EROC (EROC) Q2 2026 Earnings Report, Transcript and Summary

Greetings, and welcome to ERock's Second Quarter 2026 Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to introduce your host, Ted Durbin, SVP of Finance. Thank you. Please go ahead.

Thank you, operator. Hello, and thanks for joining ERock's Second Quarter 2026 Earnings Call. I'm joined by John Carrington, CEO; Ian Blakely, CFO; and Corey Amthor, President. In a minute, I will turn the call over to John and Ian for their opening comments, after which we will have a question-and-answer session. Please note that we have also published a supplemental earnings presentation on our Investor Relations website. Later today, a replay of this call will be available on our Investor Relations site. As a reminder, we will provide forward-looking statements during this conference call. These statements are not guarantees of future performance and involve a number of risks and assumptions. Please review our SEC filings for a list of actual risk factors that may cause our results to differ materially from those presented during this conference call. We will also discuss non-GAAP measures during this conference call. Reconciliations of adjusted EBITDA and certain other non-GAAP measures to their closest GAAP counterparts can be found in our earnings release and presentation available on our website. With that, let me turn the call over to John.

Thank you, Ted, and thank you all for joining us on our inaugural earnings call covering our second quarter 2026 results and 2026 guidance. We executed well during the quarter with our backlog up 10x year-over-year to a record $1.7 billion. We contracted with Anthropic for 470 megawatts of generation capacity for a new data center. We commenced construction on the 366 megawatt El Paso Electric project supporting Meta's data center campus in El Paso. We began operations at our Hyperion facility here in Houston, which will expand our assembly capacity to 1.2 gigawatts by year-end. And finally, we priced our IPO, raising approximately $400 million for the company. Having just gone public in June, I'd like to spend a few minutes reviewing our history, the robust market backdrop and the compelling solutions we have purpose-built for this environment. I will then discuss our commercial and operational progress during the quarter. Ian will follow with additional detail on our financial results for the quarter and our outlook for the balance of the year. ERock provides vertically integrated power systems to our customers. We design, install and operate distributed power systems for commercial and industrial, data center and utility customers across the country. We have a 15-year operating history of delivering 99.999% of reliability across over 400 operational sites, which represents more than 1 gigawatt of installed capacity. This long-term track record of high reliability sets ERock apart from most of its competitors. As many of you know, the market backdrop for our solutions is strong and growing. AI-driven demand has pushed load growth to the highest levels in over 50 years, which combined with an aging grid is creating tremendous need for additional generation capacity. Supply chain constraints have extended gas turbine lead times to over 4 years. Additionally, interconnection queues can extend to over 7 years. Together, these trends have accelerated speed to power needs for hyperscalers, utilities and C&I customers. And our contracted power system sales backlog reflects that, having grown to approximately $1.7 billion, which is up roughly 10x year-over-year or an increase of $400 million quarter-over-quarter. ERock was built for this environment. We provide solutions that help customers meet growing power demand safely, reliably and cost effectively. For many customers, addressing these needs independently can be challenging as they may not have the dedicated in-house energy expertise or the resources to manage permitting, gas supply, construction, commissioning, utility interconnection and ongoing optimization across multiple stakeholders. As we are seeing the regulatory environment evolve to address the intersecting needs of power, grid infrastructure improvements and community impact, ERock is ideally positioned to meet the emerging bring your own power model. In order to allow our customers to bring their own power quickly and efficiently, we provide a turnkey solution. Our vertical integration means that we design, assemble, install, operate and monitor our distributed power systems. Through our proprietary Granite software, we optimize the value of our customers' assets. Once our equipment is operating, we provide long-term services over the life of the system, typically contracted for 5 to 15 years and generating high-margin recurring and predictable revenue. A key metric we track is repeat business and renewal rate. Our services renewal rate is running at 100%, clearly demonstrating the value our customers see in our O&M services. There is also a positive feedback loop as the more systems we deploy in the field, the faster we learn, the more real-world operating data we collect, which feeds back into our design and installation process. I'd like to share some comments on our differentiated technology platform. At the core of the platform is RockBlock, our proprietary natural gas generator. Traditionally, high emission diesel generators have been the only option available to meet fast start and transient response needs required for large systems. With our proprietary design, our RockBlocks have been engineered to outperform other reciprocating natural gas engines meeting fast-start and transient response attributes, but with reduced emissions, far quieter operations and no water requirements. Regarding fast-start, our generator can take full electrical load in as little as 10 seconds compared with competitive natural gas solutions that may take up to 45 seconds or longer to fully ramp. Given the differential in start times, our competitors' units cannot be used in place of traditional diesel backup. The multiple use cases of our RockBlock lowers the total cost of ownership while supporting customer reliability requirements. RockBlock's proprietary technology is also designed for strong transient response, enabling rapid ramping to match the variable load profiles of AI data centers, both across training and inference workloads. Our ability to maintain stable voltage and frequency through significant load swings is a key point of differentiation versus competing generation options, including turbines, fuel cells, renewables and other gas-based solutions. Competing technologies must often use batteries for smoothing, while our system need fewer or no batteries because of our superior transient response profile. From an environmental and deployment perspective, RockBlock is the first natural gas engine on the market that meets the California CARB DG standard, one of the most stringent distributed generation emission standards in the country and is up to 99% cleaner than diesel with no water required for operation. Its lower noise profile also supports deployment closer to load centers and communities. RockBlock is modular with building blocks as small as 500 kilowatts and its patented design reduces footprint by approximately 50% compared with similar natural gas units. That modularity allows us to tailor systems to a customer-specific load profile, building redundancy and deploy more quickly and repeatably than larger single block alternatives. RockBlock have lower emissions, are quieter, require smaller footprints and use no water, all strong competitive differentiations that matter to our customers. What further differentiates ERock is the integration of hardware, proprietary software and long-term services under one accountable operating model. Each RockBlock is delivered with our embedded Granite software platform, enabling sub-second data capture from the engine and controls. That data supports predictive diagnostics, site-specific performance optimization and maintenance planning based on actual operating conditions rather than fixed service intervals. Granite enables us to monitor and control thousands of systems remotely 24/7 from our network operations center. Customers that procure a conventional gas generator often manage separate supplier relationships across installation, maintenance and service response. With ERock, those responsibilities remain integrated within one platform and one accountable partner, eliminating the handoffs that can create complexity and execution risk. We assemble RockBlock systems in-house, which protects our intellectual property, gives us greater control over supply chain execution and supports a capital-light operating model. Because our systems are pre-kitted and modular out of the assembly facility, installation requires less specialized labor and heavy equipment than many competing technologies, improving deployment speed and repeatability for our customers. This unique technology platform results in a single installed asset that is capable of serving multiple use cases. First, it can serve as a bridge power solution for customers waiting to connect to the grid. Our systems run 24/7 at 99.999% reliability as prime power until grid interconnection is established. Second, when permanent utility service is available, the same asset can be used for backup power, protecting operations from outages. And third, enabled by our proprietary Granite software, it can also serve as a source of dispatchable capacity, supporting the grid during periods of peak demand or system constraint. The unique ability to provide dispatchable capacity helps create a more resilient grid benefiting the utility and the broader community. With our multi-use solution, ERock benefits customers, utilities and the community. A good example of multiple use cases in practice is our partnership with El Paso Electric and Meta in El Paso, Texas. El Paso Electric was facing an estimated 5-year timeline to deliver grid interconnection for a 366-megawatt Meta data center complex and ERock was selected to close that gap. EPE purchased our systems, and we are on track to commission the site approximately 15 months or about 4 years faster than waiting on the grid. As I mentioned earlier, we have already begun mobilizing on site in El Paso. During the initial bridge period, El Paso Electric will collect payments from Meta via PPA. Once grid interconnection is complete, the highly depreciated asset can move into their own rate base, offering ratepayers a significantly lower cost of peak capacity. This outcome is a clear win-win for both our customers and the El Paso Electric ratepayers more broadly. Meta gains access to an operational data center years earlier than may otherwise be possible, while El Paso Electric gains a flexible, low-cost, low emissions grid asset that can support long-term system reliability. At a time when AI growth is placing greater pressure on utility systems and rates, ERock offers a differentiated asset that increases reliability and lowers costs. This is not a new model for ERock. For more than 15 years, we've partnered with utilities, grid operators and the communities they serve to deliver reliable power when it's needed most. Another example of that partnership in practice is our nearly 150-megawatt deployment with the California Department of Water Resources, which marked our first project in California. We believe our selection for this project reflected the differentiated combination of low emissions, highly responsive technology and a turnkey delivery model. From contract signing to first commercial operation, deployment was less than 1 year. Our low emissions profile helped accelerate permitting, which remains an important advantage versus many competing technologies. And because of our kitted modular design, we installed approximately 50 megawatts in 2 days at the first site and replicated that deployment model at the next 2 sites. Those systems have now been operating for more than 2 years, and we are proud to continue supporting the California grid during periods of high demand. Now turning to the second quarter. A big focus during the quarter was on our commercial efforts. We've added resources focused on the AI ecosystem, which has resulted in an increase of our pipeline with multiple large-scale hyperscalers, AI data labs and neo clouds, both directly and indirectly via our utility partners. We also made good progress operationally, building out our assembly capacity and adding project execution resources. The investments in our commercial team are starting to show traction as our pipeline grew over 3x in 2Q. On the commercial front, we received an order for approximately 470 megawatts from Anthropic, one of the world's leading AI research and development companies. This order reflects the equipment purchase agreement only. As with other large projects, installation and service agreements are typically negotiated and signed separately, and we expect those to follow in the normal course. Anthropic selected ERock to ensure power supply certainty for their quickly growing compute needs. The Anthropic contract reinforces the momentum we are seeing across our broader AI and hyperscaler customer base. Together with other recent orders, it contributed to a record level of contracted backlog and reinforces the growing demand for rapid, reliable on-site power solutions for large-scale data center infrastructure. Turning to our assembly operations, which are based in Houston, Texas. This quarter, we transitioned generator assembly into our new Hyperion facility in Northwest Houston. Hyperion is targeted to increase our annual assembly capacity to approximately 1.2 gigawatts by the end of this year at a cost of only $15 million in additional capital. With that, I will now turn the call over to Ian to review our financial results.

Thank you, John, and good morning, everyone. I'll walk through our financial results for the quarter, then cover our post-IPO capital structure, our balance sheet and our cash flow. Before I get into the details of the quarter, it's worth a quick reminder of how we generate revenue. We have 2 main revenue streams, power system sales and ongoing services. Within power system sales, we generate revenue from both the sale of the equipment itself and from the design, construction and installation work to get it commissioned. We recognize revenue on generator sales when control transfers to the customer, which can happen when the generator passes internal factory acceptance testing coming out of our production facilities or when the generator is delivered to site depending on the deal. We recognize revenue on construction, installation and balance of plant equipment on a percentage of completion basis as we incur the cost. Within ongoing services, we generate revenue from our O&M and asset management work that is performed once the system is up and running, typically under contracts ranging from 5 to 15 years. We recognize ongoing services revenue on a monthly basis as services are performed. Total revenue for the second quarter was $39.9 million, up 26% sequentially from $31.7 million in the first quarter. Results included certain customer deliveries completed ahead of schedule, including earlier-than-expected production of generators for the El Paso Electric Meta project. You should think of this production as a timing shift rather than a higher level of production for the full year. In the second quarter, Power System sales revenue was $26.5 million, up 67% from the first quarter as we delivered more generators and performed more installation work. Ongoing services revenue was $13.3 million, down 15% sequentially as a result of some campaign work that was performed in the first quarter and did not recur in the second quarter. As a reminder, we forecast a baseline of activity for ongoing services, but some quarters will have revenue increases due to higher levels of corrective maintenance or campaigns that happen from time to time. Gross profit was $7.4 million, representing a gross margin of 18.6% and adjusted gross margin was 22.2%. Adjusted gross margin excludes pass-through revenues and costs in the ongoing services, which are not intended to generate gross profit and have no economic benefit or cost to us. Net loss was $67.7 million compared to a net loss of $17.2 million in the first quarter. The second quarter's result included a $48.8 million loss on debt extinguishment related to the conversion of certain convertible notes prior to our IPO and the repayment of a $30 million term loan with IPO proceeds, both of which are nonrecurring. We delivered adjusted EBITDA of negative $14 million compared to negative $12.4 million in the first quarter. As implied by guidance, which I'll speak to in a few moments, we are expecting to turn EBITDA positive in the second half of 2026 as we ramp up deliveries against our meaningfully larger backlog and realize increased production from our new Hyperion facility. Since this is our first call as a public company, I want to take a moment on our share structure before moving to the balance sheet and cash flow. ERock's capital structure has 3 classes representing Class A equivalent economic interests. Those are Class A shares held by public shareholders and Class B and Class M shares, both held by pre-IPO owners and each exchangeable into Class A shares over time. Together, all 3 classes represent approximately 272 million Class A equivalent shares. Because of this structure, our GAAP net income is split between the portion attributable to Class A shareholders and the portion attributable to noncontrolling interests, meaning reported GAAP EPS reflects only the Class A portion, not our total consolidated net income. For the adjusted metrics we're discussing today, we present results on a fully exchanged basis, reflecting a total share count of approximately 272 million shares. Over time, as Class B units are exchanged for Class A shares, we'd expect the noncontrolling interest portion to decline. We completed our IPO on June 11, pricing 27.9 million shares of Class A common stock at $21.50 per share. Approximately 1/3 of those shares were sold by our existing sponsor and pre-IPO holders in a secondary offering. Gross proceeds from the primary offering were $400 million and net proceeds were approximately $369 million after paying underwriter fees and other offering costs. As of June 30, we had strong liquidity with $627 million in unrestricted cash, no outstanding debt and an undrawn $250 million credit facility. We also had around $34 million in restricted cash. We believe this strong liquidity gives us significant flexibility to fund our growth without needing to return to the capital markets in the foreseeable future. Turning to cash flow. As John mentioned, our business model is relatively capital light, which means over time, we expect to have a high conversion of EBITDA to cash flow. Because our model is fundamentally assembly rather than heavy manufacturing, we can scale capacity quickly without specialized equipment or heavy machinery. Our year-to-date CapEx of $9 million includes around $7 million towards the Hyperion facility, and we have less than $10 million of additional CapEx to fully build out that facility. Our cash flow also benefits from favorable working capital dynamics. Our commercial and structuring teams have negotiated firm payments at various stages of a project, including upfront payments on contract signing and milestone payments as we purchase long lead time equipment and deliver finished generators. Turning to backlog. Our contracted power system sales backlog stood at approximately $1.7 billion at the end of the second quarter or up 10x year-over-year, as John mentioned. We expect the majority of our backlog to convert to revenue by the end of 2027. With that visibility in mind, we are initiating full year 2026 guidance today. We expect revenue of $435 million to $465 million, with the midpoint representing growth of approximately 2.5x year-over-year and adjusted EBITDA of $3 million to $9 million. Guidance reflects a significant increase in generator deliveries and installations through year-end, which will drive much higher revenue and positive adjusted EBITDA in the second half of '26 as we execute on multiple large data center projects. We continue to see opportunities for increased gross margin and significant operating leverage and EBITDA margins as we ramp revenue and more fully get the benefits of investments we have made in G&A over the last several years. Of our $1.7 billion backlog, we expect approximately $360 million to $390 million to convert to revenue in 2026 with much of the remainder in 2027. With the addition of the Anthropic contract, our production capacity is sold out through 2027 and into 2028. We will continue to prudently consider additions to our capacity based on the demand we are seeing from our customers. Now, back to John to close out our prepared remarks.

Now that Ian's walked you through the details of the quarter, I want to wrap with a few comments before we open up the call for questions. Executing on and building backlog is our top priority. We're committed to delivering on time, on budget and at scale using our vertically integrated platform and capital-light model. Near term, the biggest focus areas are scaling up on the El Paso Meta project, fully ramping our Hyperion facility and ensuring our supply chain partners are aligned with our long-range forecast and significant market demand for our solution. In addition, we're seeing multiple ways to grow and strengthen the business further. We are expanding our partner network with leading AI developers as evidenced by the Anthropic announcement, data center operators, engineering firms, consultants and other potential partners in this ecosystem and growing our presence with utilities and grid operators who increasingly value the kind of robust, flexible capacity we provide. Regarding the technology platform, we are focusing on growing our proprietary Granite ecosystem software, O&M and asset management services. This capability continues to build our growing base of recurring revenue as our installed fleet expands. And we are continually investing in R&D work on our proprietary generator to make it more efficient, more robust and lower cost with a focus on extending our leadership and delivering additional value for our customers. Finally, I want to recognize our ERock team. The execution across every function while taking the company public, commercial traction and project execution in addition to rebranding the company over the last 8 months has been remarkable. Thank you all for everything you do in making ERock what it is today and the significant role you will have in our future success. With that, I'll turn it back to the operator to begin Q&A.

[Operator Instructions] This morning's first question is coming from Angel Castillo of Morgan Stanley.

Congratulations on the IPO and on a strong first quarter here. So John, Ian, just can you give us a little bit more color on just the backlog and the pipeline, I guess, in particular, one of the more unique definitions of backlog with it only including things that are -- orders that are kind of firm and where you have deposits and you kind of protected your economics. So I just would love it if you could give us a little bit of additional color on just orders that you may have for either equipment or ESI that you just have -- you're waiting for them to firm up? Or similarly, just any color on how the evolution of the pipeline, quoting activity and customers' interest. We've seen a pickup in pushback to data centers, but it seems like you're seeing very strong growth in your pipeline. I think you mentioned 3x versus where you had kind of -- where you were previously. So just curious on that evolution of the pipeline and that conversion rate.

Sure. Thanks, Angel. Thanks for joining the comments. This is John. I'll kick off. And Ian, if you have anything to add, please do. I think you're right, we do continue to see strong demand for both our power systems and our turnkey services. As mentioned in the prepared remarks, our pipeline is up 3x, and that's from 1Q to 2Q, and we see that continuing to grow. I would also say that larger deals like Anthropic and other data centers, while they can be lumpy and hard to predict quarterly, we do see significant momentum and more logo development. And a large part of that has to do with our investment in that specific data center ecosystem with the commercial team that we've added. We've put a lot of resources around this, and we're starting to see some nice traction in that area. I would also note the Anthropic deal is a new deal. It's a new customer and a generator-only agreement at this point. We are, though, also seeing great visibility for additions to existing contracts on the EPC and O&M side. And we think that those additional -- those customers will also provide additional phases to the existing projects. I would also highlight beyond the data center side, we do have 2 other key segments that we continue to see very strong momentum around that being the utility customers, a segment that we've been engaged in for really many years and it has actually turned out to be a great channel partner for us, Angel. We're seeing more and more hyperscalers come to the utilities asking who they would recommend, who they've had a good experience with and ERock continues to be the one they're recommending. Finally, on the C&I side, another important segment for us and a legacy segment that we've been with for many years. It's interesting. We're seeing more and more constraint on interconnection as low as 5 megawatts. So we do see speed to power going well beyond the data center side. And again, with our modular design, low noise and low emissions, we really have become the product of choice as that key market expands. Ian, do you want to add anything?

No. The only thing I would add is, as you noted, Angel, what we put in the contracted backlog is firm. We're not giving specific guidance on sort of what else is out there associated with those contracts. But I think in general terms, it's fair to characterize that, that continues to grow alongside the contracted backlog. So we see the business developing well in the right direction, the way we want it to be developing right now.

Very helpful. And then just from a follow-up, I guess, in the past, you've talked about being in kind of early days of testing your pricing power of the product offering. So with such a differentiated solution, just curious if you could talk about kind of those efforts and how that's progressing in the pipeline. Just as we think about, I guess, in particular, this Anthropic order, if I just take the implied increase in the value of the backlog and just divide it by the 470 megawatts, I guess, something like [ $900 to $1,000 ] per, I think, kilowatt or megawatt rather range versus that [ $700 to $800 ]. So just curious, it seems like it's progressed quite a bit, if that's the right math. So just curious if that's directionally the right way to think about this or just any other color on how that pricing is evolving?

Yes. So in general, we're not giving specific guidance on ASPs, and I'd be a little bit careful of trying to back into the math because different attributes of different deals can shift that math around. But what I'll say sort of categorically is as we look at the deals that we're negotiating currently, we continue to see positive trends in pricing and gross margin. It's worth noting that there is a lag associated with when we see the impact of higher gross margin coming through into the financials because we are sold out through 2027 and into early 2028. Really, that impact is more 2028 and beyond impact that we're seeing. But we're continuing to see good pricing dynamics and good demand in the space.

Our next question is coming from Mark Strouse of JPMorgan.

I'll echo the welcoming to the public market here. John, I know you talked about -- I'm sorry, you talked about kind of being sold out through 2027 and into 2028. Obviously, part of your solution here is kind of your quick time to power. Can you just talk about how you're thinking about additional capacity? Obviously, you brought on Hyperion now. Are there any metrics that we should be looking for as far as the backlog or pipeline over the next several quarters, maybe over the next couple of years and how you're thinking about adding further capacity?

Yes. Thanks, Mark. A couple of things there. One is we are being very prudent as we think about our expansion, assembly expansion. And the logic we hold is if it's 65% contracted or more, that triggers our perspective on doing expansions. We are continuing to review that on a regular basis, particularly with the incoming demand that you noted. I think the really -- the interesting benefit that we have is the ability to ramp up a new facility in a 12-month timeline at 1.2 gigawatts at a cost of $15 million to $20 million is something that's pretty compelling. I think as this evolves, Mark, we're going to look at where we want to put these locations. Obviously, we have a very core engineering team here in Houston, and that's working well for us. But as we go forward, we have the ability to expand into various regions of the country. And I think what's also compelling is we have the ability to put this assembly facility closer to the customer. Because our facilities are assembly, it's a pretty basic facility in the sense of it's cranes, it's forklifts, and its inventory, and we could actually go to a location for 2, 3, 5 years and move that somewhere else as needed. So we're continuing to monitor it. We do not intend to pass on opportunities because of capacity. So we'll update everyone as that progresses. And as I said, I think the velocity at which we can add is very compelling and highly differentiated.

Yes. Okay. That makes sense. And then you're describing the Anthropic deal as a new deal with a new customer. During the IPO process, there was a reference to an unnamed customer that had multiple phases. And I think part of that project or part of that contract was in your backlog. So it sounds like you're still not disclosing the name of that customer, but are you able to kind of give us any more color on how to think about additional phases of that contract coming into your backlog maybe over the next few quarters or so?

Yes. And to be clear, Anthropic is not that customer or customer A as referenced. So we are -- but at this point, we're not really disclosing any more beyond that. We obviously will as it becomes public knowledge. So we'll roll something out about it. Do you want to add anything on that, Ian or...

No. I mean, I think as far as growth and upside opportunities on that project as well as a number of others, we have additional potential. And I think the way that we look at building out our business is we want to do the right thing for our customers. We want to build those relationships as long-term relationships. And so whether it's at an individual site or across multiple sites over time. We're always looking to, in addition to what we may have under contract today, having other opportunities we're working with those customers on to help them build their business, and we'll build alongside them.

The next question is coming from Michael Sullivan of Wolfe Research.

I wanted to just ask a little more on the backlog and potential pipeline in terms of size of projects. This Anthropic one you just announced is, I think, your largest. Is that kind of reflective of kind of the new norm for size of deals? Or is it trending even higher? How are you thinking about that?

Yes. Thanks for the question. Look, I would say that it depends on which segment you're referencing, right? I think the training side will be probably a little larger than these. We're seeing anywhere from [ 5 ] to over 1 gig. The inference piece is a little bit smaller. But look, I think this is the right ZIP code and higher for the majority of the deals that we're seeing, very strong momentum around multiple phase projects. So while you might have one that could be larger than this one, then when you tack on 2 or 3 phases to it, it becomes significantly larger. So in general, I would say the momentum is moving to larger scale, yes.

Okay. Great. And then just in terms of regionally, where you're seeing the most opportunities? I know you guys kind of being down in Texas, there's been some news on that front just recently last week with the Governor's letter, delays to the Batch Zero process. Is that creating an opportunity? Or yes, just how you're thinking about the landscape regionally for your pipeline?

Yes. I would say that from a geographic diversity standpoint, seeing opportunities in the Midwest, the West and certainly core southern markets that you've highlighted, Southern Southwest. Do you want to highlight anything on the regulatory front? With Texas in particular, Corey.

Sure. This is Corey. Yes. So Batch Zero, the Abbott comments that came out -- Governor Abbott that came out a few weeks ago. Really, we've seen no impacts on that. We are seeing a lot of interest in Texas. It seems to be the easiest place to get a site set up. And although we've seen no impacts, we think this will cause a slight delay in the Batch Zero process. But most of the people we're working with, especially like even if you mentioned the El Paso deal with Meta, right, they've come out and vocally supported Abbott's comments on that and are aligned on that. And they've even set up a fund to establish money for the communities to support different parts of the community. And so we think that's very positive for that facility. On other sites, we're seeing most of our potential customers as being aligned with those comments, too. And we think we're a good fit for that also. We have ultra-low emissions. We don't use any water. And then our gensets, we're generally trying to get people on the grid rather than keeping them off the grid. And then once we're on the grid, then our gensets generation behind the meter is supporting the grid. And oftentimes, that will be subsidized by the customer that will help keep rates down for the rest of the community, commercial, industrial and residential.

The next question is coming from Nick Amicucci of Evercore ISI.

Just wanted to kind of drill down a little bit on the utility relationship. It seems like, obviously, interconnection queues continue to expand and the speed to power, I mean, is paramount. So just if we could kind of just discuss kind of how those conversations are going? And I guess, the sense of urgency associated with -- from the utility side of the house. Obviously, it seems pretty -- it seems there from the hyperscalers. But if we could just kind of drill in on the utility side, it would be helpful.

Sure. This is Corey. Yes, we're seeing increased interest on the utility side. And a lot of it is related to the data center side, too, or just load growth within their region. I'd say, really in the last 6 months, we've seen a decent increase in that and concern on speed to power also. A lot of those projects would be in front of the meter, but they still operate as providing grid support kind of like a peaker power plant could run anywhere from 300 to 3,000 hours per year based on the certain geographic area. But we're seeing that across the country really. It's really interesting.

The only thing -- this is Ian. The only thing I would add on top of that is, as you noted, and I think you implied in your question, Nick, historically, hyperscalers and others in the AI ecosystem have been the faster movers. We have seen some acceleration on the utility side in the speed at which they're moving. Obviously, until we have dealer deals to specifically discuss it, that will be the true proof point, but we are seeing some acceleration in those conversations and some willingness to move a little faster than they were.

Great. And then Ian and John, I know we've discussed in the past, and I just wanted to kind of -- just kind of hone in on it. If we think about kind of -- so obviously, a big announcement with Anthropic, and it seems like that's more on the training side. Are the discussions then kind of -- is it the intent that they have that they kind of deploy the generation assets for the training and then that proof point is then kind of continued on to the inference side? I know you had touched upon it a little bit before, John, but if we could just kind of drill into the inference opportunity, it would be great.

Yes. I think it's a blend, Nick. We're not -- it's not as differentiated as you would expect at this point. And it's not necessarily let's make sure that the training model works with your product, then we'll go to inference. So we're definitely seeing activity on both. I think the training piece is still the lion's share of what we're hearing. And again, the feedback we're getting is because of our modular design, the low profile, the noise, low noise attribute and the fact that the inference will be definitely in urban areas because of the latency discussion really bodes well for us. So we're definitely having discussions around both, but I would say stronger momentum on the training side, certainly for right now.

[Operator Instructions] Our next question is coming from Joseph Osha of Guggenheim Partners.

I've got 2. First, in going over the financials, I see there's a pretty substantial uptick in deposits, or contract liabilities. So I'm wondering if you can help us understand how bookings are converting to cash, in particular, how that compares to the revenue realization. And then I do have one other question.

Yes. This is Ian. Thanks for the question, Joe. So in general, what you'll see in our contracts is right now, just given sort of market conditions, there are significant upfront deposits that we're receiving associated with our contracts. I think we've talked a little bit about the cash payments are generally milestone-based and generally align with or are a little before we are actually deploying that capital into parts, equipment, installation services, et cetera. And so what you see there is significant cash payments that we're receiving from our customers, offsetting contract liability, which will be recognized as revenue over the course of time as we build out that project. And so really, it's just sort of a temporary phenomenon where the cash is ahead of the rev rec on those set of projects. I think maybe a framework for thinking about that is, over time, as we've discussed, as we forecast our business, we assume sort of neutral terms in the current market environment. The cash payments are a little more front-end weighted. But I think it also serves as validation of what's in our backlog, contracted backlog or things that are very firm where customers have already made substantial cash payments. So they're very locked into these projects.

Sure. So just from the standpoint of modeling, though, yes, this is pretty standard stuff. But can I think of it maybe as sort of a 1- or 2-quarter lag in terms of the deposits versus the timing at which that converts to revenue?

I would say it's probably a little bit longer than that potentially depending on the project. I would say it's typically, call it, 1 to 3 quarters depending on the specifics of the project.

Great. And then my follow-up, we've heard a lot about time to power and 24/7 and so forth. I'm hearing about some cases where utilities are saying, "Hey, we'll hook you up, but we want the customer to shed load from, say, 3 in the afternoon to 9 at night in some areas." Are you guys seeing those use cases where you're running but only part of the day?

I think we haven't had those specifically, but I do see that coming. I think that's going to be part of the Batch Zero process with ERCOT that you can get partial loads approved, right? And you may be operating for a partial load or you may be operating certain hours. But I definitely think that's the way of the future. And that's something that we've operated in those modes for a long time, for the last 10 years, we've always been providing back of a grid dispatchability. And so we're excited about that, and we think that's the right way to use these assets behind the meter.

Thank you. Ladies and gentlemen, this brings us to the end of today's question-and-answer session and today's conference call. We would like to thank you for your interest and participation in today's teleconference. You may now disconnect your lines or log off the webcast, and enjoy the rest of your day.