows. The utility just has a larger connection. Right. So, that's a perfect time to power example.
>> Yeah, that seems to be obvious priority for a lot of data center operators and historically you've done a lot of these deals with particular companies. Coreweave uh you've done deployed I guess about 100 megawatts now with Equinix but you mentioned the AE relationship which is a little bit different where you're working with the utility which then uh uses the the Bloom Energy boxes to help support uh data center campuses within their territory. I I think this has become a key strategy in the the Ohio data center cluster. How is that uh, you know, structured and and uh, you know, what are the the u what's the upside and the opportunity for uh the AE relationship?
>> Oh that's that's a fantastic question and I would say that you know Bloom is on the deregulated side of AE so they can actually provide equipment anywhere in the US. It's not just the AE territory. The AE relationship actually started over five years ago. They had owned and uh and worked with Bloom for five years where at the end of '24 they said look we are not going to be able to meet the demand from these developers and from these hyperscalers. It's just not going to happen. It's mathematically impossible. What are we going to do? Well to AE they're very forward-thinking and they do think about their customers first and foremost. Let's service this load with on-site generation. So, uh they approached Bloom. um, they have contracted for up to a gigawatt of Bloom capacity to to satisfy these customers, retain these customers. And most importantly, this is an approved utility provided solution from AE A tolling rate, AE invoice. It looks very very familiar to the developers and the tenants. I have my utility bill. I pass that through to my tenant or the contract is directly with the hyperscaler as an example. So uh it's a massive partnership for both Bloom and for AE very very well I think you've seen AWS is now contracted as is Collogix. So you have you know hyperscale and retail collocation. there is uh much more to come and at the end of the day uh they also very much looked at the fact that again on-site generation from Bloom is zero impact to their ratepayers. They they're not they're not needing to construct massive new transmission lines on an accelerated timeline or bring online some form of dirty generation. They're able to satisfy those customers with on-site generation and it's very very comfortable for the tenants and for the uh for the developers. So, uh, I think you're going to see more of this, Rich. I think the the utilities are, um, utilities are struggling obviously to meet demand. That's not news. um, I think that the timelines uh, for not just AI and ML, but let's not forget about cloud growth in general, right? The timelines are are compressing and the rack densities are going up. um, right? So, you know, you're I think you're going to see more of this where the utility provides on-site uh generation to the customer.
>> Yeah, that's an interesting development. There's uh I've been learning a lot about the utility industry and about uh how grids work and what the options are as they struggle both to sort of deal with the the load growth with interconnection queues and really with all the complexities of operating a grid. It's a it's a tricky business, but you mentioned obviously the a lot of customers are growing quite quickly. One of those that's that's working with some uh people who are scaling up pretty quickly is Oracle and you recently announced jointly with them uh that you'll work together to deploy capacity at some new data centers. Tell me a little bit about that uh that deal and uh what it looks like and and what Bloom's going to be doing there.
Yeah, I'll share what I can. Uh there was, you know, a press release as you saw. um, the capacities are confidential, but I will tell you this is not a, you know, one megawatt science experiment. This is at scale Bloom deployment for on-site generation within 90 days, right? That's an important part of that press release. Right? So, we're going to um we've already demonstrated with customers such as such as uh uh QTS here in the Bay Area that we can do this in less than 90 days and we're going to do it again. And you know, the world is going to see that look, this is truly a more efficient, cleaner, economical way for on-site generation. It's important to remember, Rich, that um, you know, Bloom's Bloom's resiliency and reliability because it's electrochemical. There's nothing spinning in these boxes. These are not engines. It's an electrochemical process. um, our resiliency and redundancy does not come from a 2N or an N plus one or a catcher system type of design. Okay? I'm not failing over to anything. I simply have 65 kilowatt Lego bricks that I'm deploying on common busways from 1 megawatt to hundreds of megawatts. What does that mean? Okay, in the event of maintenance or a failure, my blast radius is 65 kilowatts to to a 100 megawatt or 200 megawatt facility. You're probably never going to see that. In fact, I know you're not going to see that. Uh, but even if you were, it's not an issue for us to ramp up additional capacity on site and and cover that maintenance. As an example, there is no concept of um downtime for maintenance with Bloom. We're not bringing you down to rebuild a turbine, you know, for multiple days or weeks. you're not you're not deploying 25 to 50% more capacity as you would in a combustion world in order to cover those maintenance windows and those predictable outages. So um that's really what Oracle saw. um, you know, it leads to a lot of interesting conversations when you think about the fact with 10% additional capacity we'll give you we'll give you four nines 5 nines of of availability right and that's that's unheard of.
There are a lot of moving parts in the current energy and data center landscape. It's uh very hard to see into the future and figure out what's coming next. Particularly when you look at some of the change we've had just in the last few years, most of which I I don't think most folks had on their dance card. Given that, I'm going to ask you to kind of look into the crystal ball a bit. What do you see uh and what's Bloom's perspective on what the data center and energy landscape is going to look like going forward, how it might evolve, and how it's thinking about serving customers with what comes next.
>> Our crystal ball definitely says on-site generation, and that's not because we have happy ears. um, that's just math. That's just that's just the thousands of miles of new transmission we need. and looking at a trend of of building tens of miles per year, not hundreds or thousands. um, the electrons simply cannot get there to meet the demand. It's it's no one's fault. I mean, you know, the grid the grid was built multiple generations ago um as consolidated electrical distribution. um, the future is distributed generation. The future is solving for data centers at least solving our own problem. you know, what does that mean? Again, we have a reput an emerging reputation problem. I would say you're seeing that with um legislation and policies requiring these developers to for instance uh pay for at least 85% of what they request. We see that in the AE territory um related to you know rates for that load. They don't mention transmission which is the real big expense. um, you're seeing that with a with an increasingly loud voices about the noise and the pollution. um, you know, you you see watch what went on recently in Virginia with uh with with demand generation with the data centers firing their their diesel gens to ease the load on the grid and then the public uprising shortly thereafter. You're going to see more of that. There's room for multiple solutions by the way. um, Bloom happens to be the cleanest and the most reliable. um, and you're seeing extended lead times with competitive offerings for sure. One of the things that doesn't get talked about enough in my opinion in the market is, you know, the fact that these developers and tenants from a hyperscale perspective, they are entering into 5, 10, 15, 20 year uh, PPAs and capacity agreements with their vendor. You don't want to do that with a startup. You want to go with a financially robust partner, publicly traded, decades of experience. um, you know, there's there's a lot of startups out there that are claiming, you know, the next big thing. Just in my opinion, too many unknowns in that market. So, yeah, the future looks like on-site generation to ease all of the all the political the ratepayer uprisings, the and let's not forget about ESG, right? Burning stuff is bad. You can you can quote me on that, right?
>> Well Well, listen, Jeff, it's uh it's been uh uh awesome talking to you, hearing more about uh the this whole changing landscape, where Bloom fits and what you guys have been uh uh doing. If our listeners and and viewers are interested in learning more about uh Bloom Energy and what you guys are doing, uh where can they go for for more information?
>> Obviously, our our website is great. bloomenergy.com. Follow us on LinkedIn. uh follow me or reach out to me. I have a a quarterly newsletter and lots of podcasts like this to learn more. um, or or just email me directly at jeffrey.barber@bloomenergy.com. um, I'm I'm Jeffrey when I'm in trouble. Normally I go by Jeff.
>> Many thanks for listening. If you like our content, you'll want to check out the data center richness substack. We're just now launching and have lots of great insights in the pipeline. If you're a data center professional or someone tracking the sector for your business, you will definitely want to check it out. I'm going to put a link in the show notes for the podcast and also in the description for the video on YouTube. And as always, if you've enjoyed the podcast, please give it a like and be sure to subscribe to our channel as this supports my work and ensures that you won't miss out on any of our insights from data center richness.
[Music]
[Music]