Prepared remarks
Welcome to the Gorilla Technology Group Inc. Fiscal Year 2025 Financial Results Conference Call. The conference is being recorded. Before we begin, we will read the forward-looking statement. Today's call includes forward-looking statements made under the safe harbor provisions of the Private Securities Litigation Reform Act of 1995. These statements reflect management's current expectations and projections about future events and are subject to known and unknown risks and uncertainties that could cause actual results to differ significantly. Forward-looking statements often include terms such as expects, believes, plans, anticipates, may, should, and similar expressions. For a discussion of important factors that could affect Gorilla's results, please refer to our filings with the SEC, including our most recent annual report on Form 20-F. Except as required by law, Gorilla undertakes no obligation to update or revise any forward-looking statements made on this call, whether as a result of new information, future events, or otherwise. I would now like to turn the conference over to Jay Chandan, Chairman and Chief Executive Officer; and Bruce Bower, Chief Financial Officer. Please go ahead.
Thank you very much, Crystal. Thanks, everyone, and thanks for joining. I will keep it quick. If you want drama, the market's already provided enough today. So I will stick to the facts. Now let me start with the headline. We reported a record full year revenue of $101.4 million, up 35.7% year-on-year. This is the first time in our history we have crossed $100 million in annualized revenue. We guided the market at $100 million to $110 million, and we delivered within that range. That matters because credibility matters, and we intend to keep it that way. Now the more important part is how we got here. We executed a real turnaround. Our IFRS operating loss narrowed to about $13.7 million from $66.9 million last year. That was a remarkable improvement of $53.2 million or 79.6% reduction in the IFRS operating loss. Now our IFRS net loss narrowed to about $11.3 million from $64.8 million last year and 82.6% improvement.
And IFRS basic EPS improved to about $0.51 from negative 6.13%, which is a 91.7% improvement. So yes, it was a proper swing. It was not just a cosmetic one. We did all of this while keeping the underlying profitability at scale. Adjusted EBITDA came in around $19.1 million and adjusted net income was about $19.9 million, with our adjusted basic EPS being $0.89 and an adjusted diluted EPS at $0.88. What I can tell you is that it is strong, and it is very disciplined. Now I know what comes next because investors always ask it: how did we do versus expectations? For the fourth quarter, the market consensus was roughly around $34.75 million of revenue and adjusted EPS of $0.30. Based on our full year results, our fourth quarter revenue was approximately $35.6 million, which is well above consensus. And based on the implied fourth quarter adjusted earnings, our adjusted EPS was roughly around $0.37, which is about a 22% beat versus the $0.30 consensus.
Now for the full year, the market consensus was approximately $100.6 million of revenue with a $0.80 billion for adjusted EPS. We delivered roughly around $101.4 million of revenue and delivered about $0.89 adjusted EPS, which is about a 6% beat versus consensus. So the message from my side is simple: we delivered record revenue. We delivered a major IFRS turnaround. We delivered underlying profitability that exceeded expectations. Now let's just talk about the broader market because it has been volatile. The market conversation has shifted from 'you beat the quarter' to 'will AI spending hold up?' And I'm sure all of you have seen this in the last few days and weeks. That is a fair debate, but personally, it misses the bigger picture. AI is no longer a discretionary software trend. It's rapidly becoming a national capability and a core operating layer for enterprises and governments. Now the next phase of AI demand cannot be defined by one buyer or one deal.
It will be defined by many buyers across various sectors, building permanent capacity: governments, regulated enterprises, telecom operators, logistics networks, financial services platforms. This list is long, and the spend is becoming structural. The compute is also evolving at a rapid pace. This is what the market is really missing. Now AI compute is actually shifting from the training cycle to an inference-led cycle. This is important because this does not reduce demand; it broadens demand. Inference pushes AI into everyday workflows and mission-critical operations, which increases the need for distributed compute across regional data centers and edge environments where latency, data residency, and resiliency requirements matter. Now this is where edge becomes a major driver: as most of you know, we were one of the leading edge companies when we went public, and we continue to invest heavily.
Edge compute expands what AI can do because it moves inference closer to the decision point, closer to the sensor, closer to the customer interaction, closer to regulated data. It becomes a force multiplier for adoption in public safety, transportation, logistics, financial services, telecom networks, industrial applications, and the whole plethora of smart cities. Now let us talk about the scale of the infrastructure market in our region. We're not relying on slogans. We're tracking that data very closely. We have an internal team, we have a research team, which is doing that, and we use external data at the same time. Now we see Asia Pacific data center investment growing from roughly $30 billion in mid-2026, up approximately to about $90 billion by 2031. We see installed capacity broadly doubling from about 29,000 megawatts today to about 63,000 megawatts by the end of the decade. Now Southeast Asia also follows the same trajectory, going from the low teens in billions to roughly $30 billion by 2030, as more capacity is being built in the market rather than exported offshore.
India is another example. It’s scaling rapidly. From just over 1 gigawatt of installed IT load today, they're moving towards about 1.8 gigawatts by 2027 and to multiple gigawatts by 2030. We are seeing the same trend in the Middle East, with market growth from low single-digit billions to high single-digit billions by early 2030, as governments and national champions scale local compute and secure infrastructure. This is the structural build cycle we are positioning Gorilla for. So what are we doing in '26? We are advancing our AI infrastructure and data center build strategy across Malaysia, Thailand, Indonesia, Singapore, and other regions, including Taiwan. We're expanding our evaluation work in India and progressing in the Middle East, including Saudi Arabia, where an MOU has already been signed, and we're exploring data center development opportunities in that region. We're also exploring opportunities to buy and/or build our own data center assets.
Ownership changes the model; it gives us more control over our delivery and stronger long-term positioning and the potential to build recurring infrastructure-led revenue streams rather than relying on project cycles. Now in parallel, we are also strengthening our product edge for its next stage of adoption. Our first quantum cryptography is targeted to be ready in April 2026. Our local interception product suites remain in continued research and development as we expand sovereign-grade capability across security and intelligence as well as compliance-led deployments. Now come 2027, we're also assembling a team, which will be investing heavily into local interception as well. Now we currently have about 300 full-time employees today, a little over 200 contractors working on all the projects we have signed. Based on just the projects we have recently signed, we anticipate growing to about 1,200 to 1,500 full-time employees by mid-June next year, and that would be about an additional roughly 700 to 800 contractors.
So we'll have roughly between 2,000 to 2,500 employees for the company at any given point in time. Now investors want proof. They want execution, not a narrative. So I will speak directly about the signal that matters: delivery and collections, more about the cash conversion. Our top customers are progressing very strongly, and our customer satisfaction is reflected in our payment behavior. In the first two months of 2026, we have collected more than $22 million from our largest customers for solutions delivered and invoiced in 2025. We also expect meaningful collections in the coming weeks. Now we finished the year 2025 with about total cash of $104.8 million. It was very important that we did all this by reducing the total debt load to about $13.8 million, which is 35.6% lower than the $21.4 million in the prior year. Through the refinancing of certain lending agreements and the repayment of others, we also reduced our debt, releasing more than $5.3 million of deposits previously held as collateral against some of these loan obligations.
Now this kind of balance sheet gives us very meaningful flexibility to execute existing programs, fund working capital delivery cycles, and scale our infrastructure strategy with discipline. We've also spent at the same time, more than $11 million on buybacks today, which we believe the market continues to undervalue Gorilla relative to our performance and strategy. Personally, I think you could call this confidence. I call it arithmetic. Right? Why? Because that leads me to my next point. We're aiming to be cash flow positive in 2026. That's not just a slogan for me; it's an operating objective that comes with very disciplined delivery, disciplined overhead control, and a very disciplined cash collection. And finally, a lot of people have asked me this question over and over again: Gorilla Technology Capital. Personally, it's a game-changing catalyst for our next phase. It's designed to expand our ability to execute larger infrastructure programs by structuring capital efficiently, aligning long-duration funding with long-duration assets, and enabling our customers to move faster with clear financing pathways.
Some people said, 'hey, maybe they're buying the bank.' No, we're not buying a bank. You have to understand, what Gorilla Technology Capital does is strengthen our ability to scale data center build, accelerate GPU infrastructure deployment, and importantly, we participate materially in larger mandates with institutional-grade structures and governance. So if I summarize 2025 in one line: we delivered a historic revenue milestone, we executed a major profitability turnaround, we strengthened the balance sheet, and positioned Gorilla for the next stage of AI infrastructure, which is sovereign and regional, more importantly, distributed, which is becoming increasingly edge-enabled. In 2026, we shift from proving we can build to scaling what we can deliver, converting execution into cash, expanding our data center footprint across India, Malaysia, Thailand, Singapore, Indonesia, Middle East, and more importantly, using Gorilla technology to unlock materially larger programs without compromising.
All this while accelerating our product roadmap, which means we're investing heavily into R&D. Thank you for your time. I will hand over to Bruce, who knows the numbers well enough to recite them without blinking. Bruce, please go ahead.
Thank you, Jay. I think you covered the main points regarding the financials. I wanted to highlight a few things. First, we mentioned that the cash balance at the end of the year was $104.8 million. I want to emphasize that due to the collections so far this year, the cash balance actually increased. As of February 26th, it was $108 million in unrestricted cash and $116.6 million in total cash, despite spending $3 million this year on share buybacks. We have managed to increase cash while also buying back shares this year, marking a strong start to the year. I would also point out that when we discussed reducing the debt load and freeing up cash deposits, some people asked why we didn’t pay off all the debt. The remaining debt of $13.8 million has an average interest rate of 3%, so it makes sense to keep it as flexible capital instead of repaying it and borrowing at higher rates. Lastly, we issued guidance last year with a revenue range of $137 million to $200 million for this year, and we are maintaining that.
At this time, we are not ready to provide guidance on gross margin or EBITDA, but we will have updates in the coming months. The wide range depends on the delivery schedule of certain data center projects we are pursuing with Freyr and others. I believe we will have a great update in about one to one and a half months regarding the timing of those projects and delivery schedules, which should help clarify the guidance. I’d like to reinforce what we mentioned in the press release: we believe our balance sheet has improved to the point where we can fund growth initiatives and buy back undervalued shares. This includes not only the anticipated revenue growth this year, which is projected to be almost a 70% increase, but also the contracts we have in the pipeline, representing a $7 billion revenue opportunity. We are confident we can fund much of this through access to debt facilities, primarily via project finance, as well as the cash we currently have on the balance sheet. With that, I’ll turn it back to Jay for questions.
Thanks, Bruce. I'd love to open up the questions to all standing by. Thank you.
Questions and answers
Your first question comes from Brian Kinstlinger with Alliance Global Partners.
Yes. Close enough. You've come certainly a very long way over the last 2 years. Congratulations on that. Has anything changed in terms of your best guess on timing for the first 3 phases of the Freyr partnership? I think the plan was project financing to help you start in April for Phase 1, September for Phase 2, and December for Phase 3? And then the second part of that question: outside of financing these projects, are there any gating factors to starting these projects? If so, what needs to happen in those time frames?
Brian, good to hear from you, and thank you for your kind comments. We are on track with where we are today. Obviously, considering the market forces today, we have had some slight delays in terms of the delivery. But that said, let me walk you through what has happened: some programs have moved in terms of timing. We talked about the Freyr contract, for example; that is on schedule. We are currently in the final stages of getting our first set of GPUs coming through over the next few days, and we'll be deploying them as we speak. We have also accelerated the timing on some of the data center discussions. When we spoke last, I think we were looking at about 12.5 megawatts of data center. If you recollect, we were looking at roughly hundreds of high-density allies. What we've done is rather than commissioning them all on a single day, we're slowly putting them in place. Power cooling network zones are all commissioned.
Revenue ramps are going to be energized as we speak. As the racks go live, we will drop in the clusters through our partner ecosystem, which also drives what I call the GPU-as-a-service usage line. Now what's very exciting for us, and I can tell you today, is we have realized we need to deploy a lot of capital in the data center space ourselves because we've been inundated with a ton of requirements. We are currently looking at about more than 600 megawatts of capacity rather than the 12 megawatts alone. That allows us to control our destiny over time, meaning we're looking at several hundred million dollars per year once all these racks and GPUs are in motion. From our perspective, Brian, the part that I want to highlight is a very controlled ramp-up, not a single bank. Now, you mentioned, are there any delays? There are no significant delays so far. The Thailand MOU, for example, has been delayed because of the political transition, as there is a new prime minister elected.
So we're just waiting for the post-election leadership and sign-ups to settle. But otherwise, we are not facing any delays. We're going ahead with all of the approvals, all of the permitting, and all of the customer prerequisites as we go into it. So when the customer gates reopen, I believe we will start our billing on time. I hope that answers your question, Brian.
It does. My second question is you've got this large pipeline of other data center opportunities you've discussed. And not to say that your business development has been slow—it’s been very fast—but do you think those customers are waiting to see how execution is on the first Freyr contract? Is that going to, in the near term, hold back agreements? Or do you believe those will be able to move forward without delivery on those 3 projects?
Absolutely not. Like I mentioned, our pipeline is exploding. We have not been slow in our sales, Brian; I can assure you. The only thing we've been doing is restricting. We've been inundated—‘inundated’ is the right word for that. First of all, the deals are mature. At the start of the year, we were looking at POCs and MOUs and so on; it was very promising. But since then, we have moved into late-stage commercial structuring or improving force contracting, which significantly increases the scale and certainty of the pipeline. If you recall what I shared at the end of December, we ensured we have certainty in the pipeline. Now I mentioned the $1.4 billion Southeast Asia contract; that was only a catalyst. Once the government and telcos saw what we could deliver and we started signing up with the first 12.5 megawatts, suddenly it triggered some sort of a sovereign-grade AI infrastructure requirement and a huge surge in interest for us.
So I don't want to give you names, but what's happened is the demand behind that is significantly larger than Freyr itself. That is one of the primary reasons our pipeline is now in billions of dollars. The third most important part: things have changed from ambition to urgency. Governments are no longer looking at AI capacity as an ambition; it has turned into urgency for us. Now, I mentioned before that not only are we looking at GPU capacity as strategic infrastructure, but the shift into edge computing distributed environments is taking shape right now. And as I said, the market has already missed that. People think, 'oh, is spending going to continue?' It is going to accelerate; it's not just going to continue at its current rate. It is going to go exponential. We are engaging with every major customer on the planet, and I can assure you that these platforms will explode in terms of compute requirements and demand.
Finally, our execution has not been on just one data center; we've been doing data centers for a long time. We've built data centers on behalf of governments, for example, in Taiwan, Thailand, Egypt, and so on. So when we deploy large-scale local interception programs, which are more complex than simply setting up data centers, the governments and organizations look at what Gorilla has delivered, and their confidence grows. So we are not resting on past accomplishments; we are putting everything into motion. As I said in my previous response: we are now targeting over 600 megawatts of power. So the opportunity ahead is substantial and continues to grow comfortably, Brian.
Great. My last question relates to the recruitment needs. Your type of business is always a great leading indicator. How would you characterize the recruiting market in the geographies you're hiring? Additionally, outside of the execution staff, are there significant AI HPC senior executive levels that give you that additional strategy and expertise at a high level?
That's a really good question. We are hiring at a rapid pace. As you know, we are building our teams worldwide. In Thailand, for example, we're currently hiring about 80-plus people. In Taiwan, our data center team and R&D team for cybersecurity products have been deployed significantly. We've done that through what is called a hub-and-spoke model. This is crucial because our R&D platform and engineering needs to accelerate both our product and services capabilities. On the services side, as you know, Satish came in mid-last year, and he's been driving client impact and deepening our technical capabilities. On the R&D side, we’ve been hiring for SD-WAN, post-quantum cryptography, local interception capability, and video analytics; we've been expanding that product. By April, we expect to fully launch the world's first fully ready post-quantum crypto SD-WAN, and we're already working on substantial proof of concepts with customers as well.
But this is what matters, Brian: localization. Every region we work in, whether it’s India, Middle East, North Africa, Southeast Asia, or East Asia, is inquiring about how we are building stronger ground capacity. So we are building teams in Thailand; for instance, our team in Thailand, because we're pursuing very large data centers here, will be about 1,000 people by the end of this year. In India, we will have between 200 to 300 people on our team, and in Taiwan, we will exceed north of 200 people. We are also hiring senior executives at the same time. Thomas has joined us as CTO of Infrastructure. Jackie has come from the hardware side and become the GM for Asia. We are hiring next-level capabilities under both of them as well. Furthermore, we are tightening finance and compliance processes; thus, we are hiring to improve cash discipline, collections, controls, audits, and so on. The hub-and-spoke model is becoming the central framework across these regions. As we continue to expand and grow, we'll develop our teams rapidly over the next course of months and they are ready and operational now.
Thanks for the presentation. Just a few questions from me. To start off with the gross margin, just wondering about the mix, which resulted in a slightly different gross margin than what I was expecting. I wanted to understand what the mix of revenues is? And the second question is given that you're going to deploy the latest compute for data centers in Southeast Asia, what kind of level of revenue are you modeling per megawatt there? What sort of use cases are you thinking about for that?
Bruce, do you want to take the first part of the question? I'll take the second part.
Sure. I think a better way to think about it is that in 2024, we had abnormally high service mix in the revenue mix, so the majority was service. Then it was a higher percentage of hardware in 2025, roughly 40%. That’s why the gross margins were a little bit lower than you would expect. The other thing is that we announced last year that we had signed two major law enforcement customers in Asia, and in at least one of those cases, the margin going into the project was a little bit lower than we normally expect. That's because it was a key win for us as a client and as a solution to demonstrate our capabilities. So altogether, that explains why the margin drifted a little lower. I would say that going forward, building on what Jay mentioned about the pipeline, we will have the ability to be very choosy about the projects we undertake. So with considerable demand, if the margin terms, credit terms, or the credit profile of the customer aren't right or if payment terms aren’t favorable, we can simply say, 'I’m sorry, you either come in line, or we will move on to the next project.'
The GPU-as-a-service offering holds an extremely high gross margin; it’s 70% plus, with 70% being the minimum cutoff. There is, of course, a depreciation hit because an SPV would hold the equipment, and then that would be consolidated onto our financial statements, resulting in a depreciation charge. I’d estimate at scale it would yield about a 25% operating margin, but again, I’m not providing the exact forecast for margins for this year yet, as we want to wait until those details are firmed up. However, I think 2025 will reflect a dip in terms of gross margins ultimately improving over time as we secure contracts.
Just to add to that as well, Bharath, more importantly, we are investing heavily into building the business for sustainable long-term growth and gross margins. Now that brings me to the second part of your question. In terms of pricing today, in Asia, it's structured either in capacity per server per month or in terms of usage per kilowatt hour depending on the customer and program. Typically, for sovereign enterprise deployments, we are targeting contracted multiyear take-or-pay agreements, where the pricing, sustainable margins, and cash conversion are predefined. So we know exactly what we are getting into. I avoid quoting a single rate; personally, I don't want to quote a standard rate because it varies by GPU class, term length, utilization profile, power, cooling specs, location, land values, service level stacks, and so on and so forth. The proof point for us only arrives once we sign these programs where the unit economics are very disciplined, and our collections and milestone payments protect our cash.
So there isn't a single pricing model for Asia, by the way, not just in Southeast Asia; there's no single price for Middle East or Asia. That said, I can tell you that typically, if you're looking at cloud GPU rack capacity, they can range in high four figures to low five figures per GPU per month, bundled with power, floor space, connectivity, managed services, while also considering that these are long-dated, fixed milestone agreements. Additionally, we often layer what we call service level fees with compliance components, etc. Each of these can vary; for example, in the U.S., spot rents for top-tier GPU can be two to three times what you see on structured regional capacity in Asia. However, we’re positioning ourselves in such a way that we’re not reliant on setting a standard rate. Since our compute requirements are more stringent here and our contracted deals are longer, we can provide highly competitive pricing compared to even the United States.
Think about it this way: compliance premiums and service premiums would typically add 20% to 40%, as we include covenants, telemetry, managed ops, and so on. However, the energy cost differentials mean that Asia deals are often much more profitable. So to compare Asia and the U.S. is like thinking of a hotel in Vegas being cheaper, but penthouses in Bangkok being much more expensive than some in Manhattan. Does that clarify your question, Bharath?
Yes, absolutely. May I just sneak in a couple more quick ones? Regarding the Astrikos acquisition, does that carry—are you planning to have explicit pricing and margin contribution for the new contracts you've signed for this? Or is it currently bundled to strengthen your competitive advantage and increase long-term customer lifetime value?
That's a great question. Let me update you on why we invested and what we are integrating. First of all, what is Astrikos? Astrikos is a real-time infrastructure intelligence engine that does monitoring, prediction, and optimization for critical systems. It’s already deployed in very serious environments, including high-stakes smart city platforms, for example, the new Indian Parliament complex, and major initiatives in the Middle East. That matters because Astrikos is not a demo; it is a fully deployed solution. Regarding the second part of your question about our intentions with it, we are integrating Astrikos into three main parts of our stack. First and most importantly, smart city and national infrastructure operations. It provides us telemetry and prediction layers that make national infrastructure measurable and optimizable in real-time. This enables us to sell outcomes, not just technology, with real uptime and response times, thereby generating high operational efficiency for the customer.
The second part concerns video intelligence and security. Astrikos facilitates real-time monitoring and decision-making around critical infrastructure, security, and operational workflows, enhancing our video intelligence stack and improving our operationalization of data across our SOCs and NAC environments. Thirdly, as for GPUs, which data centers require, Astrikos helps in running heavy GPU environments by ensuring continuous telemetry, predictive optimization, integrated security, and operational automation. Hence, Astrikos addresses that need. On the topic of our springboard with Astrikos: it provides us substantial presence in India, significantly shortening our sales cycle and enhancing our delivery readiness. In the UAE, we're already working on expanding our Middle Eastern footprint; in the USA, it's a standard, partnership-driven market. So we are progressing on market-level entry work in that region as well. To summarize, we are a significant minority investor; we have an option to materially increase our ownership and are gaining the flexibility to integrate and build traction on a large scale.
Congrats on a great year, excellent results there. I guess just a couple of things. You talked about maybe more collections coming in this quarter. Can you frame that a little bit more? Are we talking a few million dollars? Over $10 million or maybe you can't say, but just curious?
Bruce, do you want to take that?
I would say it’s plus or minus—it's $10 million plus or minus a few million, $2 million to $3 million on either side.
Okay. And that relates to the 2025 effort?
It's solutions that were delivered and invoiced in 2025, yes.
To keep it simple, the large Southeast Asian deal sounds like no change there in terms of total value or value for the first three data centers. Is that right?
That's correct. But that has become a catalyst, as I mentioned previously.
Great. And then Jay, you mentioned maybe seeing your first group of GPUs in the next few days. Can you provide a little more clarity on that? Does this specifically relate to the Southeast Asia deal? Also, will you be receiving GPUs weekly that will gradually build over time?
Sure. We are creating a flywheel effect, if I may, Mike. What we’re doing ensures that deliveries are coming in every week. The latest agreements we have with our OEMs mean we will start receiving deliveries next week. Remember, we’ve actually won other contracts as well, so we are delivering against those contracts simultaneously. You will see a regular flow of deliveries. Thus, those data centers are driving GPU demand, and for us, our GPU demand unlocks deeper national engagements. This isn't merely a one-off related to the Freyr contract. As I said, it's a catalyst for a number of very large contracts we've signed. We also have agreements with local OEMs in the region; we've signed all the necessary MOUs, LOIs, pricing agreements, BOMs, and completed SOWs. We're now just working on delivery schedules over the next few weeks.
So these GPUs will go to more than just the Southeast Asia customers, it sounds like?
Yes. If you give us a few more days, please, I’ll provide a concrete schedule.
In terms of the Southeast Asia deal, the first data center, you're still thinking it will be up and running in the second quarter?
We're trying to push it to the first quarter, depending on the delivery schedules. But I’m confident it will be live in the second quarter. We've just completed the BOM agreement and sent it to our OEM partners. As you can imagine, it's not just the GPUs coming in; a whole bunch of networking equipment needs to arrive, too. We believe we will scale up with the customer demand. One important aspect is we have been struggling to meet the compute demand from our end customers. The U.S. is investing hundreds of billions of dollars; we don’t see that kind of investment within this region. Yes, KKR recently acquired STT for $10 billion; however, scaling data centers requires much more compute. Therefore, we've decided to build our own capacity using modular technology, and we are currently targeting over 600 megawatts. We are hoping to complete all signings by the end of this year, and will go into full-scale production towards the latter part of this year. We’re anticipating creating a new market that currently doesn't exist.
And regarding the strategy to buy and build some of your data centers, you previously mentioned in your business update call in January that you were attempting to lease out any available capacity you could in current colocations across the regions. Any updates there on newly executed leases?
Yes, we have signed quite a few deals in the region. It's fascinating, but the reality is, as I mentioned, the available capacity today is limited. You are spot-on, so we will build new capacity and deliver infrastructure to customers ourselves. I've perhaps not made this clear previously: our demand exceeds hundreds of megawatts. Asia, as a whole—this applies not just to Southeast Asia or East Asia or South Asia—does not have the necessary capacity right now. For instance, India has only about 1 gigawatt of fully utilized scale. Recently we held an AI Summit in India, which is exhibiting expansive deployment. However, there are other structural issues pertaining to power, water, and so forth. Thus, we are cooperating closely with the Indian government to assure that we get our infrastructure ready to meet various demands and architectures for edge deployment. To sum up, we are definitely making progress.
Adding on that, when we’re reserving or lining up capacity or constructing it ourselves, this is a different— we’re not in the business of building scale and hoping customers come around. Here, the business focuses on purpose-built, AI-focused data centers or GPU-as-a-service. This means that we will not invest capital until clear customer demand surfaces. The second aspect involves demanding customer prepayments, which reinforces our position. In many cases, the customer prepayments form an integral part of our financing strategy, allowing us to secure 90% or more of a project’s CapEx costs. Therefore, when customers demonstrate commitment, it generally makes us more confident to move forward and certainly increases the likelihood that the economics favor us.
Some things changed over the weekend. I was wondering if you could provide any update on operations or outlook for the Middle East given the Iran-U.S. situation?
Mr. Roy, thank you for your question. First of all, to everyone listening, I'm genuinely sorry for what's happening. My heart goes out to the families caught up in this situation and to anyone who's lost a loved one. I'm feeling very, very deeply saddened. I have friends on both sides of the pond. From a business perspective, John, we’re monitoring the situation closely, and as you know, we maintain a very disciplined risk posture. Currently, we’re not seeing any material impact on our operations. Egypt is progressing smoothly; our delivery continues as planned. Across the region, we are executing with appropriate caution, strong compliance, and clear operational controls. The practical factors matter to us, such as logistic routes, supplier lead times, local security conditions, FX exposure, collection cycles, and any regulatory changes that could impact movement of goods or personnel. If anything changes, any impact would likely manifest in timing rather than demand; we will respond quickly to protect the quality of delivery and keep everyone updated with something definitive. The trends favor us very strongly, and they are accelerating, not slowing down.
Speaking of trends, you talked about AI in India. Can you take a step back and provide your macro thoughts on the AI environment? What do you see happening?
Sure. That’s a good question. Many people keep asking me, and I’ve spoken about it at various events. I would divide this into three different trends. First, AI is becoming a national and regulated infrastructure. Governments, telecom operators, and regulated enterprises are now treating AI compute as strategic capacity tied to sovereignty, data residency, compliance, and critical services. This transforms demand from optional pilots to targeted programs with long-term intent. Take a look at Asia; they are rapidly establishing serious plans to avoid falling behind, thereby allocating larger budgets with a long-duration intent, as I mentioned earlier. The second side to this is the center of gravity; I must emphasize this point. The market is misunderstanding this completely. People are discussing if investment into AI will sustain; companies are investing hundreds of billions in the U.S. and in China.
The center of gravity is shifting from training to inference and from inference to distributed inference. Training is very lumpy, while inference is persistent. You have to grasp that. Inference penetrates into everyday workflows, which disperses compute demand across regional hubs. This necessitates more build of regional data centers. It will not slow down—it’s only going to rise exponentially. Lastly, edge is broadening the addressable market tremendously. It brings AI to crucial decision points where latency, privacy, and resiliency matter. This means increased adoption across public safety, transportation, telecom networks, logistics, and industrial operations. These changes will not replace data centers; rather, they will augment them by creating more endpoints that require regional capacity and orchestration. Consequently, in the future, expect to see considerably more distributed inference points, leading to a greater need for regional capacity.
That’s why firms like OpenAI, Meta, Google, and others are transitioning toward distributed environments. All these trends distinctly favor us, and they are not decelerating; they are indeed accelerating.
Jay and Bruce, congratulations on the transformative 2025. I just had one question regarding quantum-safe networks and your SD-WAN product. Can you share some concrete milestones that investors can look for?
Sure. As I've mentioned previously, we have created a very strong product, and we have effectively tested it over the last few months. Roger's team is confident they will be able to launch it by the end of April 2026. When we deploy AI infrastructure, we're not just dropping GPUs in a room; we're talking about secure connectivity, telemetry, orchestration, and compliance layers. These key components must be highlighted. We’re not merely selling or renting hardware—we're providing a service. Our SD-WAN plus our quantum-safe encryption allows us to control the network edge to the core securely. That enhances our solution’s value and improves the margin mix. Secondly, regarding our quantum solutions, some may believe we chase it merely because it has the term 'quantum' attached to it. But this is not the case. In fact, quantum solutions render edge AI viable. Edge compute operates efficiently at scale, but connectivity must be intelligent and secure.
So our SD-WAN does precisely that; it provides traffic optimization, segmentation, and performance control. Moreover, post-quantum crypto future-proofs the transport layer. Once we construct the transport layer, it will secure that and along with a distributed AI architecture—this will effectively deploy in both national and enterprise environments. This positions us not as a compute-for-rent provider, but rather as a trusted operator. It allows us to offer sovereign-grade, quantum-safe, policy-compliant AI networks. More importantly, our SD-WAN ensures seamless functionality even under complexities, which is crucial in our current global environment.
That's very helpful. And congratulations again.
This concludes the question-and-answer session. I would like to turn the conference back over to management for any closing remarks.
Thank you very much, Crystal. That was helpful. Some very impactful questions caught me off guard as well, which is interesting. To all our investors, analysts, and everyone supporting Gorilla, thank you. You have trusted me, us, and the entire Gorilla team long enough to let results replace speculation. There are individuals out there who claim our contracts and numbers are garbage; that's alright; it’s speculation. We are constructing the AI infrastructure that governments and critical industries rely upon, and we intend to execute with discipline. To everyone who knows me, they know I execute with discipline. Thank you all, and I’ll stop here before my tea gets cold—it's 5:25 a.m., and that would indeed be a genuine crisis for me. Thank you, everyone, and have a lovely day.
Ladies and gentlemen, this concludes today's conference call. Thank you for your participation, and you may now disconnect.