Thank you. We will now begin the question-and-answer session. The first question comes from the line of Neel Munot with Eco Capital.

Sir, this current quarter revenue and MRR jump, can you give us a mix between volume and utilization versus pricing? How much is driven by utilization? How much is driven by pricing?

So we are talking about revenue. , of course, there is expansion both in terms of utilization and addition of capacity, which have both majorly driven by revenue as opposed to just pricing alone where we have seen a very moderate impact in this previous -- the quarter that went by.

Understood. And sir, the new B200 that we've deployed mid -May, have they been committed fully in terms of capacity, as in, in terms of utilization?

Yes, it's quite a maximal utilization of capacity that we have.

And sir, on Indian AI nation, did they commit to Indian AI nation or on our own platform?

So currently, we are in a process of working with the customers around these GPUs, and kind of like, in a couple of weeks, we can figure out how this moves forward in the future.

Next question comes from the line of Bharat Gulati with Dalal & Broacha Stock Broking.

Just regarding utilization levels for our H100, H200 and B200, so what would have been the exact mix in the exit MRR for this quarter? And what would -- how much more runway do we have to grow? I understand we've announced a price hike in the month of July . So, what kind of impact is that going to have on us in Q2 in terms of MRR growth?

Got it. And sir, you just spoke about SovCloud. So just trying to understand that these Sovereign workflows that we are hosting, are they yielding a better EBIT margin? Are they sort of asset light in nature and the private cloud business that we had itera ted a couple of quarters back? Has that started to kick in? If you can show -- throw some light on that?

You see the operating leverage is obviously kicking in as we have increased the amount of capacity. Now you asked a bunch of questions. So SovCloud essentially is a subsidiary, where the idea is to be able to hold and contract large-scale CPU clusters. Then you asked like a couple of more things like if you could repeat.

Next question comes from the line of Bhavya Gandhi with Bajaj Alternate Investment Management Limited.

Congratulations, sir, on very good set of numbers. Commendable. I just wanted to understand next the B200 deliveries, when can we expect any timeline, if you can provide?

So, we expect them to be here in the next couple of months. We'll obviously be keeping everyone involved once they are there.

Okay. And sir, would you like to guide for exit MRR full year basis this year?

See, we don't provide guidance on MRR. We always, advise everyone to look at what we have done in the past rather than, predict the future. So again, we would like to reiterate that let's look at the past one by quarter exit MRR, whether they predict something for the future.

Next question comes from the line of Nishant Joshi with Equisense Advisors Private Limited.

I have a query regarding B200. Can you give a bifurcation how much of the revenue is coming from inference and how much comes from training? And can you give an outlook for the next 2 to 3 quarters, how this ratio would change?

See, we are pretty small in terms of size and volume. So major volumes, of course, continue to be on training. Now that being said, we have increasingly seen that while people run their training clusters during their working hours, they also utilize the same clusters for running inference during the night. So pretty much hard to pin down the fungibility of compute to say that, okay, when the workload is training and when the workload is inference, we've seen both the use cases being run in parallel where you can't put a specific number to, how many GPUs are on training and how many GPUs are on inference. Pretty much both the workloads along with agentic workloads, continue to be run on the same infrastructure.

It's very hard to measure like what part is there on training and inference. Now eventually, at a global level of compute, inference will have a preponderance over training. But on a very small subset of GPUs, we are literally talking about 5,000 or so GPUs, It’s very hard to, say, what percentage is on training today and what percentage is on inference.

Next question comes from the line of Keshav with Niveshaay.

Congratulations on the great quarter. So if you could help us understand the recent subsidiaries, the Delaware entity and the GPU infrastructure entity, and I think you already spoke about SovCloud. So whether this, any of these vendors financing or maybe asset vehicle or maybe if you can just throw some light on that?

Sorry, I didn't really understand your question. Can you repeat? Your voice was not super clear.

Yes. Actually, we recently set up entities in Delaware and the GPU infrastructure entity and the SovCloud. I think you have already spoken about that. So , if you can give us some color on the subsidiaries, please?

Sure. So SovCloud is essentially an infrastructure subsidiary, very focused on build -out of the infrastructure. And the U.S. entity, the Delaware entity, obviously is focused on selling in the international market and being able to establish presence outside India, where we can work both towards market excellence as well as do alliance management, as well as service customers who are outside India.

Sure. And I think on the presentation you mentioned under the SovCloud section that large -scale GPU deployment and enabling funding arrangement to accelerate this AI infrastructure. If you can provide some strategic insights into the plan?

So very early days. as we do more, we'll obviously announce that.

Next question comes from the line of Varun Gandhi with Finavenue Growth Fund.

Again, great set. Very positively surprised. Number one would be on the margins. Going forward, should we expect these sorts of gross margins to be normal, because it's the best margin we've seen in the recent history?

Yes. So it's a combination of increased utilization, and a very robust demand scenario in the market. And also, majorly driven by operating leverage as we grow larger. Now as we grow in size, we obviously intend to have a judicious mix for both contracted revenue as well as the on -demand and spot revenue. And also explore various AI services in terms of inference and agentic based on open-source model. So, I do believe that this is quite sustainable over the medium term and potentially the long term.

See, we're quite small for that right now. So let us grow. Try to focus on growing bigger because these become tracking metrics for everyone once we say this is important and then people start tracking some things. And then there is potentially no relevant for these metrics today. So my advice would be that let us not do this today. I think at certain point of time, at a certain scale, probably it would start making sense. Today at 5,000 GPUs and another 1,000 coming in, still the size is too small to start talking about these metrics is what we believe.

Next question comes from the line of Vedant with Nirmal Bang Securities.

Firstly, congratulations, Tarun, on the great set of numbers. My question pertains to the outlook on AI industries. We are listening to some kind of AI bubble, and simultaneously China developing some AI models, which are cheap and very competitive to American model. So what are the comments on that part? And if something happen -- like, AI commoditization happens. So how do you see...

So what we believe in this is that, essentially it has still been in the world of AI. If you are plugged into the AI ecosystem, I think every week, there are some very significant pieces of news related to AI technology. Some of them tend to get fixed up b y media and blown out of proportion and people start to pay attention to that. We believe that some of those things which are equally important are completely lower. So it is still day zero of AI. We are still in the midst of largest build -outs in the history of IT itself where the whole world is building the AI super cycle. And I think we are at the beginning of that AI super cycle. And it's still day zero at AI. I don't think we should worry too much about, okay, this happened. And will the whole AI industry compress because of this one event or this new thing that is happening. So, I don't think we have come to that sort of stage anywhere where one particular new items that we are hearing is going to change the whole AI industry. I don't think that's the case today. I think pretty much we are seeing a very broad trend where most organizations in their AI adoption are fairly early, except for AI native, everyone else is fairly early in the ecosystem. And I think the runway for AI is very, very long. I think in the past , we have spoken about this is a decadal theme. I think we've just begun that decadal theme of AI adoption amongst the organization. I don't think we should start worrying about this for that news item. Like I said, there is a lot of news on AI which doesn't get captured by media, which is equally important. Some of the things get captured and then get blown out of the position. So I wouldn't worry too much about any of the news items related to AI, which either are predicting tomorrow the world of AI is changing or tomorrow there is a scenario in AI, I don't think either of them are true. I hope I have been able to answer your questions.

Yes. You expressed your views. And my second question is, there was some computing shortage, I think so, that's why your realizations have increased some bit. So how you see it going forward on the realization part, the demand supply gap between the computing GPUs?

We have to look at over a couple of years. So, the compute needs would come in phases. So, there would be phases where there is shortage of demand, there would be phases where there is a shortage of supply. Eventually, if you look at like a long enough cycle of 5, 6 years, or any particular set of assets, they will see both good times and maybe slight bad times, where hopefully there's increasing adoption, all of these would smooth out a lot more where customers would be able to predict demand in advance. And the supply side would react accordingly. And eventually that demand would match supply in a way that like nobody is losing money, whether it is the buyers for AI compute or whether it is the sellers for AI compute, I think we'll all find our equilibrium in the medium term. Majorly we are not really talking about the increased pricing resulting in the tremendous set of numbers. Majorly, our numbers reflect the increased capacity and increased operating leverage more than anything else.

Next question comes from the line of Shubham Tamrakar, CFA with Alturas.

Great set of numbers. Sir, I have a question with respect to the price increase. I'm just trying to understand how much capacity is still there at old prices? And - is there any pushback you're seeing when we increase our prices?

Okay. So essentially, we are in this for a long term. Now what we find is that, for every generation that we deploy, there will be ups and downs in each of those generations, and eventually each generation of GPUs will find their sweet spot in terms of price performance ratios. And obviously all contracts need to be honored wherever we enter into contracts with customers, whether they are advantages to us or whether they could be slightly disadvantageous to us. Essentially, we are here to play for the long term. So with regard to pushback on pricing, I think making sure that the customers are able to find value in our software and orchestration platform and are able to experience our solutioning and support. And we are able to access higher quality customers is more important to us than worry about pricing alone as a way to do business. So, our goal is to kind of -- have a very harmonious relationship with our customer. And I do believe that customers do understand that the market is offering a certain set of goods and services at a certain price, and that needs to be respected. And I don't feel that there is a very inactive, pushback from the customers this is something that has to be available at its price and continue to be available at this price. Of course, post-contracts, the price needs to be renegotiated and most customers are happy to renegotiate the price based on the current market scenario and current market conditions.

Fine. And sir, second question with respect to the time lines of full capacity and the capex plan that we are thinking of. So, if you can help us to understand this also?

We obviously plan to deploy the B200 as soon as a new lot becomes available to us. And we are expecting that to happen soon over the next couple of months. And apart from that, we obviously plan to build more capacity in terms of more Blackwell including non-flagship Blackwell as well. And also, we have plans to expand our capacity into Vera Rubin. and when we have immediate figures for, we are going to deploy x number of GPUs of this generation. We'll obviously come and talk about that. So, for now, I would like to maintain that we are both aggressive as well as judicious in terms of our capacity expansion. And being a AI factory, a neocloud owning its own platform, for us, definitely getting more GPUs is like the way forward for us, and we would continue to do that.

Next question comes from the line of Rohan Nagpal with Helios Capital.

Great set of numbers. So, I had a question on renewed CPU pricing that you had intimated customers about? So, two questions on that. One, is it driven purely by market conditions changing and demand increasing? Or is there an underlying cost pressure as driving the cost increase over there? And the second thing is you intimated a price increase going live and you rolled it back and you push it out a little further. So, if you could just talk about the dynamics over there.

Sure, so see, primarily on CPU, it's not a very major volume from a revenue perspective today for us. That being said, we would like to continue to add CPU, which is what necessitates increase in pricing for the CPU customers because at whatever price point we offer our services, we would like our services to be of the highest quality. And to maintain that quality, the requirement was that we should be able to charge more. Now we obviously propose that to our longer-term customers. There was an ask from the customers that, whatever we are already using, is there a way to preserve the price point for us. So, where we offered that kind of like contract some of the capacity for a slightly longer term, say, 1 year or 2 years, then we can hold the current price on the duration of the contract, which is an option that many of our customers choose. Then they said that, we'll pay you some advance, and we'll increase the contract length from normal 1 month to about a year or 6 months or 9 months or even more than a year. So many of the customers came back and they agreed to that. So that's the main dynamic around CPU pricing where the customers wanted to hold the current price, we were able to give them an option that, yes, why not contract out the capacity that you are already using for slightly longer terms. And then we can proceed with the revision of your current contracted capacity, which allows us to service you at a higher quality and which allows us to continue to expand the capacity, especially since the price of hardware from a CPU perspective has incr eased by a much larger percentage compared to the price of hardware on the GPU side. because e the cost of memory that has increased has hit the CPU side much higher in percentage terms than the GPUs.

So, to accommodate that, we had to, in fact, increase the price of both the GPUs and CPUs. And it's not just about the higher demand scenario that's associated with the price increase. So it was also that some of these inputs cost actually increased for us.

Next question comes from the line of Ashish Ajit Golechha with Bee Ventures LLP Fund.

Congratulations. Tarun, the question was that, in the PPT you mentioned that you have started preliminary standing for next -generation architectures like NVIDIA, B300 and Vera Rubin. Now you know that considering the current compressed life cycle of AI har dware, which you always mentioned in earlier calls, how do you protect the ROIC for the older Hopper and early Blackwell customer, because what happens you only, in your last 2, 3 con calls, you've mentioned that client preferences shift aggressively downstream. So this is my first question. And second question, I will ask when you finish this question?

Okay. Sure. Let me try and answer the question as best as I can we've always maintained that at least at the minimum a 6 -year life cycle for all the GPU generations. And we have seen even our older generations continue to run very strong with our customers. So every generation will run through its life cycle where eventually it finds a sweet spot of price and performance thresholds aided increasingly by increase in performance and efficiency -- of the open source model and fine-tuning that a lot of people share in open source that, how do we get an H100 to perform this much on this open source model in terms of tokens per second or latency or H100 or H200 or B200 or the new generations of B300 or Vera Rubin. So, all of that put together gives th e GPUs quite a long-life cycle. So, where we believe that long life cycle is conducive to monetize those GPUs through almost like 5 to 6 years. And, kind of like, we don't see foresee that there is an active price compression because of the newer generation GPU coming in because newer generation will have its own sweet spot for price performance ratio for some of the newer models. But like whoever is using the smaller model or older models, or people who have built their own proprietary SLMs, they don't necessarily intend to go for the latest generation GPU at a much higher price point because a newer model will obviously have like an increasingly higher cost, whatever percentage that number is.

Okay. Tarun, other question is, the accelerated computing now is deeply anchored to NVIDIA ecosystem. So , considering I wanted to talk regarding the ASIC threat, how does E2E threshold propose to integrate non-architecture -- non-NVIDIA architecture into the TIR platform, because if the enterprise demand diversifies, how do you address this issue? I hope I am clear on this?

Not super clear, but like what I heard I'll try to answer that. So, we have worked the longest time on GPUs with NVIDIA all the way from 2019 onwards. And obviously, they have certain best practices, and they have released a lot of open -source around how to kind of integrate and wrap around your GPUs in your platform. So, there is a lot of opinion over there from NVIDIA. And we continue to

incorporate a lot of opinion and advice and their open source software into our platform so that we continue to do. That being said, we are also open to integrating any accelerated computing platform into our platform wherever our software can be compatible with anything. We are super happy to extend our software to encompass any technology from any providers. We are quite vendor-neutral in that respect. I hope I was able to answer your question.

Next question comes from the line of Abhishek Shindadkar with InCred Capital.

Congrats on an amazing quarter. Just 2 questions, one is on the sustainability and the visibility of the current performance. So, basis all the operating metrics that you may be tracking internally, if you can just give a color in terms of the sustainability and the visibility of this performance going into this 2Q and 3Q. And second, you highlighted it, but just wanted to understand, are customers now trying to lock in capacity for longer given both the increase in the prices and what we are seeing in the entire hardware ecosystem, especially from a supply chain perspective? And does that kind of improve the annuity visibility for us, given historically we had short cycle capacity as well.

So, definitely, like, there are customers who do prefer, , longer-term contracts because they would not like their prices to increase, very nearly, because they were on-demand or monthly or very short- term contracts. And there, again, we have seen increased demand from the customer side where they would like to see longer term contracts of 1, 2 to 3 years coming their way. And so, as we judiciously move some percentage of our capacity, we haven't decided what percentage that should be. So that definitely builds in predictability for the demand that we have for our capacity and build some predictability into the revenue, so we'll ob viously continue to explore those options as well as execute those options. And we have done some of that and we have honed some of that in the past as well. And we would continue to increasingly do so as well. And, some of our capacity to have the flexibility of where we are able to charge higher where the demand goes up. So that also will continue to do. You had one more question, can you repeat the second part of your question?

Yes, the first question was about the visibility and the sustainability of th e current quarterly performance?

Right. So, this is, in our belief, quite sustainable because I think each of the generations which are already there are finding their sweet spot in terms of price performance. And we believe that over the period of terms, they are quite sustainable. And as those price points, quite a few customers are very willing to sign up longer -term contracts. So, we find there is a lot of predictability over there. Both predictability as well as sustainability of the revenue.

So just two parts to my question for Tarun, sir. Firstly, the breakup last quarter for domestic and foreign business was 60-40, and India AI mission was about 40% last quarter. So , what has that trended to this quarter? And in terms of your ambition for, let's say, 2 years out, how many GPUs should we be sitting at? And just I have a follow-up for Nitin sir, after this.

Sure. So, with regards to capacity building, we have already said that we intend to be more judicious as well as aggressive. Now from a revenue composition in the previous quarter, I think India AI revenue was about 20%, 21%. And international was closer to 37% or so. And rest is like all domestic revenue. So yes, those are the numbers.

Got that. And just one for Nitin, sir, is with regards -- just a couple of bookkeeping questions with regard to the interest cost spike we've seen during this quarter. Is that due to some debt we've taken on for the capex? And regarding the DC cost that hits the P&L, has it completely got baked into this quarter's number? Should we see that flat lining going forward?

So, in terms of the interest portion, you are right, the interest is because of the loan that we have taken across to fund the first lot of B200. And second, in terms of the cost for DC costs, it would fluctuate to the extent as the usage would be increasing. So, the electricity portion would keep on increasing.

And what percentage would that be of our entire DC cost? I understand our DC costs for the quarter was close to 23...

See, these are not like super granular numbers. Broadly a lot of the cost for existing capacity is already baked in for this quarter. And in the coming quarters, some of that cost would increase, but I don't think that would be a very massive percentage.

Next question comes from the line of Bhavya Gandhi with Bajaj Alternate Investment Management Limited.

Sir, two questions. One is regarding the absolute loan and capex for the loan amount outstanding as of the quarter and the total capex plan for the year. And the second question is regarding, you mentioned in the PPT, we are looking for funding requirements for Sovereign Cloud,. So, if you can throw some light on that as well.

Sorry, I didn't understand the question. What you're talking about, SovCloud, what is the exact question on that?

Yes. So, what are the funding requirements that we are looking for SovCloud? Any plans if you can share on SovCloud. I believe in the PPT you mentioned about..

Absolute loan amount, if you can mention short-term, long-term debt for the year and the capex plans for the year.

I'll let Nitin answer the question related to the loan.

So, the loan, which stands as of now is broadly INR 450 crores, which with the other lot coming in picture, it would be increasing across in the near term.

Yes. Possible to quantify that, sir, total loan amount, peak debt for the year?

We will not be quantifying that amount.

Next question comes from the line of Ashish Ajit Golechha with Bee Ventures LLP Fund.

My question was, the preferential issue total around 2 years was INR1.9 crores. So , I wanted to understand, are you planning further equity raise also because, point is, whatever the capex which you have basically shown earlier, wanted to understand where has that INR881.1 crores capex gone? If you could share the utilization. It would be really great. This is my first question, Sir.

So let Nitin answer the question about utilization of preferential funds, which were allocated towards the capex. Th at is one. And second, whatever GPUs we have already announced, their funding is fully arranged and backstopped by debt or internal accruals or previous equity raises. So that is already arranged for. And with regard to like any future for equity raises? Obviously, we would let everyone know if and when that happens.

Okay. Nitin, sir, another thing was, in the Q4 call, the 3,900 figure was described as CPU and GPU storage, all capacities put together. But the current PPT which you have shared yesterday, which is very good, present that 3,900 and 5,100 on a chart titled GPU trajectory. So this 5,100, is this what you are accelerating the blended capacity unit. I hope my question is clear.

Yes. So nearly 5,100 is the current capacity, which is live on our platform today. That's the number of GPUs which are live today. So that does not include another 1,024 B200 we are expecting soon.

Next question comes from the line of Chirag Satiya with Satiya Investment.

Sir, I just wanted to understand the financial dynamics of the strategic partnership with L&T?

Okay. So, what is the exact question over there?

I mean, in terms of revenue, how does it work out for you? How do they pay you or something?

Sure. So, let's see, L&T is an arm's length partner, where we are working with them to, kind of like, utilize their data center capacity. So, we are a buyer over there. They tend to utilize our cloud infrastructure where they are the buyer. And then jointly with regard to our platform, we do go to customers.

So that's how the arrangement is basically. And tomorrow, if L&T is building compute capacity, then we are also looking to partner with them and offtake some of that capacity. So all relationships at arm's length are present and we'll continue to kind of operationalize over a period of time.

Okay. And second question is, how do you see the growth trajectory in the next 2 to 3 years?

Okay we have mentioned that basically AI is entering into a huge super cycle. So we fully intend to be a part of that build -out in terms of both the infrastructure capacity is agai n aggressively and judiciously. And in terms of the build -out of our platform, where our platform is able to increasingly help our customers do more in terms of, whether it is training their in-house models or fine tuning open-source model, or running inference at a scale or running agentic harnesses that are able to take care of their enterprise AI-driven processes. So, we want to be able to help our customers qualitatively in all these aspects.

Next question comes from the line of Sucrit Patil with Eyesight Fintrade Private Limited.

I have two questions. The first question to Mr. Dua is - this quarter's growth is very good. But the cloud industry is entering a tougher phase, hyperscalers are scaling aggressively, customers are getting harder on pricing, and technological cycles are also getting short. How do you see the company positioning itself to...

See, we have been in this business for 16 -plus years. So, we have seen everything. So cycles will continue, there will be good times, there will be bad times. But we are very confident that for every cycle, we will continue to grow. So that's what we believe in.

My second question for Mr. Jain is, how -- I just want to understand how are you approaching the balance between funding expansion, keeping liquidity strong and ensuring shareholders return ? Looking ahead to the next quarter, any road map that can give a layout?

So, I think from a funding perspective, we keep a balance between the funding and the building up capacity aggressively and judicially. So there, we keep quite a balance that we should not be too aggressive or we should not miss the bus, which is the India AI bus, sorry, which is the AI bus.

Thank you. Ladies and gentlemen, due to time constraints, we have reached the end of question-and- answer session. I now hand the conference over to Mr. Tarun Dua for closing comments.

Thank you, everyone, for joining us this early in the morning. And I hope we have been able to answer most of your questions. And we do welcome this interaction. And any further questions, please do reach out to us at investors@e2enetworks.com. And we will reach out to you and answer all your queries.

And we hope to keep you updated about our ongoing expansion plan and , keep you informed about what we are building on the platform, how we are seeing things and we hope to keep communicating. And a very good morning once again to everyone, have a good day. And we would like to thank all our investors, all our partners, all our team members. It is with all your support and love that we have reached where we have reached. And we hope to continue to enjoy your support in our ongoing journey. Thank you, everyone, once again.

Thank you. On behalf of E2E Networks Limited, that concludes this conference. Thank you for joining us. You may now disconnect your lines.