David Holmes  00:00

We need a lot of energy to power the AI systems that are in production today, and that are being deployed and scaled out in the future. AI for energy, though, is thinking about how can we apply AI to solve some of the challenges facing the energy industry as a whole, and that's not just challenges created by AI, it's some of the broader themes that are happening within the industry around broad scale electrification, the growth of intermittent generating capacity, the stresses being placed on the electrical grid. AI has a huge role to play in addressing some of those challenges.

Nico Johnson  00:37

Welcome back to SunCast, and thank you for giving us the only non-renewable resource you've got, of course, that is your time today. I'm joined by David Holmes, the Global Industries CTO for energy at Dell Technologies. I'm sure many of you are familiar with Dell, but did you know that Dell now sits at the center of some of the largest conversations happening around AI infrastructure, edge computing, and yes, energy. David has spent much of his career in and around the energy sector, and I think it gives him one of the best vantage points on the convergence of energy and compute. David, welcome to SunCast.

David Holmes  01:14

Thanks for having me. It is

Nico Johnson  01:15

fabulous to finally get a chance to sit down with you. I've had the pleasure of becoming a follower on LinkedIn. You've got a great perspective, in my opinion. You've described this challenge that we are facing as both energy for AI and AI for energy. I thought we'd start there, but before we get into the specifics, can you define those two ideas for us?

David Holmes  01:39

Energy for AI is a very simple one. We need a lot of energy to power the AI systems that are in production today and that are being deployed and scaled out in the future. AI for energy, though, is thinking about how can we apply AI to solve some of the challenges facing the energy industry as a whole, and that's not just challenges created by AI, it's some of the broader themes that are happening within the industry around broad scale electrification, the growth of intermittent generating capacity, the stresses being placed on the electrical grid. AI has a huge role to play in addressing some of those challenges.

Nico Johnson  02:16

Let's take the first half of this equation, energy for AI, everyone is talking about load growth, and at first it was, we're seeing a tremendous amount of load growth, and then it was, well, it turns out all these AI data centers are precipitating most of what we might consider to be load growth, obviously there's electrification and EVs, et cetera, but as AI infrastructure scales, what changes in the way we have to think about electricity demand? The

David Holmes  02:43

biggest change that's happened is that energy is now a board level concern. If we go back a few years, all board members of large companies were concerned about what the cloud strategy was, and then a couple of years later everyone was concerned about what the cybersecurity and cyber resilience strategy was, and now energy is becoming a board level concern as well, and the reason for that is the way in which AI is changing the way the energy is being consumed. Data centers that used to consume 510, 1520, megawatts of power are now consuming hundreds or even 1000s of megawatts of power, and the load profiles are changing as well, and this is causing significant challenges for electric utility companies that are not used to rapid load growth, and they're certainly not used to sudden changes in the shape of loads as well. So this has caused a big conversation in the industry about how we can address and rise up to the growth in demand for energy and the changes in the load patterns that AI, in particular, is is causing.

Nico Johnson  03:52

Yeah, definitely, we'll get into some of the trends we're seeing around load patterns that AI is causing. I guess that also factors into the second half of this framework, AI for energy. I'm curious, where you're seeing AI create meaningful value for energy systems.

David Holmes  04:09

That's a tricky question, because there are very few places where you don't see AI being deployed across the energy ecosystem. So, Dell Technologies, we think of the energy system as being kind of two real businesses. There's a molecules business, which we used to call oil and gas, but is now oil, gas, co, hydrogen, ammonia, anything that's molecule-based, LNG, sustainable aviation fuels, and all those things. And then there's electrons business, which is power generation, transmission, distribution, storage, and consumption, and so when you look across that entire ecosystem of the molecules business and the electrons business, AI is being deployed across that to solve lots of different problems, so to pick a couple of different examples, if you take to. Additional oil and gas company, the way in which they're characterizing reservoirs, the way in which they're delineating reservoirs, the way in which they're understanding how to produce hydrocarbons in the safest, most efficient, and productive way is being augmented by capabilities that are providing to analyze vast amounts of data and provide new insights. If you think about the electricity ecosystem, there's lots of examples of how people are applying AI to do things like look for stranded electrons on the grid and figure out how they can be accessed. To look at how we can build much more contemporary demand response systems that allow us to balance load and supply, and perhaps solve one of the most pressing challenges around how we ensure we're maximizing the utilization of intermittent energy generation. We've seen huge problems where there's been huge amounts of curtailment of renewable energy generation, because there's nowhere for the electrons to go,

Nico Johnson  06:00

and

David Holmes  06:00

then in the same month, you'll see some lockdowns on nobody can charge EVs from six till 9pm because they don't have the power being generated, and so AI provides a perfect set of tools to be able to address some of those challenges.

Nico Johnson  06:16

Yeah, and as we'll talk about there, there has been up to now a big gap, technology, technology-wise, that needed to be bridged around both the intermittency and the shoulders of where electricity can be harvested versus deployed or needed in the supply chain. Most people hear the word Dell, and they think laptops, that's not the Dell that you work inside of every day. In fact, if anyone's paying attention, the time of our recording, you all just vaulted. I wish I had become an investor in Dell two weeks ago when we were planning this interview. 33% jump in stock, principally driven by the AI and data center business, expecting a record 61 billion in in revenue in the coming quarter. Talk a bit about the piece of Dell that you inhabit, what global industries is, what it means to be the CTO of energy within that.

David Holmes  07:13

It's a really good question, and you know, honestly, I've been on executive energy round tables where people have wondered why is the lap Dell. Yeah, I'm not quite sure what their perspective is. So, let me explain a little bit about global industries, because we're quite an interesting part of Dell. Inside Dell, we have, you know, 30,000 odd people who are really good at selling it to IT people. Now, Dell's portfolio is much larger than many people realize. Everybody's familiar with the desktop, the gaming PCs, the laptops. Many people will be using them, but actually, Dell is also one of the largest infrastructure providers across data storage, data protection, compute, AI systems, edge systems, operational technology, and so we're a huge part of our business comes from that, and it's perhaps not something that many people see in the mainstream if they're not involved in the IT organization. Now, where my role is quite interesting is, you know, we have a global team of subject matter experts across multiple verticals, key verticals we cover are healthcare, life sciences, manufacturing, retail, and energy. So, in the energy field, we have subject matter experts across the electrons business and the molecules business that I talked to a moment ago, people who've got lots of domain expertise in that field, who are looking at the application of technology to solve the specific challenges of the energy industry, so we're really interested in how we can help deliver business outcomes for the energy industry. So, if I get an account executive from Exxon phone me up and say, 'Hey, I think we've got an opportunity to sell some new Windows laptops to Exxon, then that's great, but there's not a huge amount my team can do to really help and support that type of conversation. If, on the other hand, we get a phone call from an account executive from Center Point Energy saying, hey, they're looking at how they can apply digitalization to substation operational technology systems, or they're looking to deploy an advanced distribution management system. This is something where we have a point of view. We partner really closely with all of the key industry vendors, companies like SLB, GVNOVA, ABB, Schneider Electric, Emerson, all those types of companies to validate, optimize, and certify their platforms to run on our infrastructure, so we can provide an easy button to delivering outcomes faster to solve some of the specific problems that energy companies are wrestling with.

Nico Johnson  09:55

I'm not sure I understand what it means their platforms to run on your infrastructure. Could you. Pack that for me a bit.

David Holmes  10:01

Let's take electric substations as a really good example. A typical electric substation has 15 or 16 boxes that look a lot like a PC. They have a CPU, they have some memory, and they have some storage, and those systems will run a whole variety of different operating systems. It could be Windows XP, it could be Linux, it could be a real-time operating system, you name it, it will be deployed out there. And these systems are typically connected via RS 232 serial cables into a serial concentrator, and it's, it's how substations have operated for several decades now. That creates some real challenges, because upgrading these systems requires you to go and physically visit the substation and do work. If one of these systems breaks, it requires a visit out to the substation. So, in partnership with Intel and VMware, we put together a solution called Virtual Protection Automation and compliance, it was a virtualized system for electric substations. We said we're going to replace all of these single function devices with a highly resilient, rugged cluster of computers that comply with the IEC 61 853 standards. And once we built that, we then built a consortium, the VPAC Alliance, which has includes companies like Schneider Electric and many others, each of whom are subscribing to this new way of operating substations that says actually we can deliver things more efficiently, more reliably with lower levels of emissions and with low levels of cost and improve the cyber resilience of the delivery of compute within a substation, and so that's the kind of thing where we'll work with a company like Schneider Electric to validate their software running on our virtualized platform.

Nico Johnson  11:56

At what point did energy become a strategic decision or a strategic focus for Dell.

David Holmes  12:02

Well, that's interesting. I think Dell has been a significant provider to the energy industry for nigh on 30 years now.

Nico Johnson  12:15

Yeah,

David Holmes  12:16

it was one of the, you know, Dell was founded in up in up in Austin, and so not very far from a lot of very large energy companies, but actually the history of my team is quite interesting, because there was a company called EMC, large infrastructure provider, EMC owned VMware, RSA, Pivotal, and had its own storage portfolio. That company was acquired by Dell, Dell Technologies came into existence out of the merger of Dell and EMC. I was an EMC employee, so inside EMC we had the Global Industries Organization focused on those verticals I mentioned earlier, and so with the acquisition there was a question mark about, well, how will Dell Technologies approach addressing vertical markets, which have higher growth opportunities, where you can demonstrate subject matter expertise in that particular industry, and so the organization came from EMC, was augmented by Dell Resources, and so that became the kind of like the real genesis of a focus, and it was enormous for me, because I went from having EMC portfolio of products to be able to build solutions to having the entire Dell Technologies portfolio, and it gave me a lot more flexibility to build solutions that would more broadly address the needs of the industry.

Nico Johnson  13:42

It's fascinating when I I, when I interviewed someone I've recently learned is a longtime friend and colleague, Mark Spieler from Nvidia, I didn't realize the specialization inside of Nvidia to different verticals, which, as you're explaining the Dell Technologies approach, it sounds very similar. I mean, it sounds to me like you and Mark have identical jobs at different companies. When I started pulling back the layers, I learned a tremendous amount that I think our listeners and most of the public don't know about how Nvidia seeds ideas more than anything about how people can build on their infrastructure. How does that match to what you're doing at Dell?

David Holmes  14:26

Oh, very, very closely. I mean, it's interesting when I, when I think back, kind of 12 years ago, what I was doing inside the Global Industries Organization at EMC was quite groundbreaking. This idea of having industry subject matter experts, people who are experts in the industry first and technology second, was kind of quite, quite groundbreaking.

Nico Johnson  14:48

So, Nvidia followed you.

David Holmes  14:52

I, I would say we had a, we certainly had an energy team before, before Nvidia did, but, but, but it's, it's a model. Been well replicated. I think it's recognized that if you really want to help an energy company solve its problems with technology, then you need to understand what those problems are. You need to be able to speak the language, understand the terminology and phraseology, and you need to be able to communicate with stakeholders outside of it, I mean, another important difference, and Mark would say exactly the same thing. Mark does not spend all of his time speaking to people inside IT organizations, neither do I. We're interested in speaking to business stakeholders to help explain how technology can address challenges of the industry, not get into conversations about speeds and feeds and networking, and how all these things work. Not that those matters aren't important, but they're not how you get to solving the business problems that stakeholders are interested in. The bill of materials almost comes at the end of the conversation. You know, it should really be about you have this challenge, we have a solution. This is how we'll address it. By the way, this happens to be the collection of components that will address the problem, and this is why this collection is the right one to solve the problem. But really, you start with the outcome first, and then technology follows along. I

Nico Johnson  16:16

mentioned that you've had an illustrious career in and around the energy sector. When you look back, what prepared you the most for being able to recognize, and I presume you recognize, before many others, the convergence of energy and compute that many people are talking about today.

David Holmes  16:33

Well, it's interesting. I started my energy career in oil and gas. Interestingly, my, my background really started out of before the cloud was really even thought of. We had this idea of ISPs, internet service providers. They were the people who glued everybody and connected them to the internet. And then we had this idea of ASPs, or application service providers, and somebody had the crazy idea that we could run applications on computers connected to the internet, which today seems like why would you do it any other way, but back in 2001 it was a pretty revolutionary concept. So I was, I was working with Halliburton at the time, and we were working on an ASP solution for complex d visualization applications, actually have a patent for the for the work that we did at that time on how to make d applications run on internet computer connected systems. So

Nico Johnson  17:36

it

David Holmes  17:36

was a very interesting time and a very interesting introduction to bringing together the world of data centers of oil and gas workloads, oil and gas data management, understanding the relationships between the services, the stakeholders, the different software providers that operate in that industry. After that, I spent - I was lucky enough to spend five years working in a startup that was working on workflow data and information management systems for the oil and gas industry, and that was incredibly insightful, because it was really working around a lot of the transitions that were happening just as the cloud was just starting to be talked about, so the time we were one of the first people to, you know, run data management systems on the on the public cloud. In fact, there's a famous presentation at a big Norwegian data management conference where myself and a colleague stood up and we talked about the potential for cloud computing, and everybody pretty much laughed at us. They said, in the energy industry, we'll never ever use the cloud, because, wow, you know, there are latency issues, there are regulatory issues, there are security issues, there are all these different concerns, and it's interesting how that, that, that came to change, but that, as that company was being acquired, and I was kind of like, I was done with the 80 hour weeks, and I'm working for two candy bars and a packet of crisps, although it ended up pretty well, but, but at the time, it sort of, it was, it was a, it was, it was an interesting period of exploring technology and software development in a, in an internet connected environment, but that was the point at which EMC phoned up and said, hey, we're looking for an oil and gas CTO. Would you be interested in the role, which I absolutely would. And what was interesting about that, everyone knows that oil and gas is a cyclical industry, and so during one of the downturns in oil and gas, I said, you know, we focus a lot on oil and gas, but actually electric utilities is every bit as important. So we renamed our vertical from oil and gas to being energy, and it was a time at which all of the big, you know, the Shell, BP, Total, Equinor, they were all rebranding, we're energy companies now, not. Oil and gas companies anymore, and so it felt like a good time to think more holistically. So we hired an experienced CTO from the electric utilities business, and we started building out the capabilities in that space, developing a lot of new partnerships and learning really, really fast to the point where actually you know three years ago as AI started to become a serious part of the conversation when, when we were talking about, you know, the intersection of energy and AI, it had kind of led us to that point, where we were like, actually, we now understand the entire energy ecosystem and supply chain. So, whether you're talking about behind the meter generation with natural gas, whether you're talking about deploying new distribution management systems to manage renewable energy, we had all the pieces in place to be able to understand the narrative, the challenges, and the solutions that were available to be able to drive the deployment of this technology more quickly.

Nico Johnson  21:10

I imagine you spend a lot of time as an interpreter between the tech world and the energy world. The two executives speak different languages in many ways. I'm curious, your perspective on the gaps you see between how one side understands the other.

David Holmes  21:30

There are a couple of different dimensions to that. There are absolutely times. Sometimes I describe my role as a protocol droid between IT people and business stakeholders, and I've sat in a room with a major oil and gas company, which will remain nameless, but the IT team had no clue what the 10s of millions of dollars of infrastructure was actually being used for. They didn't understand the data that was on the infrastructure, they didn't understand why the business kept wanting more and more of it, and it was clear that in order for them to be effective, they needed a kind of like a 101 of let me explain to you, where's this data coming from, what are the different types of data, what are the formats, what are the characteristics that your business users are looking for? What's the really important data, and what's the data that you can kind of archive away and move out of the way? And so often you're exactly right, you do sit and you translate, because, because if you're talking to a geophysicist, they're not really interested in petabytes and gigabits and GPUs, they're really interested in how they can build an understanding of the subsurface faster, more accurately, and with new new insights.

Nico Johnson  22:53

It's interesting as well, the tech companies think of the FANG they spent much of the last 15 years watching the growth of renewables, the sort of the decarbonization of the grid, making their own commitments to decarbonization. What observations do you have around tech companies becoming energy companies? What I observe, and at least as a layperson who's been only focused on renewable energy for much of my career, is that we've been service providers to and helped build these in for this infrastructure, which is nothing new for the oil and gas sector, but but I look at Amazon, I look at Google, Google just now acquiring Intersect, and it's not, it's not far-fetched to believe that they're going to compete with utilities in some way as energy providers. What does that future look like for you as a technologist?

David Holmes  23:51

It's an interesting time. I'm not sure that there are any technology companies that are really excited about becoming energy, the utility

Nico Johnson  24:01

business,

David Holmes  24:03

and if you look at the constraints that utility business operate under, if you look at the length of their capital investments, and if you look at the returns that they get,

Nico Johnson  24:13

guaranteed 9% Yeah, Dell's not interested in that.

David Holmes  24:20

I didn't say Dell wasn't interested in that. I said I don't know of any technology companies that are excited about becoming energy companies, but, but, but I think what's important, it's, uh, it's becoming a matter of necessity, is there is huge competition for electrons, and one way to secure your electrons is to acquire the means of generation to to purchase your own generating capacity or to sign in and lock in long term supply agreements from from power generation companies and what's interesting is looking at the way in which. Companies are looking at how they want to be grid connected or not grid connected, and the regulatory environment has a huge part to play in that as well. So, I think I don't think there's anyone excited about becoming an energy company, but to go back to your question about what point do some of these companies become competitors in this space? I think utility companies are under a really complex set of constraints. They are absolutely on the hook for delivering reliable, resilient, affordable, and sustainable power, and typically they're regulated pretty fiercely to be able to do that. The way in which they manage capital budgets is very different from a lot of privately held companies, and so they have constraints that they operate under, but the reality is, is that there is going to be disruption to the energy markets, and so one of the big questions is, is that disruption going to happen within utility companies, or are we going to see virtual energy companies appear that take advantage of the opportunity to arbitrage power to provide the capability to offer flexible AI-driven demand response in a way that does compete. Now I'm not sure that the big hyperscalers are necessarily the people that are that are going to leap head first into that space, but there's certainly a number of technology companies that are looking at hard at how with minimal capital deployment they can start constructing offers that help support the grid.

Nico Johnson  26:47

Sure, yeah, octopus being one of them. I think there's a lot of, there's a lot of value at the grid edge that is still woefully locked up. I recently saw a LinkedIn article that you wrote around the concept of needing to maximize the value or the utilization of every electron on the grid behind the meter that struck me as a simple phrase that carries a pretty profound idea with it. What does it mean to you?

David Holmes  27:19

Well, it sounds obvious, doesn't it? Like at a time in which we're desperate for power, we should be using all of the electrons that are that are available to us. What that means is we need to identify where there is stranded power. You know, there are lots of places where there's huge amounts of capacity, large amounts of the time. There are substations that have 456 10 megawatts of spare capacity, 90% plus of the time, you could co-locate compute there, and for the majority of the time you could absolutely power some significant AI infrastructure without needing any new transformers, any new transmission lines, any distribution lines, any of the things that are on long, long lead time.

Nico Johnson  28:05

Yeah,

David Holmes  28:06

and so I think it's beholden on us. If you look at, you know, some of the stats that Emerald AI are quoting, they're suggesting that there's 120 to 150 gigawatts of spare capacity on the grid, that if we could operate the grid more efficiently, we could take advantage of a significant chunk of that. To put that in context, you know, ERCOT right now, what are about 80 gigawatts? So it's kind of like it's one and a half times ERCOT that's available if we could take advantage of it.

Nico Johnson  28:37

You've spent a lot of time in both of these worlds, what's evolving technologically to allow us to decouple the market mechanisms that force sort of nodal constraints or force on site, off grid, if you will, support of data centers rather than moving them to more logical locations,

David Holmes  29:00

so I think when it comes behind the meter generation, I think that's just a practical solution that companies are looking at, because exactly like you've said, the internet interconnect queues are huge, and so if I can get my own generating capacity, then I can essentially be off grid, now that creates a whole bunch of different challenges for any data center operator, and it also creates potential challenges for grid operators as well. It's probably not necessarily in the best interests of a grid operator for half of the generating capacity in their region to be disconnected from the grid. That said, it is the pragmatic way in which data center operators can build and scale quickly.

Nico Johnson  29:46

It's also not in the best interest of the data center operator to not connect to the biggest battery we've built over the last 100 years, which is the grid, and have capacity to both absorb or or supply, which in the world. Where VPPs are a real and present opportunity, it further underscores for me that these tech companies, who hyperscalers in particular, who are making a legitimate run at speed to power behind the meter assets, we disrupt ourselves all the time in the industry. There was a huge arbitrage opportunity in Texas, and now there's a glut of battery storage, right, because of the arbitrage that existed. I propose that the same arbitrage exists right now for these hyperscalers, which is going to flip on its head and say, "Oh, actually, we're asset owners and can sell to the market, and especially in a freely, in a deregulated market like Texas, all the tech companies are going to have these assets, and then figure out through VPPs, not figure out, but leverage through octopus and others ways to sell that power that are challenged that are going to challenge what we consider today the modern utility.

David Holmes  30:52

It's possible, but it's worth contemplating the relationship between power and revenue.

Nico Johnson  31:00

Yeah, for

David Holmes  31:00

these operators, so if you're operating, you know, 100 megawatt AI factory, then the revenue that's generating means that your power costs are relatively small.

Nico Johnson  31:15

Yeah,

David Holmes  31:15

and so your opportunity to maximize your revenue is driven by making sure that you're driving your very expensive 10s of billions of dollars of compute infrastructure

Nico Johnson  31:27

at

David Holmes  31:28

maximum utilization, and so whilst you might feel a social obligation to support the grid in times of need, your priority is going to be to where you can to maximize the use of that that power to run your infrastructure. Now I'm not saying that at all that data center operators are bad, operate bad, bad actors, quite the opposite. In fact, you know, we, we saw numerous cases in Winter Storm Yuri, where operators shut down their operations in order to, in order to provide power back to the grid. Yeah, you know, people will do the right thing, but I don't think there's a huge opportunity for them to say, "Hey, well, I can make, you know, a billion dollars a month running my data center, but I can do this kind of side arbitrage power deal, which is going to make me another 15 20 million, maybe. I mean, it's not impossible, but it is. And again, this is where the regulatory environment comes into play as well. So that's going to drive a lot of how new data centers are permitted, and we're going to see a lot of negotiation to ensure the operators are what I would describe as positive grid assets.

Nico Johnson  32:49

I think for most of the history of the grid we've assumed that supply follows demand. I'm curious to hear your perspective on a phrase that is used within Dell Technologies, how energy aware compute begins to change the relationship to supply, and how we think about that relationship with energy and technology.

David Holmes  33:12

The first thing to say is we're working really, really hard on what it means to run a modern data center, and by extension, a modern AI factory, and one of the conclusions that we've reached is pretty much everything we learned out of the last four decades of operating data centers isn't super helpful as we look forward, you know, this growth in power density from eight kilowatt racks eight years ago to 250 kilowatt racks today means that you have to think about architecturing data centers completely differently. It requires different cooling strategies, different power strategies, different types of equipment, and that's before we get into the different types of load and you know frequency stabilization and load volatility and dealing with all those things as well, Adele. One of the key things we've been focused on is maximizing the efficiency of these new generation of systems, looking at how we can maximize the number of electrons that are going to delivering useful compute by minimizing the amount of power used for cooling by moving to hybrid direct to chip liquid cooling strategies, our red or heat exchangers for residual power, and then our next generation of Vera Rubin systems, fully liquid cooled fanless systems, where we're constantly looking at how can we start maximizing the number of tokens per watt that we're generating from an AI factory. One

Nico Johnson  34:43

of the themes that's emerging from the FERC large load interconnection proceedings, and from the conversations in places like Texas, in particular, is that not all megawatts of demand are created equal. Let me hear your perspective on what separates a flexible load from a conventional load. And how flexible can these new data centers really be? What are the practical limits, and how are you all exploring that from a technology perspective?

David Holmes  35:09

So it's a really interesting question, and I've noticed that the definition of large load seems to be getting smaller and smaller, and so now it's down to about 25 megawatts, so whereas people used to think about large loads as being well, that's just for the hyper scale data center operators, now it's becoming closer and closer. So this idea of flexible loads essentially, how does a how does an AI data center become a grid asset? Well, it becomes a grid asset by having the capability to respond to grid conditions, and that means a couple of different things. Obviously, it means that when the grid is under stress, when it's under maximum load, that the data center can reduce its load on the grid. Now, in Texas, that becomes quite easy, like if you're a 75 megawatt load, then we're going to tell you, you need to figure out power for yourself, and off you go, but more broadly, it's really about being able to respond in real time to different load conditions, because in addition, there are times at which more power is available, I mean, you only have to look at the real time prices across ERCOT to know that there are certain times of day when the grid wishes more power could be consumed, and so what I think is really interesting, and why TerraFlow, I think, is a really interesting company to be looking at, is that they have a solution that allows the data center to continue operating at 100% of capacity whilst either completely reducing the load on the grid, reducing the load on the grid by 50% or even consuming significantly more power from the grid when it's, when it's available to essentially charge up these, these, these flow batteries, and I think that model combined with the software that companies like Emerald AI are producing is really going to help change the way in which data center operators, utilities, and grid operators start working together effectively.

Nico Johnson  37:13

We talked about this concept of stranded or underutilized energy resources, particularly. I know that there's, there's not gas resources you talked about interconnection nodes. How do we think about this, and where does renewable energy fit in there as well? Because these, there are tons of renewable resources, in particular in the Panhandle of Texas with wind, that do get curtailed. Anything, I guess, anything to add there that maybe we have an impact around this concept of stranded or underutilized energy assets. I

David Holmes  37:47

guess it's worth taking a second to kind of talk about the two major classes of loads that AI generates. There's AI training, which is 10s of 1000s of GPUs working on vast quantities of data to produce a model, and then there's inferencing, which is taking that model and figuring stuff out with the model. The users actually consume that inferencing component can run on relatively small scale infrastructure, even as small as kind of, you know, 15 kilowatt systems could provide a useful level of power for doing inferencing, but you know, a megawatt, two megawatts, four megawatts, a lot of companies are looking at how we can take advantage of available power in different locations and locate infrastructure that matches that, then aggregate all of that capacity together to deliver a much larger net aggregate outcome, and again it it removes the need to deploy new transmission distribution infrastructure, substations, transformers, and all that other hard infrastructure that's on super long lead times,

Nico Johnson  39:00

you've had a lot of experience with and conversations with energy executives across all different types of molecules and electrons, as you well put it. How is traditional generation evolving?

David Holmes  39:14

Well, I think I think traditional generators are, you know, first they were trying to figure out whether this was a kind of a short term glitch in the machine or whether it was going to be a long term sustainable thing. I think I think generators are going to take a while to catch up with demand. I think there's going to be a competition for generating capacity, and I think there's going to be a lot of innovation in generation as well. You know, Dell, we really see kind of an all of the above energy strategy. You know, nuclear is going to take a few years before it starts becoming a material contribute contributor to net new generation growth. You know, it's going to get there in the short term. Natural gas is going to provide a lot of the net new. Capacity used by AI data centers, but renewables have a critical role to play as well.

Nico Johnson  40:05

In three years' time, what will you look back on and say, 'I told you so? And they will look back and say, 'How do we miss that? or 'Why weren't we should have paid attention?

David Holmes  40:13

Well, I think I think distributed AI systems are going to be huge in three years' time. We'll see massive deployment of those. I think the level of innovation the AI can provide in supporting the energy ecosystem is going to be key. I think in the next three years we're going to see some significant evolution of the operational technology systems that the grid operates on, and that's going to make that's going to make a very big difference, and we're going to see people come up with creative solutions for solving some of these issues around linking AI data centers to the grid and the behind the meter assets.

Nico Johnson  40:58

If you had to give advice to a room full of energy executives on the opportunity right under their nose, be it developers or technologists, software service providers that are trying to help make these linkages. What would that advice be?

David Holmes  41:13

Well, there's lots of different types of energy companies, so that's

Nico Johnson  41:16

let's, okay, so let's focus on just our audience. Our audience is predominantly focused on decarbonization, deploying renewable energy as quickly and fast as quickly and widely as possible, and reducing our dependence on fossil fuel.

David Holmes  41:28

So, to renewable energy providers, I'd say make your load look like a base load as much as you can, right? You know, think about how the combination of renewable energy generation and long duration energy storage can help you present a solution to a data center operator that's going to enable them to be able to use large amounts of your power and think as well about how you're going to manage reporting on the carbon intensity of compute, which is an area we haven't touched upon, but for many companies is becoming really important, maybe not so much in the United States, but in the rest of the world. Understanding the carbon intensity of AI has become a really important talking point. So, the more that you can do to support data center operators in demonstrating how they're minimizing the carbon intensity of the systems they're operating, the better they'll be

Nico Johnson  42:28

as these worlds continue to converge. What do you think tomorrow's leaders are going to need to understand that wasn't nearly as important when you started your career?

David Holmes  42:36

I think everything is connected. I think you know when I started my, in my career in energy, you know, the, you know, energy companies were all unrelated to each other. At the Dell Technologies World Conference in Las Vegas, I brought together oil and gas companies, electric utility companies, service providers, and a group of people who 10 years ago I don't think would have had very much in common, but now have enormous amounts in common.

Nico Johnson  43:05

What's a book that you have gifted, read more than once, or recommended in the last year?

David Holmes  43:11

Do you know what? Actually, there's a really great book published by the IEA, maybe 20 years ago. It's called The Power to choose, and it's one of the first books that talked about demand response, and it was written at a time before so much that has happened had happened, and yet some of the principles and some of the exploration in that book are really relevant today.

Nico Johnson  43:36

Prescient, yeah, I have a copy of the book that Dan Yergin wrote in 1986 that basically created the modern day sort of media empire for Dan Jurgen as a as an energy theorist and I keep it on the shelf like visible with all the other books and people ask about it all the time and I'm like it's just to prove that actually in this book he said renewable energy was going to be inevitable and that if one of the if one of the industry spokespersons for molecules and traditional fossil fuel can believe that renewables can coexist, but yeah, then I don't understand why we have such a such distance, but there is, there is this that's fascinating. You're the first person to ever quote a book or a recommended book from the IEA. Touche, what's a habit or practice that has had an outsized impact on your career or personal life?

David Holmes  44:28

You know, I think the biggest thing is understanding your network and understanding how different people can relate to each other, and being able to bring people together, I'd say my superpower is being able to kind of build amazing teams and to bring amazing people together to achieve better outcomes, and it's a critical part of our work, because we, we work a lot across partnerships, and so that would be, that would be my one thing,

Nico Johnson  44:57

that's magic, I love that, yeah, I. And guest really early on, who said your network is your net worth, and I fully believe that. What's something you're learning right now?

David Holmes  45:08

Right now, I have this little interesting side project looking at how ARM and Intel-based systems work together with Nvidia GPUs to deliver particular types of outcomes, and it's a little side technology project that I'm interested in. I have quite a, quite a setup in my, in my house. You might remember that Gritty kind of went out of business in Winter Storm Uri, Octopus kind of like offered the same model, and I'm still on it, so I still buy my electricity wholesale,

Nico Johnson  45:44

right?

David Holmes  45:45

And so I do a lot of things in my house to optimize my power usage, depending on grid conditions. And now I'm working on a couple of little side projects to look at how I can kind of wire GPUs into that model and take advantage of it.

Nico Johnson  45:58

That is fascinating. Speaking of, do you know, have you heard of this company, Lectra, that will put basically an AI data center in your closet in your house and do fractional compute?

David Holmes  46:11

There are a couple of companies that are working on that. We're working with Span AI,

Nico Johnson  46:16

yeah,

David Holmes  46:16

on their recently announced solution, they're building extra.ai Wow, and I think you know it's, it's a really interesting area, and it has a lot of potential.

Nico Johnson  46:30

At the founder of lecture on, I have to get the folks from Span as well. If I look outside of my traditional renewable energy circles, as I have for this interview, Who should I interview next?

David Holmes  46:42

Well, I think you said you'd spoken to Mark Spieler already. I

Nico Johnson  46:45

have fascinating episode. You should totally mark,

David Holmes  46:49

definitely get hold of Arch Rowe. Yeah,

Nico Johnson  46:53

yeah, I had Arch on way before, back, gosh, 2018 Yeah, I should get ours back on.

David Holmes  47:03

You should, because if you look at how their vision has expanded out on the set of capabilities they're delivering, they're solving a lot of the really hard problems.

Nico Johnson  47:13

Yeah, yeah, I mean, that's one

David Holmes  47:16

of my ones to watch.

Nico Johnson  47:18

Good, brilliant mind. I'll hit you up for some more, David. I really am so appreciative. One of the ideas that will stick with me from this conversation is that the future is not going to be defined solely by generating more electricity, but by finding ways to tune how the world works dramatically smarter in how we use electricity, how the electricity and infrastructure sort of cohabitate, it has so for so long been like kind of like water to a faucet, as it were. Yeah, and and it's becoming this bi-directional force, whether we're talking about AI infrastructure, grid flexibility, energy, where compute, or the simple idea of maximizing every electron. As you introduced me to that thought, the common thread is intelligence. I want to thank you for sharing your perspective and taking the time out of what I know is a very busy schedule to come sit with us, and I hope to have you back to hear more about how Dell is transforming the sector. Thanks for taking the time for our SunCast.

David Holmes  48:26

I'd love to. Thank you very much.

Nico Johnson  48:28

Twice weekly, we deliver conversations with founders and leaders on the front lines of the clean energy transition, and we're here bringing this content to you each and every week, free of charge. Of course, it's not free - our sponsors help pay the bills and keep the lights on, so that we can help you build your legacy in the clean energy transition. And so, if you'd like to say thank you to them, or learn more about what it is that they bring into the world, or perhaps see how you could reach 1000s of listeners twice a week, just like they do, check it out at SunCast dot media forward slash sponsor. Remember, you are what you listen to. Thanks again for showing up, Solar Warrior. It's half the battle.

‍