TGEN - AMEX

Greetings. Welcome to the Tecogen Third Quarter 2025 Conference Call. [Operator Instructions] Please note, this conference is being recorded. I will now turn the conference over to Jack Whiting, General Counsel. Thank you, sir. You may begin.

Good morning. This is Jack Whiting, General Counsel and Secretary of Tecogen. This call is being recorded and will be archived on our website at tecogen.com. The press release regarding our third quarter 2025 earnings and the presentation provided this morning are available in the Investors section of our website. I'd like to direct your attention to our safe harbor statement included in our earnings press release and presentation. Various remarks that we make about the company's expectations, plans, and prospects constitute forward-looking statements for purposes of the safe harbor provisions under the Private Securities Litigation Reform Act of 1995. Actual results may differ materially from those indicated by forward-looking statements as a result of various factors, including those discussed in the company's most recent annual and quarterly reports on Forms 10-K and 10-Q under the caption Risk Factors filed with the Securities and Exchange Commission and available in the Investors section of our website under the heading SEC Filings. While we may elect to update forward-looking statements, we specifically disclaim any obligation to do so, so you should not rely on any forward-looking statements as representing our views as of any future date. During this call, we will refer to certain financial measures not prepared in accordance with generally accepted accounting principles or GAAP. A reconciliation of non-GAAP financial measures to the most directly comparable GAAP measures is provided in the press release regarding our Q3 2025 earnings and on our website. I will now turn the call over to Abinand Rangesh, Tecogen's CEO, who will provide an overview of third-quarter 2025 activity and results; and Roger Deschenes, Tecogen's CFO, who will provide additional information regarding Q3 2025 financial results. Abinand?

Thank you, Jack. Welcome to our Q3 2025 earnings presentation. In the last 3 months, we have seen significant forward momentum on our data center strategy. A year ago, when we started this pivot into data center cooling, most of the leads were from independent developers. Now the level of interest has ramped up substantially. We're getting interest from well-known colocation data center developers. We have now presented our solution to NVIDIA, AMD, and hyperscale developers. Across the board, the feedback has been positive. Given the level of interest, I'm now very confident that Tecogen will be successful in this market. During this call, I'll explain what some of the bigger developers are asking of us, the validation steps, and give some clarity on the path forward. The Vertiv relationship is a key part of this strategy, so I'll explain how this fits in as well. Our initial leads in data centers were from independent developers. For example, companies in conventional real estate pivoting into data centers or former data center executives who had started their own companies. These developers were initially more open to new ideas, such as our chillers. One of these developers gave us the LOI for 6 STX chillers I mentioned last quarter. Over the last 3 months, the developer has come to visit Tecogen, see our customer sites, and as a result, is now looking to include us on 3 of their projects as part of their main AI cooling work. This could mean more sales of many more chillers than contemplated by the LOI. The total combined IT capacity in the initial phase of the build-out is likely to exceed 200 megawatts and substantially more than that over time. The developer is in active discussions with potential tenants. And given the power constraints across the country, we believe this developer is likely to be successful. The other projects I mentioned last quarter are also at various stages of either obtaining financing or tenants, but are moving along. What is more exciting is that we have started to attract the attention of big-name developers who are well-established in the industry, including those with multiple AI data centers already constructed or in construction. As we presented and listened to developers, hyperscalers, and chip companies, it has become clear that the current methods of cooling a data center create numerous problems that we can solve. For those of you who are new shareholders, I'd like to reiterate the value proposition that Tecogen offers to data centers. As chips have become more powerful, they need more cooling, and cooling systems must be designed for the worst case. So how to stay with full AI load? In some parts of the country, this can be 120 degrees Fahrenheit or more. When you design for the peak, you tie up a lot of power since you don't know when you'll need to turn on the cooling system. In the past, this wasn't a big problem, but with the latest AI chips, electric cooling could require -- consume 25% to 35% of a data center's total power, and the power requirements are only increasing. If you move all for part of the cooling to natural gas, data centers have more power for IT, increasing their potential revenue. I know that many of you may have heard of data center cooling technologies such as liquid cooling, immersion cooling, et cetera. All of these technologies are targeted at what happens inside the data centers. All these technologies still connect to chillers like ours to reject the heat, which today are powered by electric chillers requiring substantial amounts of power. Our chillers can interface with any of these liquid cooling or immersion cooling options. The graph on the left shows how much power allocated to cooling nearly doubles based on the hottest design day. This means that in Texas, a 100-megawatt data center today is allocating 35 megawatts to cooling. There's also a secondary problem for some of the bigger data centers. One of the reasons we've started to get interest from hyperscale developers is that some of them are using on-site power generation. They tell us that due to their cooling requirements, they add additional power generation capacity that sits idle for much of the year. Because gas turbines are in short supply, data center developers would rather redeploy these idle gas turbines at the next data center. These developers have also told us that the cost of adding our chillers is less than half the cost of adding an equivalent amount of power generation. The reason for this is because significant additional infrastructure is required to support gas turbines, whereas our chillers can be a direct replacement for electric chillers, reducing the need for electric power for cooling. Most data center developers we have spoken to are considering using our chillers for 30% to 50% of the data center cooling. Our chillers will operate on natural gas above a certain temperature threshold, so the data center can cap the power allocated for cooling. In order for these established developers to use our chillers, there is a validation process. This requires providing test data, computer modeling of performance at various parameters, and other information. Although many of these developers understand that we are a smaller company and will need time to ramp up, manufacturing capacity is a key parameter that we need to satisfy. To address manufacturing capacity, we have already been making factory layout changes to increase throughput. We have begun working with contract manufacturers and expect to get the first articles for sheet metal assemblies for the dual power source chiller before year-end. However, this is only one pathway to manufacturing capacity. Concurrently, we have been working with Vertiv as a secondary pathway for increasing manufacturing. I understand that shareholders may be concerned that the Vertiv Tecogen relationship has been slow to date. This has changed substantially in the last month. Vertiv has tasked the head of their U.S. chilled water Group to lead the partnership. As a result, we have seen significant forward momentum. Although I can't get into specifics, we are working on multiple avenues to jointly sell and scale up our natural gas solutions to these larger developers and satisfy their validation requirements. Based on conversations with these larger developers, the AI chip companies, and hyperscalers, I'm now confident on the scale of the opportunity and the value our chillers bring. Many of these developers need hundreds of chillers a year, and all of them have confirmed the benefits. Our current backlog is approximately $4 million and is predominantly cannabis cultivation and the Las Vegas Convention Center 10-year service contract. We're expecting some multifamily projects and some other projects in cannabis to close later this quarter or early next year. But given that we are now getting interest from some of the well-known developers in data centers, our focus needs to go on securing initial projects from them. If we can do that, more projects from other developers will follow suit. From there, there are multiple strategic options to turn our opportunities in the data center market into value for Tecogen shareholders. As a technology company of our size, we also don't need the broader AI market to grow at billions of dollars a year to generate value for our shareholders. A single 200-megawatt data center uses 100 to 200 electric chillers. To put that into context, the Las Vegas Convention Center order was just 7 of our bigger chillers. We're working on technology improvements that provide test data and engineering support and building chiller inventory for potential data center projects. We've also made several improvements to our engine platform so that we can double our service intervals and provide better performance and higher engine time. To get essential data for continued product improvement and to reduce the impact of higher labor costs in certain territories like New York City, we invested $700,000 into new engines for the service fleet. This disproportionately reduced our service margin because we expense engines, but will have a beneficial effect on profitability medium and long term. It also provided much-needed data for continued R&D. Our current cash position is approximately $14 million, but we're expecting to collect $2.5 million in the next few weeks. We also repaid the related party note, so we have no debt on the balance sheet. I'll now hand over to Roger to take us through the financials.

Thank you, Abinand, and good morning. Our third quarter results, total revenues increased $1.6 million in the third quarter to $7.2 million, which compares to $5.6 million in the third quarter of 2024. And this is due entirely to the 115% increase in the products revenue during this period. Our net loss increased in the third quarter to $2.13 million, which compares to $0.93 million in the third quarter of 2024. And this is due to a decrease in our service margin resulting from increased material and labor costs incurred, as Abinand explained earlier, as we invested capital in engine replacements. And in addition, our operating expenses increased during the most recent quarter. Our gross profit decreased 12% due to increased costs incurred in our Services segment. Gross margin for the third quarter decreased 13.7% to 30.4% from 44.1% in 2024. We will discuss the gross margin further in the segment performance slide. Our operating expenses increased just under 28% quarter-over-quarter to $4.28 million from $3.35 million, and this is due to increases in administrative and R&D payroll and increased benefits, recruitment costs, and general increases in our business insurance premiums, depreciation, stock-based compensation, and higher sales commissions. Moving on to the EBITDA and adjusted EBITDA. For the third quarter, our EBITDA loss was $1.94 million, and the adjusted EBITDA loss was $1.7 million, which compares to an EBITDA loss of $0.77 million and an adjusted EBITDA loss of $0.75 million in the third quarter of 2024. And the increases in both the EBITDA and adjusted EBITDA losses in the current period are due to the decreased gross margins in the Services segment and our higher operating costs. Moving next to performance by segment. Our product revenues increased in the third quarter to $2.98 million from $1.39 million in 2024, which is due to increases in chiller and engineered accessory shipments, and it's offset by a decrease in cogeneration shipments. We delivered an additional hybrid drive air-cooled chiller in the current quarter. Our products margin decreased to 36.8% quarter-over-quarter from 42.7% in the comparable period in 2024, and this is due to higher material and labor costs, and also the RA cool chillers are sold at a slightly lower margin due to their initial shipments of the product. The services revenue increased 2.4% quarter-over-quarter to $3.94 million in the third quarter of 2025 from $3.85 million in the comparable period in 2024. The gross profit margin decreased 19.1% to 25.3% from 44.4%, and this is due to the increased labor and material costs in our New York territory, I should say, New York City, as we invested capital in engine replacements. Our energy production revenue decreased by 34.2% quarter-over-quarter to $25.60 in 2025 compared to $389,000 in the comparable period of 2024. And this is due, as it was in the past quarter, the expiration of contracts at certain sites late in 2024, and also to the temporary shutdown of a couple of sites for repairs during the current period. Gross margin decreased 10.8% to 34.4% in 2025 compared to 45.2% in 2024, and this is due to the temporary shutdown of the sites we previously mentioned and the additional costs we've incurred to bring the sites back online. Overall, gross profit margin decreased 13.7% quarter-over-quarter to 30.4% from 41.1%. And again, this is due to a reduction in our services' gross margins. This concludes our review of the third quarter 2025 financials. And I'll now turn the call back over to Abinand.

Thank you, Roger. I'd like to summarize by saying that as a technology company of our size, we don't need the broader AI market to grow at hundreds of billions of dollars to generate tremendous value for our shareholders. To put the opportunity into context, a single 200-megawatt data center uses 100 to 200 electric chillers. The Las Vegas Convention Center, as I mentioned earlier, was only 7 of our chillers. Therefore, as a company, we need to stay focused on satisfying the needs of data centers, including delivery and performance. And from here, we can convert this into value for shareholders, either through scale-up of manufacturing or other strategic options. Given the level of interest that we're seeing from some of the biggest names in the industry, my confidence level for our strategy continues to increase. I look forward to updating investors on progress. I'll open the floor for questions.

[Operator Instructions] Our first question is from Chip Moore with ROTH Capital Partners.

I wanted to maybe ask first on the initial pilot for the 6 units. Maybe just any update on how you're thinking about potential timing for that? And then the follow-on, maybe expand on -- it sounds like that opportunity is growing. Are these additional sites in planning? Or how to think about those additional opportunities, as well as I think you mentioned they're interested in doing a greater portion of their load. Just any sense of scale there as well.

Yes. So that's a great question, Jeff. So it's -- the way we are seeing it, all 3 projects are at around the same stage in terms of planning because there -- that developer is trying to essentially get a tenant for all 3 about the same time. What seems to be the case is that some of the larger entities that are willing to lease these kind of data centers would try to take up all the capacity in one shop. So it's possible that a portion might be leased separately. So the timing could be very minute, or it might take a few months, depending on how long it takes them to get a tenant. It might even take -- it's because a lot of those things are outside our control, I would say that from what we understand, the developer is in very active discussions with tenants right now. We have been included as part of the engineering design stage. So our odds are very good, but the tenants do have a say in what gets included as part of the project. That's really where the Vertiv relationship will come in handy, because if you end up with a hyperscale tenant and you need an approved vendor type relationship, Vertiv is an approved vendor for most big companies. So we can always sell through there. So there's a few nuances that comes to that. But having said that, that's not the only opportunity that we've got on the table right now. There's multiple opportunities today that have come up. I can't speak about specifics on a number of them because we're under NDA with a number of companies. But I think this is just one of the many that are moving in sort of parallel paths right now.

And to your point on Vertiv, it sounds like things are progressing at a faster clip now. Maybe just any more you can give us on that? And then just around validation and test data that you mentioned, obviously, you're a smaller company, just a sense of what you need and timing there on some of the more detailed stuff.

Sure. So let me start with the Vertiv side of things. So I think that one of the reasons that the Vertiv relationship was a little slow to get going was because we were either dealing at too high a level in terms of seniority at Vertiv or at more of a junior level. So what Vertiv ended up doing was restructuring who our point of contact was in their company and who was going to lead the effort forward. So now we have their Head of U.S. chiller operations, who has both the authority and the ability to move this forward quickly. So that started accelerating things now. The other thing that we are working on is we're starting to see interest directly from large-scale data centers, where for both companies, there'll be a big benefit in terms of this relationship is really also if Vertiv can help us scale up supply chain manufacturing. So a lot of our discussions have also been around that area because -- and of course, providing a way for us to -- or for end customers to feel more comfortable working with a smaller company like us. So it's the nature of the relationship is not purely just Vertiv is going to do the marketing and we sell to them. Now there's multiple other avenues that are being worked on right now. With regards to the second question on validation, so part of that -- and that's one of the reasons why, as we're starting to see this real level of interest, as a company, we're focusing significantly on that, whatever it takes to provide the data. So we've got units in our test cells here. We -- our engineering team, we've expanded that a little bit to really be able to support some of this effort. And then we have -- we just run test data. I mean we have some of that data as a standard just because every -- even in other industries, the same, some of that requirement is the same. But with data centers, there's -- it's a little more expanded because they operate in different conditions. And depending on which part of the country they're in, it varies. So we have to provide a lot of that. And that -- part of that is we've got units in test cells in our factory, just running that data as needed as people ask for it.

And on the manufacturing side, Abinand, I thought I heard you say you're working on the contract manufacturing piece. Was that on the dual-source power unit? And how are you approaching the contract manufacturing side?

So we believe both from a margin standpoint as well as scalability standpoint, the pieces that are better outsourced are things like the -- especially on the air-cooled chiller, where there's a lot of sheet metal assemblies associated with that. So we're going to look to outsource that sheet metal assembly and then have the -- do the power train and the final assembly in our factory, test it, ship it. So that sheet metal portion and all of the refrigeration systems associated with that sort of the piping, we're working with a company that does a lot of overflow capacity for some of the bigger chiller manufacturers. So they have significant ability to scale up. But what we need to do is to get the first article, validate that, which we're expecting to have the first article in the next month, 1.5 months or so, so that we can verify that it meets our design and it fits with all the other pieces that we will assemble in our factory. Once that happens, scaling that up is relatively straightforward because you're just using that same -- you're using the same plant. So that is an avenue for us to really improve throughput out of our existing factory. The other piece we have done in our existing factory is to have a more flexible layout. So it doesn't matter which chiller variety we get an order for, we can respond to that.

And if I could ask one more, just on service margins and the new engines you introduced, just how to think about that, maybe more near term, when we start to see some benefits. And obviously, I imagine that should help you as you get some bigger deployments in data center, but just strategically, what that could do as well?

Yes. So if we look back at service margin over the last year, 1.5 years or so, they've sort of been inconsistent. There have been some quarters where we've had them right around the 50% mark, and some points have been in the sort of low 40s or so. And as we mentioned last quarter, the cost of operating in New York, in particular, has gone way up, especially travel time between sites. So when you look at the service like a single engine in, let's say, New York, the -- some of your biggest costs, right, tend to be the engine. If you can basically avoid any engine work, or you can avoid having -- if you can double your oil intervals, then you don't have to go back to a site anywhere near as often. And you can do this -- you can -- the delta in putting a new engine with all the approved -- like the updated systems versus maybe making some repairs or making small increments, that's -- it's relatively small. But if you could double your engine life or even increase it by 25%, 30%, that will disproportionately improve your margin. So I would say in the next couple of quarters, we're not yet sure where the service margin is going to land, just because we might choose, depending on how the initial results from this and how we're seeing things start to look. We remotely monitor every unit. So just looking at the data, looking at short-term data and saying, okay, everything is working really well. We're seeing this decline in terms of having to actually go to sites as often, then we might roll out more across the fleet. Because, as I mentioned on the engine side of things, it's a combination of not only improving the actual engine system, but it's also things like oil change intervals, where you can almost -- you can increase it 50% or even double it, which is what we've seen in the initial sites. If we can see this across this broader range of units, then we might apply it to more units because, again, it will help us get to that 50% or higher gross profit margin much sooner if we do that. So for the next 2 quarters, it's a little hard to predict until we see our own data from that. But past that point, I would say this is -- this will definitely improve overall margins on service and get us to the point where we're generating healthy cash from that, and we can really focus on growing the data center business instead.

Maybe -- sorry, one last one. Just your tone and confidence sounds like it's picked up. And obviously, getting in front of names like NVIDIA and AMD and hyperscalers quite impressive. Just any more you can expand on that feedback and what you're hearing from some of those bigger names and receptivity?

So again, I can give you broad information here. A lot of these ones, they're -- especially the bigger names, as we look at the bigger developers, we have NDAs with them. But more broadly, I think across the industry, the things that we're seeing are the power constraints are getting substantial. I mean, people are finding that utility power is, I wouldn't say nonexistent, but getting the full utility power is becoming more and more difficult. The cooling load is definitely getting higher with the current generation of chips, right, especially when you're designing data centers down in Texas or Virginia, where your hottest day because you're actually not designing just for the hottest day in 1 year, you're designing it for the hottest day in the last 20 years. So that delta, the amount of power needed, is so much higher. And as I mentioned, the feedback on the alternative, which is, okay, what if you do on-site power generation? On-site power generation is actually -- this is where I feel like our assumptions have been validated in terms of the cost delta. We're definitely getting told across the board that on-site power generation is substantially more expensive than choosing our chiller option. I think the real hurdles for us as a company to get traction in this space is going to be to figure out ways to derisk somebody choosing a natural gas chiller, right? Because you -- and that's really where that dual power source chiller is particularly attractive, because you've got 2 power sources. So you can max that chiller up with a generator if you needed to, you can run on natural gas, you can choose to switch over to natural gas part of the time. You don't -- you have many different ways to run that. But it's still going to be -- we have to work our way through these steps. But we're not getting any pushback in terms of the underlying value proposition, the cost benefits. It's really a matter at this point of figuring out ways to get projects, even if it's smaller projects with some of these bigger names, to really get people comfortable with the technology. And then from there, I think it will grow.

Our next question is from Alexander Blanton with Clear Harbor Asset Management.

I'd like you to give us an idea of what the dollar volume would be of the example that you gave, a data center with 200 chillers.

So I'm going to give you broad ballpark numbers here, right? So the -- let's just say out of the 200, about half of them went to our type of chiller, right? That would be anywhere from $30 million to $50 or more, depending on how big each of those.

$30 million to $50 million for 100 chillers?

Well, it depends on how -- so there's a lot of pieces here, right? You've got the size of the chiller, the I guess, to a certain extent, Alex, I'm giving you a very broad range here just purely because there are other people that sell this and each -- it depends on -- our pricing is a little different depending on whether it's the dual power source or whether it's a standard DTX chiller or so.

Why would you -- why did you say 1/2 of them would be yours, who would have the other half?

So typically, at least based on the early conversations that we've had with some of the bigger developers, they would typically keep standard electric chillers for part of it and use us for part of it. Some of that may be just having dual supply chains for any data center. I think in many cases, having splitting those kind of systems seems to be standard practice. But again, there's a lot of moving pieces when it comes to those bigger projects.

Well, if the savings using the gas chiller is so great, why would they have any electric chillers?

So I think over time, and this is really part of the earlier comment that I made. I think the issue on right now is very little to do with the benefits of the product. I think a lot of it is really comfort level because we're doing something different in a data center, right? Even though we might have these chillers in many critical cooling applications like hospitals and ice shrinks, this is still a new industry for us. So I think over time, you might see the full system go this way, but--

Okay. So it's a matter of having confidence in the company and the product.

I think so. I think that's really the -- at least in terms of the feedback and the ways that people are looking at it, doing a portion of the AI load initially or a portion of it or using it for things like turbine cooling, which again allows people to try it out without taking too much risk initially. And then if things work well, then I think you'll start seeing much bigger portion of the load that move over to systems like ours.

Just to shift topics for a minute. What is the status of the renovation of the Las Vegas Convention Center with your chillers?

So we have shipped our chillers to the convention center. I would estimate that, that site will probably come online early next year. So they're in construction. They're installing our chillers right now. And I would say that, yes, we're expecting that site to start up sometime early next year.

And what percentage completion is it at the moment? You shipped some, but not complete, right?

No, the chillers -- we've shipped the chillers to them now. So that project is complete. The bit that isn't complete is the service contract, which will, of course, carry on for -- like it's a prepaid contract. So we'll recognize that over the next 10 years.

Okay. So you recognize the revenue from those products?

Correct. Correct.

And as to the -- going back to the data centers, what would you estimate would be the timing of the orders in that space?

That unfortunately, is the hardest piece to predict. At this stage, I cannot make even reads. It's the one that we have an LOI for, that could move very quickly as soon as they get a tenant, right? That would be that -- but the other projects there, a lot of the ones from the independent developers that I initially -- like we started working with, those projects, they're hoping to be online by 2027, which means they'll have to take deliveries in 2026. But with a lot of those entities, again, there's pieces that are outside our control, like lining up the tenants. With regards to the bigger-name developers, it's really a matter of how long it takes to get through the validation process, what it takes because they might start using chillers earlier on. It might be smaller chunks just to try it out. But they don't have tenant problems. They don't have financing issues. So those things might move much faster. I think this is one of those cases where we'll -- I'll try to keep people updated between now and when we report next as soon as we have any traction on any of these projects in terms of getting those projects over the line.

Now these are projects that haven't started yet. What kind of an opportunity is there for you to retrofit or to participate in the expansion of existing centers? I mean you're talking to larger hyperscalers or people in the business that are already operating these centers. Do you have an opportunity to sell anything to them?

So in terms of retrofits, we're not seeing as much -- we've seen a few projects in that in terms of retrofits. But majority of the ones right now are being built ground up for the latest chips. A lot of the retrofit opportunities are with the previous generation of chips that -- I think the industry as a whole, we're not seeing -- at least from what we've seen there, we're not seeing as much activity in the retrofit market. But I think that is going to change pretty soon, given the power constraints. In the very small-sized data centers, we might see some retrofit opportunities. Some of the cloud-type data centers that are being converted to have some AI component or some computing component, that might happen. But it's -- in terms of the bigger AI data centers right now, it tends to be new builds, built ground up for the latest chips with liquid cooling and all of the other pieces associated with it.

Our next question is from Barry Hymes with Sage Asset Management.

I had 2 questions. One, just back on the Vertiv relationship. When -- how long ago was the change in that point of contact? And are you generally making joint sales calls with them? Or did you have to do teach-ins for their sales force? Or just kind of how is that process working? And if you do get an order through them, do they do the service, or do you do the service? And then second question, just quick, could you remind us what percent cost savings on average, or what's the range that you would typically quote on your system versus traditional?

So with regards to the Vertiv relationship, I'd say the change in point of contact was probably about 3 to 4 weeks ago. So it's pretty recent. Although we've been dealing at the higher levels with Vertiv since the agreement was signed, it was just trying to get things to move faster on their side and make sure that there was authority to really drive this relationship forward. So that change happened more, yes, 3 to 4 weeks ago. We have done teach-ins for their sales force already. We have done a few joint sales calls, but a lot of them have been more in terms of identifying what are the key pain points that data centers are identifying beyond purely like what is the power that's freed up there. Other things like, for example, one of those calls with the stakeholders, identify things like uninterrupted cooling. So for example, in an electric chiller, if you lose power, what needs to happen is the electric chiller will shut down, diesel generator will come on and then you will turn on the -- like the chiller will come back online. With natural gas, you could potentially keep running through an outage. You just move some of the pumps onto backup power, and you could run through an outage. So there are benefits like that, that are not necessarily initially obvious unless you talk to end customers. So those are the kind of things that we've done with the Vertiv sales team to really identify what those pieces are. Now it will start to ramp like -- and Vertiv does -- has been doing some initial quoting to potential customers. So -- but exactly where those projects are, some of that is not as clear to us. Most of these bigger opportunities that we're talking about today came from Tecogen's direct marketing. So us reaching out directly or being big data centers at trade shows, or yes, direct outreach to some of them. So I think the Vertiv side of things is spinning up. I think the right things are happening. But that change, I think the acceleration really started, I think, in the last month or so. With regards to your question regarding savings, really, you have to think about it in terms of power available, right? So when you've allocated power to cooling, that power is not available for those IT chips anymore. So it's really a reduction in revenue for data center. So let's say, 30 megawatts goes to cooling, that -- today's numbers, like if you're a colocation data center owner that's essentially having a hyperscale tenant there, that hyperscale tenant is paying anything from $150 to $200 per kilowatt per month. So for every additional megawatt, you're talking more than $2 million of additional revenue a data center can make if you're a colocation data center owner. If you're a hyperscaler, right, it's the difference between having your data center that's 70 megawatts versus -- or having 70 megawatts of computing versus having 100 megawatts of computing available. So it's really what it allows you to do in that data center. How much additional computing you can have on-site? So it's really an increase in revenue rather than purely a reduction in operating cost, which there is a reduction in operating cost, but that's actually small compared to the increase in revenue.