Battery booms and the rise of flexibility [partner content]

This is partner content from the

Sean McAvoy has spent most of his career building AI software companies. Energy wasn't part of the plan. Then one day, while working at a company called Veritone, someone handed him an unexpected challenge. Somebody said, Hey Sean, what can you do with these twenty patents around battery technology and forecasting? And all they were was patents. It was like little software. So I took the patents and built a software business around those

So Sean did what software entrepreneurs do. He turned the patents into a platform. But as he quickly discovered, the power business, a maze of markets, machines and regulations, is not an easy one to master. Of all the businesses that I've been in and software companies I work for, the energy industry is highly complex. And so that's how I got into the energy industry.

But he and his team figured it out. And those patents became the foundation of an AI platform that sat between energy assets and energy markets, forecasting prices, controlling equipment, and automatically bidding electricity into the grid. It was built to answer a set of deceptively simple questions. How do I get the solar to work with the battery? How do I get the battery to work with the EV? How do I get solar and battery?

To be able to work with my manufacturing process. And so that like lends itself to: okay, we need more intelligence. While Sean was building that platform, an Irish company called Grid Beyond was developing another critical piece of grid flexibility. The company had emerged as a leading demand response provider in the UK and Ireland and was beginning to expand globally.

Grid Beyond was predominantly like a hardware company in terms of installing hardware control systems at industrial sites to uncover flexibility in energy systems at those industrial sites. So there was kind of like two bookends. One bookend was like controls that are on site at large industrials, and the other bookend was the ability

To put energy into energy markets and make money. And over time it kind of the whole middle piece between the two bookends became some very sophisticated AI-powered technology that is kind of like driving optimization. That middle layer was exactly what Sean's team had built. And in 2023, Grid Beyond acquired the Baritone Energy Business, bringing those capabilities together into a single platform.

we uncover flexibility. And once we find it, we ask the customer what do you want to do do with it, whether it's revenue generation, cost savings or resiliency slash reliability. The customers span heavy industry, metal plants, cement plants, pulp and paper mills. Increasingly, they include data centers.

Facilities that consume enormous amounts of electricity, but also contain systems that can be carefully orchestrated, cooling equipment, backup generators, batteries, and the computing workloads themselves. And Grid Beyond is increasingly being asked to orchestrate all of it. We can then inform the data center, hey, we need to get prepared already because on Thursday at four o'clock, we're predicting an event.

Let's turn on our optimization to reschedule jobs at that time. We're going to reduce our energy on the chiller pumps. We're going to trottle down like some of these chips.

We're gonna prime our batteries from our solar, we're gonna get our gen sets turned on twenty minutes so they're all warmed up and read ready to go. That is all done by software and it's all done by intelligence, by having models built in a digital twin uh environment such that you can orchestrate and control all of the moving parts. Making all of that work requires extremely fast decision making because when something happens on the grid, the response window can be tiny.

That's where AI comes in. And the advances happening right now are unlike anything Sean has seen in his years working in the field. How does this period compare to other periods of acceleration in the tech industry? Does this feel a lot different? Oh yeah. There's not there's nothing like this. Like the pace at which it's changing, the pace of the capabilities that are at your hands these days, like to to be able to develop like something new is I've never seen anything like it. Um

Even like five years ago, you were like, Let's pull all the data, let's send it up to the cloud, let's like have our applications in the cloud, make decisions. No, now you're doing it like right there at the local site level. And in some cases you have to do it because When you're responding to extreme events or any event or even a grid event, you have to do it in milliseconds. The goal right now, I think, for a lot of people is like sub 50 millisecond response. And so

You want to reduce latency everywhere. So you start doing your inferencing and decision making like locally. very locally on the ground, like at the site um level. And then once it's done its inferencing and decisioning, it then sends it back to the cloud so you can see what it just did.

Um and so like you're seeing like innovation around that that is allowing the responsiveness of all of these things like batteries or fuel cells, et cetera, you know, to be able to act before a blackout or brownout happens. This week, Stephen Lacey speaks with Sean Mekavoy, President and Chief Product Officer of Grid Beyond, about the frontiers of grid flexibility.

As the grid gets more complex, software is increasingly keeping it all in sync. Sean and Steven talk about how battery markets boom and then saturate. The promise and limits of data center flexibility, and how AI software is orchestrating everything from industrial loads to giant battery fleets.

So you have a unique window into how battery markets play out around the world, from the UK to Texas to Japan. When we think about how these markets evolve from opportunity to saturation, is there a consistent pattern you see repeating itself? Yeah, we do. I mean Because we've been like in Europe for quite some time, like fifteen years, we've already seen um saturation in uh say for example like the UK market.

with lots of batteries like come coming on online. And because you have the availability of uh batteries and they have a lot of capacity that can be put onto the grid, prices dropped uh because um scarcity wasn't as much of an issue anymore. But when the prices drop, it also makes it hard for those batteries to actually finance themselves because they were originally put in there probably when the bat when the prices were high.

um, those batteries, lithium batteries, you're talking anywhere from like fifty megawatts, hundred megawatts, two hundred megawatts of batteries, and you're talking hundreds of millions of dollars that has to be f financed and you're expecting to get paid back through the monetization of that battery capacity in the a energy market. But once it gets flooded, like then the grid determines there's hey, there's enough supply here. I don't have to offer

you know, uh my energy request at this price, I can drop that price. And so we saw that like in the UK um probably at least five to six years before we actually then saw it in Ericot in Tech, Texas. And I think the last really good year of battery pricing, for example, in Texas was probably around like twenty twenty three. And then it started to really drop in twenty four.

And then like we saw it further in twenty twenty five, there was a short kind of A short spike in prices when ERCOT brought out like a new service that was highly priced, ECRS, but that only lasted like a few, few months. And so we're seeing this in this cyclical part like in in Ericot, but it's it is s cyclical, like and I'll tell you like why it's what we see in Ericot now is like A lot of batteries came came online over the last like five, six, six, seven, seven years in AirCON. Like a lot.

And so when the grid operators looking at that, they're going, Well, you know, we have a lot of uh solar, we have a lot of wind in Texas, we now have a lot of batteries in Tech Texas. I'm not sure if we need to build any more gen generation. And it's not costing us as much money as it used to.

Because there's so much supply coming from these um batteries and and renewables that we don't need to build like new transmission lines and I don't need to build out like more generation. But all at the same time, demand is starting to go up. mainly because of what I spoke about earlier in terms of like the electrification journey that customers and industrials like are on.

And even data centers like coming online. So the demand keeps going up, but we're not doing a lot in terms of new generation because the batteries are flooding the market. Then all of a sudden demand is like, okay, it's outpacing what is currently on the grid in terms of batteries and renewables because people have stopped putting battery projects into Texas because the prices are not good enough. And if you then have an event where

AI is exploding and there's new data centers like being built, which is really driving demand up. Now you're kind of behind the demand curve. And so we see cyclically in another couple of years the prices will start to come back in ERCOT and people will start to make m money again. And so what we're also seeing is like, you know, when I said like, you know

Five to seven years ago, battery projects started to really explode in their cot. We're seeing this in Japan right now. We're at the very beginning of it, like in Japan, where their markets are now open up to batteries and the price of energy for battery is like sky high. Like in some areas it is like two hundred thousand dollars like per megawatt with a service called S uh FCR in Tokyo. Um And we're also seeing like another evolution happening in Australia where Australia has

They've encouraged and incentivized um rooftop solar for many, many years um throughout Australia. And then they started to cause problems on on the grid, like big frequency issues because of cloud cover, et cetera. Now Australia is going through an explosion of sub five megawatt bat batteries going under the grid. Not big ones, but small ones that are manageable in communities or in in industrial parks. So we're seeing an explosion kind of there as well.

And so it's all kind of cyclical in terms of, you know, it'll get saturated, it'll be too, too, too many, and then demand will come back like as it keeps growing over a number of years, and then you'll see the prices being more b beneficial to bat batteries. But yeah, we see this kind of cyclical Evolution kind of happening from the UK to you know the US to Japan and even Australia right now, which is fascinating to watch.

And so what are you doing during that saturation period of the cycle? Are you searching for different revenue streams to stack on within that market? How do you typically change? Yeah, so I mean there's two things. You have your your existing batteries that are in a market, and I mean physically located within the market, so you're not going to be able to move those batteries. But what happens is with saturation is that the spreads get really narrow between the high price and the low price.

And you really have to fine-tune your predictive models to make sure you're accurate down to the minute in terms of when the high prices are going to actually hit. You have to build models for shoulder seasons, for main season, summer, main season, like winter, because the market acts differently across.

seasonality mainly driven by uh weather. Gas prices is an another contributor, but weather is actually a big one. And you can see like from the different storms that we've had, Uri and I think like 21, Um, you know, even like this year, um we've had like some really uh crazy cold spells that are driving up kind of e energy demand.

And so like what you're really working on is getting your forecasting and your optimization really tight and really kind of fine-tuned because the spreads are small. And like, you know, what our salespeople do is like, you know. They'll start looking to see, okay, they always want to follow the the money. That's their job. Follow the money, find where projects are going to monetize greater. And usually it's kind of like, okay, well, where is

demand outpacing supply. And like if you look at like PJM and you see what's happening in PJM right now, the prices are going sky high, capacity prices in the In the base residual auctions are at an all-time high. And coupled with that, you have long term incentives that are at a local level, a state level. So like Illinois right now in the PJM.

uh market area has great incentives for um batteries. And I think we're also kind of over the hump of, you know, where we were at this stage in 2025, where we had a new administration come in, there was a new focus. on energy and and what that strategy should be. And the industry kind of stalls stalled out a little bit in twenty twenty five and then came back when you got more um certainty.

And now that it's back, we're seeing like, you know, new markets like PJM with the capacity prices being high. Also, MISO now is starting to come back with their own. higher capacity prices where it makes it attractive um for people to be able to put batteries in those markets. So we're seeing now a big shift in our sales pipeline in terms of battery projects. almost equal to what we have existing in ERCA, Texas. We're now seeing like the same in PJM.

How transformative do you think the expansion, rapid expansion of data centers will be for battery storage? It is definitely going to have an impact. It is not the silver bullet. There's actually no silver bullet right now, as far as I can see, that's going to help. data centers get online uh quicker unless you can get your hands on, you know, some big gas turbines, but it takes three years to build them out and there's a waiting list for three years to even get them.

But in the meantime, it's kinda like what do I have at my disposal? And we look at like what we call solutions around the meter. What solutioning can you do around the meter? to help out and provide supplemental power to these data centers if they have to sign an agreement with a utility. which enforces them to curtail at the peak times during the year, like just the peak times. And so then you're looking at a combination of okay, generators.

um, you know, whether gas or diesel generators coupled with batteries, coupled with potential like solar if you have enough land. And so we look at like solutioning like around the meter to help solve uh for that. for potential curtailment of day data centers so they can get online. And um batteries are part of that solution. And batteries are like evolving as well from not just like

one, two hour batteries. Now you can go like between four and eight hour batteries, depending on the chemistry of the battery that that you want to invest in. But long duration bat batteries are a good fit. for data centers, if they have to curtail for like 30 minutes, like or or or an hour, you can get like a four-hour battery system. Six hours.

battery and make sure you have um you have enough coverage. But it's not optimal. Like data centers need firm baseload power wherever they can get it. That is number one on their list. It'll always be number one because

Reliability is their number one concern for data centers. They want to have maintain high levels of uptime, your five nines, you know, that's what they're a that that's what they're aiming for, regardless of whether they're public or they're or they're private they data centers, uptime and reliability is key.

There is a an open question about curtailment for data centers and how much they're b be willing to curtail and Certainly I think if they can get faster interconnection, many of them would be willing to, but not for very long. Um what do you think the optimal curtailment is? How much are they willing to accept and do bigger batteries allow them to play a little bit more?

Yeah. I mean I spoke to quite a number of kind of data center developers, um, data center owners. Um, and basically they don't want to cur curtail full stop. I mean, they have fitted out their data center with the best infrastructure. They've paid a lot of money for GPUs and CPUs, extremely expensive. They want to maximize like those chip assets like as much as possible.

and curtailment for them is like something that they just do not want to do unless they are forced into it. And some of them will even consider like relocating or changing their plans in terms of where they site their data centers so they don't have to curtail. Senate Bill Six in Texas is kind of like an example of that where if you're kind of 70 megawatts and and over

you need to have the ability to uh cur curtail. And other states are adopting like similar measures as well. Um and and so like you can move where you don't have to have that. kind of uh restriction. But then you're still probably time to power is still an issue for you. Big batteries can can help. But thirty minutes like to an hour, I think is like it's you know, is something that you can you can work within in some terms of a data center.

um in order to curtail. You can get really smart and there's like some solutions out there that will be profiling all of the internal jobs within a data center that is being sent to kind of your CPU, GPU kind of platforms. And if events can be predicted, like people like us, we can predict an event that's on the grid that's gonna like cause a curtailment event.

You can then optimize and reorder your jobs going to your CPU and GPU. So your non-critical jobs are going to the CPU and GPU at that time because you're not gonna turn them off, you're gonna throttle them down and you're probably gonna trottle down all your chiller pumps at the same time who are also consuming uh energy. And so there is that kind of or if you're like, you know, even more sophisticated and you have

sister data centers. You can move jobs between data centers if you know in advance when a energy event is going to happen, that's going to warrant that you will have to um cur curtail. And there's a lot of different types of data centers. Does curtailment work for any of them better than others? Yeah, I mean we have all we have all types of customers. Um the ones that are the most flexible uh by far are the crypto data center centers. And there's lots of them in Texas.

And they're the ones that have like flexibility in terms of, well, if the prices are super high, uh, I don't need to curtail. I'm making more money like mining than I I I need to be participating in energy uh markets. But then when crypto is kind of like

you know, the prices fall, which we've kind of seen in the last couple of months, the cryptos are coming back and are saying, hey, but put me into the energy markets. And they can also avail of bat batteries as well. They can put like a lot a lot of batteries around their sites. And you know, they can continue to maintain power as well as participating in the energy markets. And that kind of like irks Texas like somewhat in terms of like, hmm, these crypto guys are double dipping.

you know, they're mining uh Bitcoin, but are also playing in the in the energy market because they have all these batteries like supplementing power. It gets harder like when you're a public data center and you have multiple tenants within your data center.

The only way that you can curtail is like you have to actually ask like each of your tenants or have an agreement with each of your tenants to see if like they can actually cur curtail. And that means like putting submetering into your data center. So there's a meter by Tenant and that meter can then be uh curtailed. But there's then data centers as well that are not what are called real-time data centers. They are doing

large language model training, they're doing batch processing. They're not real-time crit critical. They're not doing inferencing AI in those data centers. And you can curtail those day data center centers like very, uh, very easily. But when you're talking about like real-time processing and inferencing, where you're consuming large amount of energy to power these GPUs, um, you do not want to curtail those. And so you have to be, you know.

smart and like very strategic if you're a data center um developer. And we speak with a lot of them in terms of based on what you're doing in your data center, are you like a hyperscaler? Are you not? Like what tier are you? Um where you're going to put your data center center kind of determines like what you can actually do with it.

There's a lot of hope and enthusiasm among, you know, clean energy professionals and advocates that we can deploy as much solar and batteries together to serve data centers. Um, what what is the role of solar and storage together in serving these new loads relative to all the other resources that these data centers are gonna be utilizing?

Yeah. Well number one, it's like I go back to it, it's like they're not the they're not the answer. They are supplemental energy that can help and they're kind of filling a gap until either more jet jet generation comes online and depends on what form of that jet generation can be. Um everything from solar plus battery plus nuclear, for example, plus gas tur turbine. Um, but what we are also seeing is like

We are seeing data centers like, you know, and even retail energy providers who are coming to a utility hand in hand and saying, look, we want to put a data center here. I'm the retail energy provider. I'm going to provide this data center with with with uh energy. I need them to get an i in interconnection, but we're also willing to put batteries down in areas of congestion on your grid to alleviate that congestion on the grid.

And it may be kind of like close to where the data center is, but it's not like directly like on the land of the day data center. So you can use batteries to alleviate um congestion in certain areas and data centers are willing to pay for that.

I mean if they were going to pay for a battery to provide like supplemental like energy to the data data center, maybe it's better to go talk to your your utility and say, where would you like me to put this to Help you give me an interconnection faster by my batteries helping to decongest the grid. And this is like the bring your own capacity model.

your own capacity, more collaboration between data centers and utilities, data centers and the grid, data centers and what I call gen tailors, like you know, big retailers where there's an energy or a con constellation energy.

who can also generate their own energy, they can supply it to a data center, but they also want to help out the l local grid as well. Because the last thing that people want to see is that the cost of the energy for the data centers is getting passed down to the ratepayers.

That is like what would give everything a pretty bad name because it sounds then like that the ratepayers are paying for all of this, but they don't typically see the benefits of it, of a big AI data data center. So more collaboration between everyone.

And really what we have right now is like we have a deployment problem. When I was working like way back in whenever at Sun Microsystems and you know, we were doing the first like web applications, it was kind of like, how do we deploy these things at scale? You know, cloud was like, you know, starting and it was like, how do we deploy everything at scale? And that was our big problem that we had way back then.

But now, like we have the same pro problem. We have the solutions like to solve the problem, but how do we deploy it at scale in a fast enough time? To get these day data centers online because the US wants to win the race of AI. And there's other countries out there that are also.

Um going at it at a super fast pace. And right now, our problem is getting these data centers like online and you know, new generation being built and new transmission lines being built. So we have a deployment problem, which I saw a long time ago as well.

Yeah, I mean energy is all about trade-offs. So can you walk me through some of the benefits and trade-offs of other resources when you think about how these sites are being powered? Obviously a major push toward gas, but huge constraints in the supply chain. of enthusiasm about nuclear but long timelines. Uh same with geothermal. Wha how do you see the resource mix and what are you seeing customers pursuing? Like everything is being looked at again. Nothing is off the table right right now.

I mean if you you know if you were like two, three years ago, yes, there were things off the table, like sustainability was number one on everyone's list. It is in in the US it is not anymore. And so people are willing to look at like all types of of energy gen generation that can be leveraged to help solve like the problems and and like you know natural gas.

you know, that is now coming back and people are like, oh, firm base load power. If I can get a a connection, I'm gonna uh a gat gas connection if it's available and where I'm gonna site my day day data center. I'll be able to get online uh faster. Geothermal is now being looked at. Like you see, like some big sites around um geothermal. Texas is putting a huge Amount of effort into uh geothermal as well. Nuclear, you see some plants reopening that were dormant.

um and being retro fit fitted. So nuclear in terms of like what we had and closed down is now starting to r reopen. Not all of them will because they because they they they can't. And then you're looking at like you know, the future, maybe five to ten years from now, which is like these small modular uh reactors that can fit on a half the size of a football field, you know, but they could power like, you know, 10,000 homes.

Uh for for for example, but that's kind of out there and it has to be um tested. So everything is on the table right right now. There's a lot of there's a lot at stake in terms of the AI race and how you get these data centers on on the grid fast enough. And it's all about combinations. It's not just one thing. It's you're going to be looking at like every type of generator, turbine,

uh long duration battery, lithium, solid state, like you know, whatever's gonna give you the the mix you need to get up and running as fast as possible. But these are kind of like short term until newer gen generation is built and tr transmission lines improve. So you were going through this kind of like chaotic phase at the moment and it'll last for like probably like three years where Whatever you can put together to get yourself online, you are going to do it.

You mentioned that you have a lot of different types of data center customers. Are you seeing customers becoming more sophisticated because they're having they're being forced to think about energy in different ways?

Absolutely. I mean time is money in terms of like not getting these data centers like up and uh running and especially like, you know, even the likes of like Meta and you know, you have your Oracles of the World and your Microsofts and you know, everyone else who are trying to build build build build D these out. They're investing even in their own teams. Like they now have their own big, sophisticated energy management teams that we talk to and and we deal with.

And it is all about technology being implemented to manage energy. It's not so much like that there's somebody there flipping switches and gonna turn this on, that, that off. It is all kind of fully automated. It is all predictable. It is all based on models. It is all even like site load modeling in terms of like how the AI is consuming um energy and at what times of the day is there spikiness in the consumption of of these um chips.

So we d what we do a lot is we build digital twins of everything that we look at, whether it is an industrial site. or it is a data center, um, or it is a big battery and so solar plant, we will build a digital twin like off that plant so we can see data flow, energy flow everywhere in the plant.

I mean it's both a nerve wracking and really exciting period. I think if you really care about driving down emissions or you're worried about affordability, certainly a lot of people are are concerned about You know, the the increasing use of fossil fuels, uh, and particularly gas to power these data centers, but it's also unlocking just extraordinary innovation in contracting and

You talked about using distributed batteries under a just bring your own capacity model. How do you think that this period is going to influence the energy transition? Where are we? Energy transition is still h happening. It has gone through a couple of bumps. They're like speed bumps and you're forced to slow down. You know, and I spoke about early 2025 with brands being taken away, cancelled, and that's kind of like we'll see like a drop in in innovation. But filling that void is

what I call the pri privateers. So it is the likes of, you know, the big hyperscalers or these big companies, like whether it's NVIDIA, whether it's like Microsoft, whether it's Apple, et cetera, they are now I would say filling that gap in terms of creating their own um energy management systems, their teams, partnering with with people. We saw we can see consortiums today, even like around DC Flex, and the fact that

you know, you can now do so much more, so much faster with AI. Solutions are being developed at a rapid pace. Like my own engineering team, everything is built like AI native and the efficiency of the engineers has just gone through the roof. you know, in terms of like, you know, whatever kind of tool of choice that they're actually using to help like develop

software and code so much faster. It it's amazing. But you know, you hit a speed bump, but you will solve around it. And as the AI continues to get faster and better, those solutions are also coming quicker. Sean McAvoy, thank you so much. This was a great conversation. Thanks for having me. Uh it was a lot of fun. Sean McAvoy is the chief product officer and head of commercial for Grid Beyond North America.

GridBeyond develops AI software that helps businesses unlock flexibility in their energy systems, connecting assets like batteries, generators, and industrial loads, and connecting them to electricity markets. Learn more at gridbeyond.com.