Gorilla Technology Group Inc. (GRRR) Earnings Call Transcript
November 19, 2026
Earnings Call Speaker Segments
Thank you so much. Good afternoon, ladies and gentlemen, shareholders, analysts, and everyone who has joined us today. I appreciate your being here. I'd like to open today's discussion with a point that appears to have gotten lost amid all the recent commotion and enthusiasm in the financial markets. Pats, real figures, and a bit of arithmetic, if you'll forgive me. We've seen remarkable volatility, geopolitical uncertainty, semiconductor control disputes, and more speculation than can keep the financial press going until Christmas — and somewhere amid all that excitement, a fairly significant development seems to be getting too little attention. Gorilla Technology has just posted the biggest revenue quarter in its history — not a forecast, not a memo for interpretation, not merely one more nicely illustrated PowerPoint deck; and I want to explain what we've accomplished, why it matters, and why I believe we remain in the early stages of building a far, far, far bigger business. Let's start with the figures. In Q3 of 2025, Gorilla brought in approximately $26.5 million in revenue. In Q1 of this year, we delivered around $28.2 million; in Q2, we delivered approximately $50.1 million. And now in Q3, we delivered a record $69.5 million. That's 163% growth year-on-year and nearly 39% sequentially. We beat the upper end of our own quarterly plan range by slightly more than $19.5 million; let me say that again. We expected up to roughly $50-plus million and delivered close to $70 million. I realize the market has plenty of great things to worry about; but surely 163% revenue growth warrants at least a short pause in the national anxiety; for all of 2025, Gorilla generated approximately $101.4 million, and our minimum revenue outlook for 2026 is roughly $200 million. That would mean almost doubling the entire revenue. Our target for 2027 remains $450 million to $500 million; those are ambitious figures. They demand execution, and they certainly aren't mere guarantees; but I'd urge investors to look at how the business is progressing rather than just the share price's movement on any given afternoon, week, or month. The share price tells you what the market thinks today. Our financial statements tell you what the business is actually delivering and has to live with. And sometimes, ladies and gentlemen, those 2 stories seem to be reading from entirely different books. One of the stranger things I've noticed is the tendency to analyze Gorilla as though we were set up last Tuesday to rent out a few NVIDIA cards, and let me correct that notion. Gorilla has an established business, delivering security intelligence, network intelligence, AI analytics, systems integration, and critical technology infrastructure. Most of our current revenue continues to come from these broader technology and infrastructure operations. GPU as a service remains a relatively small contributor today. Egypt still remains an important customer; we continue to deliver against our contractual commitments. And importantly — most importantly — we're collecting cash. Since September closed, roughly more than $22 million has come in from Egypt, and we're currently seeing additional collections of $15 million to $17 million. Over in Taiwan, we just locked in another sizable multiyear national intelligence contract worth tens of millions of dollars, with rollouts scheduled over the quarters ahead. Our established technology operations keep running in other markets too. These businesses haven't simply vanished just because the market has latched onto a new acronym, GPU, as a sell signal. In fact, they're a key foundation for Gorilla's next stage, though our strategy isn't to own the GPU itself. What we want is to pair computing infrastructure with software, security, data intelligence and all the related services. GPUs supply computing power, but the real question is what you do with that power — and that's where the broader opportunity gets really interesting. Gorilla plans to play in both areas, which is why we're building that infrastructure program. Now let me tackle the big elephant in the data center. Everybody wants recurring revenues. And apparently everybody expects the whole process to wrap up before the next earnings call. Unfortunately for me, the laws of physics stay stubbornly resistant to financial modeling. You can't order 5,000 GPUs on a Monday morning and expect a fully operational AI cloud by Friday. The equipment has to be manufactured, shipped, installed, powered, cooled, networked, integrated and tested. Then customers finish acceptance and start using their capacity. Only after that do the relevant recurring revenue economics take shape. Take Yota, for instance: we've already delivered 512B 200 service — yes, delivered. And 128 are scheduled to complete our plan of 640 service in the Phase 1 deployment, which represents 5,120 GPUs. We're also moving forward on the considerably larger B-300 programs, and that schedule is complete — to be complete by the end of December. Now I want to stress something. A server delivered isn't necessarily a server commissioned — and a server commissioned isn't necessarily a server accepted. And a server accepted definitely isn't automatically generating cash at full utilization. Between those stages sits an entire commercial and operational life cycle. And interestingly, an AI chip may generate a poem in 3 seconds; sadly, it can't commission its own data center in 3 seconds. So we're working through these stages systematically. As the infrastructure enters service and customer workloads ramp up, we expect recurring compute revenue to become a more meaningful part of our business — and that's exactly how we intend to reduce the lumpiness of our revenue over time. Now I know the analysts will want to discuss gross margins with us, quite rightly. Our gross margin in Q3 '25 was 37.3%. This quarter it was 12.8%, but let's examine the sequential performance. During Q2 of 2026, our gross margin came in at negative 4.2%, and we lost roughly $2.1 million. Then in Q3, gross profit reached approximately $8.9 million, which works out to about an $11 million swing and a 17 percentage point recovery within just one quarter. We managed that while revenue rose by nearly 39%. Delivering hardware and mobilizing infrastructure comes with economics that differ from recurring compute and managed services. Higher margins don't just appear once the equipment is bought. They hinge on commissioning, utilization, commercial pricing and running things efficiently. Bruce has told me that my optimism still isn't an approving account under standard IFRS. Unfortunately, I've appealed twice, and both of those applications were turned down. So yes, margins need to get better, and they will improve further. But an $11 million gross profit recovery in 1 quarter, I believe, is a meaningful step in the right direction. Our focus is on keeping that improvement sustainable. During Q3, Gorilla also spent approximately $218 million of cash on property and equipment. Our total property and equipment asset base now stands at nearly $250 million as of the end of September. We also ended the quarter with a little over $5 million, nearly $57 million of cash, while operating activities consumed about $30 million. These are substantial figures, and we take our financial responsibilities seriously.
AI infrastructure requires a lot of capital, but I want to be very clear about something. Just spending money doesn't automatically generate value. In my view, a data center full of expensive gear with no paying customers is essentially just a costly lighting arrangement; so we need to ensure our investments are commercially sound — properly financed, successfully brought online and capable of producing sustainable returns. Let me also briefly address receivables, where we're seeing real progress — as of the end of September, our receivables plus contract assets totaled approximately $203 million. The key thing for us to look at is that we reduced unbilled contract assets by over 50%, going from roughly $104 million plus down to $52 million, as we progress through project milestones at Bill. That improvement is also reflected in cash. We've already brought in significant cash from several other customers. As I mentioned, Egypt has been paying us more than 22, and we anticipate substantial customer collections — as we finish more contractual milestones, we expect to turn additional unbilled revenue into receivables and eventually into cash; so for me, the direction is very clear: speed up billing, reinforce collections, steadily bring down outstanding balances, and we are simply not generating revenue. Our attention is shifting more and more toward turning those revenues into actual cash. Alongside that, we're advancing arrangements for institutional and project financing that will back our infrastructure build-out. What we're after isn't merely a larger Gorilla — we want one that's bigger, financially more robust, and in the end enormously profitable. I'd be neglecting my duty if I didn't acknowledge the enthusiasm around semiconductors; all of us have witnessed remarkable market swings, policy shifts, and major regulatory changes affecting advanced computing. We treat this with the utmost seriousness; at this point the U.S. The Bureau of Industry and Security has put out further restrictions that touch advanced chip shipments as well as some remote-access setups involving data centers located overseas; with these changes, we really need to look closely at where equipment is deployed, who the customers are, and who the end users are. As far as we're concerned, our stance has been clear all along. You have to build compliance into the infrastructure from day one. That's the reason we've had export control teams in place for the past couple of years; we're tightening up our procedures for verifying customers, checking beneficial ownership, tracing equipment, handling things such as export classification, licensing requirements, access controls, continuous monitoring, and more. All of that is already underway. So none of this is new to Gorilla. Our business model is designed to be highly transparent, with customers who can be identified, infrastructure that is controlled, and clear accountability. Also, we will keep obtaining whatever authorizations are needed for the transactions involved. It is not my intention at all to spend $1 billion building an infrastructure business and then find out that someone raised a concern to verify who was using the equipment. So we place our faith in robust governance and credible compliance, which become a significant competitive edge when working with institutional customers and financing partners. And regrettably, the semiconductor manufacturers are not under our control. Clearly, we also have no control over the financial markets. What is within our control is the way Gorilla reacts, the way we deploy capital, and the way all of us carry out our work professionally. So where does that leave us going forward? For the rest of 2026 and on into 2027, our priorities are clear. First, infrastructure deployment is already in progress, and we need to finish it and put more capacity into productive service. Second, we need to raise customer utilization and grow the recurring revenue contribution from our AI compute business. Third, we will keep expanding our established technology operations, which will stay a substantial portion of Gorilla’s revenue. Fourth, we need better gross margins, operating performance, customer collections, and cash generation. And last, fifth, we have to strengthen governance, financial controls, and regulatory discipline—even more importantly, so that we can run a substantially bigger company. Our floor for 2026 revenue remains $200 million. For 2027, the revenue goal remains $450 million to $500 million. We understand the execution demands and the risks that lie between us and those goals; but please think again about what we have already accomplished: quarterly revenues are up 163% over 12 months; we have produced more than 11 million in sequential gross profit recovery. We have significantly grown our infrastructure asset base. Our established technology business continues to bring in revenue, and our GPU as a service business is still at a very, very early stage. My take is that this makes for a fairly intriguing mix for a business whose fundamental momentum the market at times appears to miss. Let me close with a point I hold strong views on. I know shareholders have felt let down by how the stock has performed lately. That frustration is something I feel myself. Even so, I will not conflate share price swings with what this company has actually accomplished operationally. Over time the two are linked. On any given trading day, they are not the same thing. Our revenue growth has been exceptional. We're putting a lot of infrastructure capacity in place, and we keep making the organization stronger—but none of that puts us above criticism. There are margins we need to improve, capital we need to manage, equipment we need to commission, customers we need to collect from and, even more importantly, profitability we need to show. We will be held accountable, and accountability cuts both ways for us. Should the market want to question their performance, I welcome that scrutiny, though my one request is that it start with the facts instead of finishing with the share price. Our team has put in tremendous effort in many different countries. I want to be sure I thank each of them, along with our customers, our partners and our shareholders. I should probably also thank my family, who likewise prohibited discussion of the globe's GPU megawatt data center at dinner. I've been contesting that ruling through all my prospects, though they appear to be losing, along with my IFRS application listing Bruce as a payable. So let me wrap up by saying this quarter is the biggest in Gorilla's history — and it's now behind us. We aim to make it a milestone instead of a high watermark. Thanks. That said, let's proceed to the questions you have.
I'd like to add my congratulations for a strong quarter, well executed. Jay, I'd like to raise a topic that wasn't in your prepared remarks but has been in the news lately — news about your company — namely the proposed purchase of data center assets based in the U.S. As I recall, it was only a preliminary review at the time the press release came out, so perhaps you could update us on where due diligence on that potential acquisition currently stands, along with any other details you're able to share?
Of course, Gary, and thanks again. Look, the recent regulatory changes carry plenty of implications. At the moment, we're taking a close look at the most recent rules issued by the U.S. Bureau of Industry and Security, and at what they mean for our global AI infrastructure initiative. Our stock has dropped — I won't claim otherwise. As a result, raising money has proven rather hard for certain other ventures beyond Yota and the other significant investments we're seeking to complete. Where we stand today is that, out of caution, we intend to go through every project that's been affected and evaluate the regulatory situation — while simultaneously cooperating closely with our regulatory advisers so that we remain entirely in compliance with all U.S. requirements. On the export controls point, once we've finished reviewing the matter, any significant determinations will be shared via an official announcement at the right moment. Is that clear?
It does. Thanks for that. And following up, I'd like to ask about the financing environment. I believe it's widely reported that debt markets have grown less accommodating when it comes to this kind of capital raise for this kind of capital expenditure. But perhaps you could tell us, of that $450 million to $500 million revenue expectation for fiscal year, how much is backed by what you currently have in deployed PP&E, plant, property and equipment, and by your existing capital on hand, likely sources of capital. And anything else you can share there.
Bruce, do you want to take that one? I'm sure you're eager to jump in.
Absolutely, and thanks for the question, Gary. The first item we'll cover is that Jay said we've taken delivery of 512 servers, which works out to roughly 4,000 GPUs, and we're in the process of setting those up right now. Those are designated for the primary offtaker at T1. Once billing begins, that primary offtaker would generate a little more than $5 million per month in revenue. When construction starts is something I'll defer to the commercial team, though it will certainly happen in the first quarter. The remaining 128 servers, meanwhile, would bill around 1.5 million per month, so that comes to $6.6 million a month that we expect to be building during the first quarter of 2027. All of that is supported by equipment we've either already received or have the resources on hand to pay for. Turning to the second part of your question, I'll start with the markets in general and then get into where we sit within them. It's no secret that markets have been somewhat tumultuous over the past few weeks. That said, the genuine chaos has been concentrated in the unsecured market for certain neo clouds. Everybody is talking about coiveunsecured and so forth. The senior secured market, meaning project level finance, has been somewhat volatile but relatively more stable, and where projects can show strong cash flow generation along with a comprehensive security package, financing is still available. If you look across the market, a handful of deals in the billions have closed both internationally and in the U.S. that were senior secured project level financings with solid off-takers. At the Yost project, we believe we have a very solid set of offtakers, anchored by 1 of the strongest balance sheets and operating performances around, and we're also quite confident in the structuring work we've completed. I know everyone is following this closely, so where things stand is that, together with a Tier 1 investment bank, we are working on a $1.7 billion bond, and I would expect it to close within weeks. The purpose is to refinance Jon, for which we've already deployed the capital, and then to finalize the financing for Yont2. I'd add, of course, that this remains subject to market conditions and final documentation. We're also working with an anchor investor who is a very well-known participant in this space. Getting this to the finish line has involved numerous work streams, and they've been considerably more complicated and, I would say, more stressful because of market conditions. Everyone wants to triple check their work. Every 2 weeks there's either market demo, announcements from the export control authorities, or something else that makes it difficult. So, but it is very much on track, which is how we would characterize it. So that EUR 1.7 billion would cover the financing for every one of the Gott projects, and once the Yolt project is deployed, it would take us to the guidance number. Okay. That's really helpful detail. Appreciate — if I may answer that.
Sorry, may I tack on one quick point there? Bruce mentioned that $1.7 billion financing program and how it progresses; obviously this is a bond program; that being said, what I think, Gary, I would also like to add is that up to now, we have actually deployed $331 million in total payments, okay? That covers all of the networking, all of the prepayments, the advanced equipment tranches, and so forth. And more importantly, within the next few days, we are not even 2 days, just a few days, perhaps, Matt, we expect to receive another $130-plus million in relevant payment milestones from the customers as well; so we are pushing forward, and whatever liquidity we get from our customer payments as well at the same time, we will keep honoring our contractual obligations to Yota and all the other projects as well.
Could you break down how much CapEx, GPU purchases included, is still left on YODA and on UA2, each on its own? And once project financing is in place, how soon do you think YODA 2 could be up and running — sorry, those are my first 2 questions.
Perfect. Bruce, would you like to handle the first half? I'll take the second half.
Right, so on the 512 servers, one installment payment of EUR 50 million remains. Then for the 128 servers, another EUR 50 million is due, which means EUR 100 million on Yolt, and we're comfortable covering that mainly through customer prepayments. After that, moving to point 2, it's $1.5 billion. Then, as I mentioned, we're working on a bond program of $1.7 billion, so that would comfortably cover not just the CapEx for GPUs and networking but also any OpEx reserve or DSRA we need, and in fact there would be some refinancing with money coming back from the project level to the parent company level. That's the outlook there; and then, Jay, the second part of the question goes to you.
Sure, Brian. So the one key choice here is that OTA is already fully — it has shipped and is currently undergoing testing, and we'll finish the entire deployment by the middle of December. As a matter of fact, we began seeing revenue traction already last month — or this month, I should say, and we'll keep bringing in revenues during the quarter this year. On Yota, we'll be at full capacity by January of next year. Then for year 2, the deployment of the B-300, that comes to 2,736 V300 GPUs. The point is we aren't beginning from zero. Every one of the broader deployment programs is already finished. The RFS for 3 of the 5 floors is done as well. So we're proceeding with FMC, our partner, which has already begun production, and over the next 30 days we'll be deploying our first networking equipment. All deliveries are scheduled — again, contingent on the planes being available — for completion by the end of December. So the main takeaway is that we have the revenue transition you asked about: OTI will move into full revenue by January this year, since all deployments will be finished by mid-January and January, with full deployment by February. Our priority is deploying those GPUs as fast as we can, so teams are working around the clock — the SMC team, the Yoder team, Gorilla, which is a big team in Mumbai and in Delhi — and at the same time the NBIS team, the NVIDIA services team, is also working very closely with us to close everything out by the end of this year. Brian?
Just to follow up on U2 — I'd assume that until the financing is in place, you can't buy the GPU — so how long is it — what's the lead time at present, once the financing comes through, is it weeks? Is it a month or 2? Maybe just walk me through that, if you don't mind; then I have 1 last follow-up.
Again, Bruce, would you like to take the first half of the financing of the delivery process.
We coordinate quite closely with those suppliers, so it isn't as though we're talking to them all the time. What we do is match the delivery timeline — or the production timeline and delivery timelines — to the financing timeline, so you won't see a gap of weeks. In the case of Hilton, it was largely a matter of us telling them the date we expect to pay; they get production ready; we pay and they ship, and in some cases that happens within a couple of days. So it's hard to be precise. So it ought to move very fast. That's the reason we feel confident about the timelines we provided for the financing and afterward for the deliveries as well.
Great. I have one final question. Sorry, Jay — did you want to say something? My apologies; it's possible that's the case.
Yes, Brian, I did. Sorry — consider how the catalyst is financed for releasing and delivering the equipment, rather than for beginning the manufacturing process, all right? SMC is clearly a partner, and simply having a nice photo is not what we plan to do with Charles. What we have done is actually pay them what you might call an enormous amount of money. And they have already received the advance payment as well. They've actually begun delivering some of the networking gear for all the RFS that's needed — I was over there last week too. We finished the RFS, the data center evaluation; that's why I mentioned there are 3 floors that are fully ready, and 2 floors that'll be completely fitted out by the end of this month. The age wand cooling for those 3 floors is now in place as well. So what SMC has been through is that they've kicked off manufacturing, and it's well underway; and as soon as we finish our testing — remember, every single piece of equipment that comes out, we test it. Once testing is complete, we get final approval from Raj, our Group leader/officer, and after that we release the funds and send it out. Normally, since we use chartered aircraft for these shipments, delivery generally takes just 24 to 48 hours. I hope that as you train.
That's genuinely helpful. I have one last question: on Neutral DC, you brought up the regulatory changes that are being talked about at the moment — where do you come out on that? Is it parked for the time being, until you can file cleanly, with Yoda 1 and 2 being your main focus? Could you take us through it? And has anything changed in how approvals are required going forward in India? I presume not, but I just want to check whether there's been any change to the scrutiny or the process.
Excellent question. Forgive me if I come across as somewhat diplomatic. As things stand today, we've adopted a precautionary step. Nothing has been cut off, ruled out, or paused. What we mean is that for the projects in question, we are now accepting complete commercial commitments—largely neutral DC, as an example. We are collaborating extremely closely with our legal and regulatory advisers, and we have brought on a couple of them in the U.S. We're being careful not to lock ourselves into anything before we have complete clarity — and of course, the BIS has put out a 132-based document that remains rather unclear, so we want to avoid landing ourselves in any kind of trouble — and Brian, another thing you'll notice is that we already maintain a sizable export controls team, one we've been relying on all along. So none of the Yota implementation is touched by this specific BIS rule; it could, however, carry some consequences for the neutral DC. Even so, we will make absolutely certain that we stay committed to pursuing every one of these infrastructure opportunities. I do want us to have full clarity on BIS compliance. When that moment comes, I would prefer to return to you, share an update, and tell you exactly where we stand on it. As of today, things remain — I wouldn’t call it business as usual, but this is a precautionary step.
Hello, everyone. I have a question about the possible repayment impact on the convertible, in particular how you intend to repay this note. Will that come from project financing proceeds, or from cash on hand, and can it be done ahead of conversion at a discount to market becoming available in late December? Because if I remember right, and please correct me if I'm mistaken, if project financing has slipped partway to date, would you settle conversions in cash as the note permits at a 5% premium purely to limit dilution? I'm just curious about your thinking here.
Yes.
To clarify how this instrument actually works: we drew down EUR 25 million, and the face value we’re required to repay is $27.75 million. Should we fail to repay that face value, the note holder then has the right to convert into shares over time. There’s also a one-time option for us: once the registration statement for the NOCs becomes effective, we can repay the notes at face value plus 5% and avoid conversion of the note. That’s our base case for handling the note — upon effectiveness, we’ll be in a position to pay it off. Jay mentioned that we are imminently in line for up to $130 million of prepayments from our customers, and that will form the liquidity source we use to pay this note off. As a second step, if that isn’t available to us, or for whatever reason we choose not to use it, then there’s the project financing — we’ve already put $330 million into the Yota projects — so we’ll be able to take some of that back once the project financing clears, and that can be applied toward repaying this note. And then last but not least, if market conditions are such that, for whatever reason, we’re not able — the noteholder can’t convert it all at once; they can only convert $6 million a month. So we have several months before it’s fully converted into shares. So on balance we thought this was actually a very good outcome for shareholders; the headlines might be scary, but when you look at how the instrument works, it’s fair for us and for the note holder. One thing I’d also add is that if we do need equity, this facility provides for up to $100 million overall. So rather than having to go to market in uncertain times and wonder where we get the money and on what terms and so on, this is already lined up; and then we can draw on it several times and then do the same type of repayment at a small premium. So our philosophy is always to be overly cautious rather than hoping that everything works out. I hope that answers your question — add something to that as well, Steve.
Yeah. One thing I can tell you is that this has pulled in a great deal of market attention. I'm not sure why? Because that range bridge financing is not a preferred permanent capital structure; and the market, I believe, has to understand that, since the knee-jerk response has honestly frustrated all of us. So we obtained initial funding based on our intention, which Bruce and I have stated very clearly, to pay off the outstanding obligation in cash during the coming 4 to 5 weeks, okay? As the customer receives become available. Now we're gathering cash at a very fast pace. And this comes from all our customer prepayments, our customer those payments, any deals, we've managed to collect everything. The market, in my view, isn't grasping something important: people should examine the cash sitting on the balance sheet and recognize that this company has the means to pay. We'll ensure every contractual repayment obligation is met. And more importantly, our goal is to seek a resolution that keeps their equity dilution as limited as possible going forward. So we'll make certain that we collect from every one of our customers without fail and get this repaid. That's essentially what Bruce and I are working toward. Honestly, it bugs me that the market just doesn't get this at all, and you see this huge reflexive reaction around dilution and all that sort of thing. I'll hand it over to you.
That's really useful. As a follow-up, on the U.S. data center and the tenant there: where does the acquisition currently stand relative to the end-of-November closing target — and do we also expect the planned project financing to fund the initial GBU deployment at that site? And how advanced are the talks with a tenant for that capacity?
No, that's a very good question. So we're still very keen on setting up our first AI infrastructure in the United States. We've been making progress in discussions with the facility owners, or potential commercial partners, and so on and so forth. But the difficulty we face — and let me be very honest and open with you here, okay? — is that the share price has generated a lot of promotion for us, because we aren't able to raise capital effectively and efficiently. Okay? We're going through the motions; we've put in the offer, and by the way, we've taken an extension. I don't think we'll be able to close by November. Our share price is in the dumps right now; and unfortunately, I have to ensure the company isn't overcommitting and overburdening itself; so we'll make a decision based on how our share price improves over the coming few weeks. But right now, there's an extension through the end of December, and we'll make a decision — if the share price doesn't move, then we'll have to take a radical decision on whether we actually proceed with the transaction or not. But we're hoping the tap price will keep improving, because we've continued to deliver. Our intention is basically to secure the infrastructure, the GPU deployments, and so on and so forth. And we're already in conversations with a couple of offtakers who are willing to take the entire 6 megawatts going up to 36 megawatts straight away. I hope that answers it.
There’s one more point to add here, Filter: the project financing we’ve been discussing applies solely to Yota, up to $1.7 billion. So that’s — we’ve also held separate financing discussions beyond that. But I don’t want to give everyone a play-by-play of each financing conversation we’ve had for every project we’ve announced or haven’t announced, right? We want to stick to the high-level points that matter and could potentially move the market. We’re not here to share what amounts to my daily log of calls with banks.
Nice work on the solid growth this quarter. So regarding Yoda programs 1 and 2, are you still expecting those same EBITDA margin targets you mentioned a couple of months back?
Yes.
Yes, absolutely. We get the quick answer yes.
And regarding India, on those India projects, what sort of tax rate do you expect there?
Yes, so we've been working with tax advisers in India and in Singapore. Singapore is where the assets are actually held, in the Singapore SPD. It's the first choice of the financing counterparties, which is why we chose that. And a couple of announcements have come out recently, like in the last 10 days, about different tax treatment for GPUs, especially for financing of GPUs, where there's a blend of onshore and offshore capital. So I wouldn't want to give a definitive answer at the moment. I think the overall group tax rate is sort of around 20%, and I would direct you that it would be around the group tax rate. But hopefully we can deliver an answer that is optimized and improves on that.
Let me expand on that too. Mike, domestic companies in India usually face a tax rate of about 20 — somewhere between 20% and 22%, all right? But concessions exist. There's something known as the concessional regime. Don't hold me to this, but I believe it's Section 115 or 116. I can't recall exactly; typically, though, you'll land between 22% and 25%. With the new GPU structures, you have 2 options for doing this. You set up what's called an SEZ structure, or what's called the Gift City IFSC structure, all right? Now, something has happened over the past few weeks that's genuinely exciting for us: the International Financial Services or IFSC, as it's known, the Indian regulator has officially recognized GPU and associated data center equipment leasing as an eligible financial activity inside the IFSC. So we've already locked in the SEZ for all of our withholding taxes and our corporate income taxes and so forth. We've also secured — we're currently in the process of establishing our IFSC, and this only happened about 3 or 4 weeks ago, just a day or 2 before devalue. And that enables — this is quite interesting. From a tax standpoint, qualifying companies like ours will gain from what I'd call a substantial income tax holiday, partially covering roughly around 20 consecutive years within a 25-year span, all right? So — and then afterward — per the framework, I believe once the 20 years pass, the interest rate — sorry, the proposed tax rate is roughly around 15%. What's also quite interesting is that we have — our structure sits in Singapore. Now in Singapore, we've been working only with the ED and the — all the top on shows in Singapore, to build a very, very significant capital efficient, capital investment efficient, tax optimizable structure. So the free trade agreement between India and Singapore and especially India and the IFC as an IFCIndia and Singapore operations, lets GPUs be owned and hosted in Singapore while leased or rented to India, and that gives us a chance to build a very efficient financial structure. So on taxes, we're aiming to bring that down to about 2% overall, all right? And in Singapore, some taxes will apply across a 5-year period. The upside there is that we could write that off against certain other projects, which we're clearly securing as we speak. I hope that answers your question, Mike.
So would it be reasonable to describe 20% as the worst-case tax rate?
Yes, that is genuinely the worst word, worst case.
Got it — and has anything new come up on your Kora project?
A lot, actually. I was starting to wonder when you'd ask about it. So let me lay out where things stand — and let me give you some numbers on Kora, because I think the two of us have already gone over this in our Q&A. Our Cohort project has really come together much more than we expected. As you know, we've secured over 142 rights, roughly 60-plus acres of land at the center. Power remains 1 of the biggest constraints for any infrastructure build. On Kora, I'm glad to report we've locked in a 280-megawatt power allocation with full redundancy — that means two separate 280-megawatt supply configurations, giving 560 megawatts of total provision capacity. On fiber, we have 6 drop lines; we also have roughly 18,000 cubic meters of water allocated to us each day, 18,000 cubic meters per day. Even bigger: 30% of the energy delivered to us is green. And most important: both the local authority and the national grid levels sign off on all of it, which adds resilience so the entire infrastructure campus gets electricity in full wherever it's needed. We're also watching major development milestones as we build this app. At the moment we're nailing down the core infrastructure required to build and operate at stage 1. We've made progress on environmental testing too, which we finished in the last 45 days. Environmental testing, ground analysis, utility planning — water and fiber are being dug up and examined right now — and we're moving forward on choosing engineering and construction partners. We've invited around 14 of the top players across Asia to come in and submit a response, and at full development and utilization, our current illustration projections point to potential annual revenue of roughly about USD 4 billion. You might look at this yourself, Jay, and think it's an unusually large opportunity you're chasing — but whether it pays off still hinges, in my view, on landing the customers. So let me say right now that we aren't simply putting capacity in place and crossing our fingers that buyers show up. We're already talking with several of the biggest customers anywhere in the world. American — and I want to be crystal clear on this, American — and the task now is converting that demand into signed revenue. And Mike, I'll repeat once more, I want this to be completely clear. We don't make a practice of constructing a data center without having our customers lined up first — before any ground is broken, the customer will already be locked in. And as I said earlier, Cord is, for us, a long-term infrastructure asset. Our target is a 50-megawatt footing: we're putting up 50-megawatt data facilities with capacity fully commissioned and customers under contract. One more thing that might really surprise you is just how many people we've brought on board — the team in Thailand has already exceeded 75, and by Q1 or Q2 of next year it ought to land somewhere in the range of 150 to 200. Every one of those people works on land, power, engineering, and utility grid capacity engineering, and each was hired to help make sure we succeed. And when everything is fully developed and utilized, our present illustrative projections point to potential annual revenue of roughly about $4 billion U.S. Mike, over to you.
Great, that's helpful. Broadly, I believe this will be my final question. Given the political change in the U.S., there appears to be somewhat greater certainty surrounding data centers overall. So does that create extra demand in India, beyond what you had previously — are you receiving more inbound calls, or something like that?
Mike, your questions are always a pleasure. The new BIS rules do affect Thailand, easier Singapore, you know, every country we're discussing. Yet we're taking them with the utmost seriousness; even so, I want to make sure I'm very clear here. The following is a compliance requirement—it does not forbid us from constructing air infrastructure in Thailand or Indonesia. And Matias, no construction is being stopped on my end, since nothing improper is taking place; we are not illegally moving GPUs into China. That is not what we do. As I said, our core development stands at 200 megawatts. That said, we have additionally signed up 40 to 50 megawatts of extra capacity. Right now, we are looking at Indonesia for a further 300 megawatts. Our planned commercial model is simple; we have customers and offtakers—including U.S. offtakers—who are properly screened, we will screen them, controlled in structure, and we will prohibit any remote access to or diversion of GPUs. So Mike, what we've done at this point is build a really strong compliance structure that addresses export licensing, verifying customers, beneficial ownership, GPU traceability, physical security, remote access controls, independent monitoring, and everything else along those lines; so let me come back to what you asked. I realize your question was about whether this generates demand, but we're securing every license we need, confirming identity, and so forth. Honestly, I'd prefer to put money into compliance immediately, which is what we've already been doing, instead of having to explain to Washington why someone failed to verify who's using our equipment; and as for demand, India clearly stands to benefit. The India situation also presents a difficulty—I believe the earlier question concerned taxation. Handling taxes there is an enormous undertaking. To give a sense of the scale, I've brought on 30 finance staff, purely to handle the tax arrangements linking Singapore and India, all right? The structure is extremely intricate, and we must ensure we can meet whatever the authorities on either side require, along with what the U.S. government needs. So India will face the difficulty of securing very large orders, though in my view building out the infrastructure will require considerably more time while the momentum carries on. Other regions are clearly going to keep drawing market scrutiny. You've surely noticed players such as Microsoft, Google, Oracle and Meta. Every one of them — Anthropic too — is paying serious attention to Southeast Asia. And among those here, we're likely one of very few aiming to turn that model into a commercial offering. Mike?
Bruce, congratulations on the progress. Turning back to Core, our assumption is that construction there gets underway within the next 2 to 3 months? Do you figure that window still holds up?
Yes. Yes. That's quite reasonable. So, as I said before, a substantial amount of work is already done, all right? The 1 key thing to factor in is that every bit of testing analysis and utility planning is finished — all the water and fiber requirements are done. So at this point it's about ensuring we build the substation; just so you know, we've more or less begun constructing the substation already. There are 2 substations, like I said, 20 to — and what we're also observing is considerable demand coming from the offtakers; the 1 point I raised earlier is that our development strategy splits into 4 data halls at 50 megawatts apiece. We have enormous amounts of land now. We could actually turn that into the 6 data halls of 50 megawatts and scale it up, since power availability is rather high. But we're also ensuring that our planning through to execution wraps up within the next 3 months. One more thing you'll notice, Michael, is that we aren't doing a traditional old-school build, all right? Ours is a hybrid build. You'll see us deploying hybrid modular technology that we're quite proud of, and we're actually considering bringing it in-house too, the modular technology. And I want to be sure we can complete construction from the time of build within a span of 6 to 9 months, rather than taking 2 to 3 years. Very interesting. And on that — go ahead, Brito. One more point to add, Michael, is that the preparation, the site acquisition and prep, hasn't cost us an arm and a lake. Getting from land acquisition all the way to a site ready for building should run about $14 million, $15 million all in on our side. That's why we jumped on the project — it's obviously a massive opportunity, but much of the infrastructure, et cetera, was already in place or there was a fairly straight line to getting it in place; that's why we see the opportunity.
That's useful, Bruce. And on the CapEx side of the hybrid modular design, the speed involved and the cost savings too are quite interesting for us. So do you still expect something in line with roughly $6 million to $8 million per megawatt?
Mike, I love that. So let me give you the high-level picture on this, okay? One thing you'll have to factor in is that the engineering approach differs very, very, very substantially. The prefabricated data halls, what you're looking at there is factory-built modular data halls, standardized back layouts, containing we and high-density GPU infrastructure. When it comes to electrical systems, we produce medium voltage switchgear, transformers, the full range of UPSs, backup storage, prefabricated distribution, kids, and more along those lines. Then on the cooling side, there are all our factory-assembled cooling distribution units, liquid cooling, lubes, heat exchangers, central cooling plants, and the high-density B20 designed for high-density B20 deployments of 300-plus GPUs. We also build the redundant independent electric supply portion, so it arrives with what I'd describe as an N+1 or a 2 architecture for the critical systems. And here's the part that will surprise you: our cost per build comes in at USD 4 million to USD 4.5 million per megawatt, rather than 6 to 7. The total spend for the complete 200-megawatt buildout comes to somewhere in the range of $800 million to $900 million. That figure would encompass every bit of the civil works and foundations, the building's electrical infrastructure, transformers, UPS, the full set of liquid cooling and mechanical systems, network security controls, contingencies — what else am I forgetting — plus module fabrication, integration, and installation.
Really interesting, Jay. So covenant PAUSE came in below what we expected, and that's on the favorable side. One last question. It's somewhat high level. Given the increasing emphasis and attention on cybersecurity — what are you hearing from customers? Do they want to add Gorilla's cybersecurity capabilities as a managed service alongside GPU as a service?
For sure, Michael; that's 1 of the most exciting opportunities we have right now, because cybersecurity and AI infrastructure personally, when we speak with customers, are growing more and more interconnected — now what we hear from customers, especially, you know, we work very largely with sovereigns and governments. So they come to us and say, we're not looking anymore at computing capacity. They want all of this to be secure, resilient sovereign infrastructure. Now we are a cybersecurity video intelligence network intelligence community, right, out of a company. What we have — if we've already got this operationalized, we have the experience. We have all the relationships and we're expanding our footprint. So what that has let us do is create additional revenue targets; now a lot of people kind of list the fact that we actually invested in the company called Astros, by the way, in India; that lets us actually expand our capability to provide recurring software and managed services to sit on top of the underlying contracts. So you will see — that's why in my first introduction, I'd mentioned that Gorilla is not just trying to build GPU as a service. Our underlying business is what people seem to be forgetting; and those commercial discussions, by the way, are already in place. They're creating huge opportunities for us for a fully integrated solution; and that is a model, Michael. We're just not simply looking to rent computing power. We're building the AI infrastructure ecosystem and creating multiple recurring revenue streams. I hope that answers your question.
The majority of those points have actually been addressed already. Perhaps just a few for me, if you don't mind. Since Jensen was a major backer of start-ups and the largest deporter of AI trade build-out, — do any concerns exist for Gorilla regarding the future references you might receive? And what is the media relationship like since this came about? That's the first one.
I'll attempt to respond to that in a very diplomatic manner, alright. To begin with, our relationship with NVIDIA continues to be important for our AI infrastructure strategy. And we will keep engaging within the BD ecosystem and with all of our equipment partners, whether that's Dell, Super Micro or HP and so forth. Now I would like to draw a distinction between the wider industry commentary regarding AI financing and how Greta's business fundamentals are going to be impacted and some folks will begin speculating. We aren't building a speculated business. We are collaborating very closely with NVIDIA. Incidentally, NVIDIA services are still sort of running operationally on the ground. Nothing has changed following Jensen's departure. And we are placing increasing emphasis on all the principles that underlie the data center and the projects. Now naturally, we don't take future allocations for granted; and we are certainly not saying that NVIDIA has made any commitments beyond what has already been agreed. But what has been agreed, they are fully honoring and it is heading in the right direction; in fact, we're pushing forward even more quickly than before, as I mentioned on our previous earnings call. Now having said that, we are also examining alternatives. I mean we have already been examining AMD. We met with the whole AMD team. We're working on them; we're trying some POCs. And we anticipate that the Helios will also have a significant effect on some of the future allocations as well. So we are going to be working very closely on both of them; and we're also — we've had some conversations with the ceteris team, which is also actively engaged with AMD. So there are multiple conversations we are having simultaneously. so that we can deal with our business as well.
The following question concerns what's...
Sorry, one point I'd like to make is that my microphone does lag on me. It was, sort of, made public that certain circular financing arrangements are being scrutinized. It's worth noting that no circular arrangement exists in our contract with NVIDIA as a customer. There's no back star, NVIDIA has made no investment. Nothing there resembles the sort of circular financing currently under scrutiny. So, in that respect, I think we have less to worry about than other players would.
Understood, that helps. My second question concerns what your backup plan might be — if the worst case occurs and the $1.7 billion bond financing you're working on right now fails to close, do you have any thinking on what you'd do, or anything you're able to share?
Would you like to handle that one?
Sorry, could you repeat anything the first.
The basic question is this: if that $1.7 billion bond program doesn't go through, then what? What's your plan B?
Discussions are still ongoing with 2 private credit players.
All right. Okay, fine. So I suppose the final item is cost keeping for the model — apologies, please continue.
Just to build on what you were saying, correct? What I mean is the project financing comes through what we — the traditional bond growth, right? Bae had over 330 million plus invested, and we're carrying on with the conventional model; but at the same time we're collaborating with a couple of other institutions on long-term associated contract financing as well; we're also in talks with banks and other strategic capital participants who are able to offer alternative equipment financing too; so we are definitely not sitting here and putting with atoms. I assure you of that; but we're working very closely to close our bond program. Naturally, it's run by the largest institution on the planet. But we'll make sure we have plan B and C, Plan B and Plan E simultaneously; so we're working on all plants at the same time in Para.
All right. All right. The final 1 is really just housekeeping on the model side. With several projects underway and each at different phases and on different time lines, what ought we to keep in mind regarding the gross margin and EBITDA margin range tied to the 2026 guidance of $200 million of revenue? And is there any color on how much contribution is expected from the GPU as a service and any of those data center projects you're working on?
For 2026.
What I'd expect is for gross margin in the fourth quarter to improve sequentially, though it won't be a dramatic change. As Jay said, we'll recognize some revenue from opus service in the fourth quarter, but that won't be enough to dramatically shift gross margins for the full year. I think the real impact from the improved gross margin from GPs a service will come in the first quarter, when the revenue numbers will become quite large and the margin impact will also be tremendous. So you can infer from that, that gross margin will improve sequentially in the fourth quarter, and that overall, the — I'm trying to answer my words. I don't want to give way too much; so you can imply that the EBITDA the EBITDA performance will improve, but PAUSE — but the real improvement will come in 2027.
The first question is a bit backward looking — when I look at third quarter revenue, how much of it came from hardware purchases that hold margins down, the kind that are onetime as you build out infrastructure, versus legacy services contracts, smart contracts, the portfolio you had before moving into the data center business.
Brian, would you say most of the hardware — meaning the revenue recognized in the third quarter — came from either systems integration or hardware carrying a depressed gross margin? And the other point I'd stress is that some of the Egypt related revenue likewise showed a depressed gross margin for the same reason as in the second quarter, namely that we bought it in dollars but deliver it and recognize the revenue in Egyptian pounds. Given the geopolitical situation, the Egyptian pound has been quite weak against the dollar. So that appears in the financial statements as a weaker gross margin, because the COGS is recognized as we — the cost of the goods purchased is actually recognized when we deliver it and recognize the revenue.
Great. So my final question — when I look at next year's revenue guidance, what exactly is baked in? How are you factoring in the portion of revenue that doesn't come from data centers — I mean things like the Egypt contract, or any other deals outside the data center business?
Taiwan is business as usual; the core mix of maintenance and recurring customers will — and the contract we announced in October should add about $25 million — we also assume some contribution from the net project. The precise figure — I'd rather not commit to one right now, since the contract's end date is July. After that point wraps up, we're currently in discussions about what follows, whether that's maintenance, support, or something along those lines. So pinning down a number at this stage isn't something I want to do. Our approach is to only give out figures once they're under contract, with both a date and an amount attached — prior to that, it's just not something we're comfortable doing. So then — the bulk of the revenue in our guidance comes from the Yota project, where, as Jay mentioned, we've already taken delivery, and we're configuring for Phase — and then Phase 2, we have a delivery schedule aligned to get up and running and billing in the first quarter; and to that as well. Sorry, Brian, just to add to that. The revenue targets you're talking about are completely man-dependent, or entirely now dependent on GPU as a service. We're expecting major contributions from both the technology operations expanding AI infrastructure business and so on and so forth. As Bruce mentioned, Egypt is still a major factor for us. Yet when you consider every Taiwanese project we've signed lately, those also turn into a huge platform for our cybersecurity, network intelligence, air infrastructure, surveillance, and similar areas. And what matters more — I want everyone to take this point away — we are not walking away from our technology heritage business in order to turn into a data center operator; data centers are one piece of the infrastructure story we've already put in place. I believe the market has to acknowledge that; and going forward, Brian, I think this year will certainly see all the technologies in our data center at Kora fully valued, with 200-plus megawatts coming from Gorilla. Gorilla is going to provide every piece of software, the whole security platform, all managed services, and every infrastructure service for Gorina. In practical terms, that means in next year's revenue, and as we head into 2028, you're going to see an enormous — possibly steep — increase when it comes to GPU as a service — but once 2028 arrives, things will start to flatten out. At that point, you're looking at a substantial share coming from multiyear contracts tied to your network surveillance and AI surveillance offerings, your infrastructure management solutions, your intelligent cybersecurity solutions, and so forth — leveling out and pushing GPU as a service to roughly 40%, 50%, 60% of the business, while the remainder takes over a far larger piece as well. That is exactly why this has us so enthusiastic, and not merely in our role as a data center operator. I hope that addresses your question.
Thanks a lot. I mean, some really intriguing questions came up today. I genuinely appreciate them. I'm certain there are plenty more urgent questions out there. At the close of these calls, I always notice folks asking why you didn't address this point or that one, and so forth. Time constrains us, but we'll do our best to ensure we can answer every last question. Let me express gratitude to all our shareholders, especially those who have stayed with us through the recent market volatility. I know our stock price is doing badly, but I promise you that while our share price has been dropping quite eagerly, fortunately our revenue seems to have lost its bearings and headed solidly the opposite way. Put another way, we're up 163%, which is the biggest quarter we've ever had. I'd love to take the credit. Unfortunately, it's completely the team behind us who delivered those results, and all of them are on the call listening right now. So let me share something a little more entertaining, or somewhat funnier. At dinner not long ago, my wife asked me what was on my mind. Let me tone that down a bit, though I do recall over dinner bringing up power distribution, GPU utilization, and data center cooling, and she came back with, Jay, what ever became of road maps? So I told her, look, I've locked in 560 megawatts of redundant power — if that doesn't count as commitment, then I honestly don't know what does. And you know what? I'm still sleeping in the spare room as things stand. I'm not sharing this just to fill you in. I invented that story because a CEO has to supply ambition. Bruce supplies the financial discipline. When I look at a $1 billion revenue company, I can spot 7 ways to make it more profitable. All right, in all seriousness, folks — to our shareholders, employees, customers, and partners, thank you for continuing to place your trust in us. I know how hard everyone is working, and letting anyone down is the last thing I want. We've shown we're capable of delivering growth. Our job now is to turn that ability to grow into profitable growth, cash generation, and shareholder value that lasts. Gorilla just had its biggest quarter to date, and we plan to work incredibly hard so that this is only the start. So thank you, ladies and gentlemen. I need to head back to my spare room now — it's 5:45 a.m. where I am. Take care. Bye-bye.
That brings today's conference to an end once more, ladies and gentlemen. Thanks to everyone for taking part. You can disconnect at this point.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Gorilla Technology Group Inc. transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.
Get an API key View API docs →For developers and AI pipelines
Programmatic access to Gorilla Technology Group Inc. earnings transcripts and 132,000+ others is available through the
EarningsAPI REST API and the hosted MCP server.
Quarterly plans from $162 - full transcripts, speaker segments, full-text search,
and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.