Ladies and gentlemen, good day and welcome to the E2E Networks Limited Q2 and H1 FY 2026 earnings conference call hosted by Go India Advisors. As a reminder, all participant lines will be in listen-only mode. There will be an opportunity for you to ask questions after the presentation concludes. Should you need assistance during the conference call, please signal an operator by pressing star and zero on your touch-tone phone. Please note that this conference is being recorded. I now hand the conference over to Ms. Raashi Khatri from Go India Advisors. Thank you, and over to you.
Thank you, and good afternoon, everyone. I welcome you to E2E Networks Limited H1 and Q2 results and earnings call. We have with us on call today Mr. Tarun Dua, Managing Director; Mr. Nitin Jain, Chief Financial Officer; and Mr. Ronit Gaba, Company Secretary. I must remind you that the discussion from today's call may include certain forward-looking statements that may be viewed in conjunction with the risks that the company may face. I now bring over to Mr. Tarun Dua to take us through the company's business and financial highlights. Subsequent to which we will open the floor for Q&A. Thank you, and over to you, sir.
Yeah. Thank you. Hi, everyone. Welcome to the E2E financial year results by E2E Networks. Let me very briefly talk about our company. So E2E Networks completes infrastructure tier including cloud computing across India. We have built a capacity of more than 900 cloud computing, and we have been in this business providing complete infrastructure since 2009. We have built a very strong engineering team over the last many years and we have built our from-the-ground-up self-service platform for running our entire cloud infrastructure, which is accessible on the web on myaccount.e2enetworks.com. TIR is our AI-enabled platform, which is about infrastructure. Our platform allows for, like a lot of features which are independent of any large cloud provider on any infrastructure. Let us talk about some of the updates, some of which I have already mentioned to you.
You have recently seen two reasonably large orders from IndiaAI Mission for INR 88 crores and INR 177 crores. These are primarily for customers who want to run their own LLM training for building large symmetric LLM models. This is primarily focused on India and also like MIT, The Chinese University of Hong Kong, United Women's Hockey League. These two large orders, we are in advanced discussions with both the IndiaAI Mission team as well as with us on when to go live with these. And we expect these two to go live very soon. Now, we had made a guidance for FY 2026 March for the monthly run rate, where we mentioned a number of anywhere INR 35-INR 40 crores per month. With these two orders, it looks like we should be able to meet our monthly run rate this quarter, hopefully much earlier than March.
And obviously, there is a great deal of confidence because of these AI orders that it could be sooner. We are also continuing to see a very robust demand for cloud computing. The demand continues at a high spread and at a reasonably higher price point for actual computing. We are also seeing a lot of demand globally. Essentially what we had was that as we have visibility into our capacity utilization going forward, we would be looking at procuring more racks soon. Given the market scenario of a lot of demand coming up for racks soon, we are already in very advanced stages of placing orders for nearly 2,300 Blackwell units, majorly B200. This is primarily funded through our internal accrual, previous fund raiser and debt from institutions. We spoke about how the Chennai location had a limited play.
Our Chennai location went live as well. All of our capacity play is basically online and available for use by our customers. I would also like to update and share with you about the ongoing acquisition of the AddressLab assets. Primarily the reason for acquiring the frameworks of SilverTap, was essentially to be able to deliver a lot of additional customer demand for a lot of customers we were currently servicing, especially customers who operate globally. Even more importantly for us was the people who were a part of SilverTap, now they are a part of E2E Networks itself. Essentially, we believe they are able to contribute overall to building up a very fine DNA for E2E Networks itself. That's one of the reasons for the SilverTap acquisition.
Hopefully in the medium and long term, we will see a lot of benefits from having access to great talent, which we have already hired within E2E Networks and acquired through SilverTap. They will continue to perform and provide more tech talent in India. Over a period of time, we will see the benefits of that. I think our focus continues to be, from a management perspective, our focus continues to be to right-size the cloud GPU versus SRE acquisition in a way that is prudent and as well as also aggressive, prudent and aggressive at the same time. I'm trying to hit both sides. Especially prudent from the point of view, that if customers have tendency on how they subscribe to SRE, we want to buy more, and buy as much as, actually, we don't lose value or lack of agility.
That balance we have always tried to maintain, and in future also our strategy remains to balance both too. That being said, overall the global market, if you look at it, we are still very small. This essentially means that there is plenty of addressable market for us to pursue. Whatever we pursue, we are very confident that we will be able to utilize that for revenue. That continues to remain our strategy. On the platform side, on technology side, we are committed to continue to invest in our sovereign technology. When I use the word sovereign, I essentially mean that we are building a technology which is primarily reliant on Open-Source with minimal reliance on international software, which will be open to say, encompasses initiative.
We continue to invest in our homegrown technology for our customers, for our cloud, and we will continue to do that. We also see long-term benefits of continuing with that strategy of having full stack sovereignty. Essentially, sovereignty should be designed for our customers. The example is that we are an Indian cloud computing provider where the infrastructure is owned by us. The data centers are physically present in India, and the software being operated on our cloud is primarily built by us or is open source, and we need to re-implement that open source also in India. You have three levels of sovereignty over here, and we believe in India's strategy of data sovereignty over the AI model, which is what we are seeing the investment from IndiaAI Mission on, where they are trying to ensure sovereignty of the Indian AI model.
That ensures four levels of sovereignty. As a country, we will continue to invest into the entire sovereignty platform, just like the past year I have invested into the digital public infrastructure. Things like the partnership between public and private enterprises could lead to optimum results and cost efficiency. We believe that we are in a state of that situation in India. With that note, I would like to hand over the session to Nitin, our CFO, who will present. Then we will open up the session for Q&A session.
Thank you, Tarun. Good afternoon, all. I would just run you through the financial highlights. This quarter ended September 30, 2025, we reported a revenue of INR 50.8 crore, which is 21% over the last quarter, which is April to June 2025. EBITDA margin has significantly improved to 41% from 29% in the previous quarter. Other income, mainly comprising of interest on deposits, has declined on account of utilization of deposits for the payment of CapEx. As a consequential to the depreciation, we reported a net loss of INR 13.5 crore.
Now we can open the floor for the question and answer session.
Thank you very much. We will now begin the question and answer session. Anyone who wishes to ask a question may press star and one on their telephone. If you wish to remove yourself from the question queue, you may press star and two. Our recipients are requested to take a queue, while asking a question. Ladies and gentlemen, we will wait for a moment while the question queue assembles. The first question is Bhavya Gandhi from Dalal & Broacha Stock Broking . Please go ahead.
Yeah. Hi. My first question is regarding the potential of fully utilized, the group that we have received recently. What are the plans for this year and next two years you can provide in terms of amount as well as in terms of GPU?
Let's talk in terms of GPU. We believe it's not like globally what we have seen, and we hope that that will be replicated in India, is that the Blackwell are going to be bigger than Hopper in terms of global utilization. The price points for Blackwell and there is no utilization for Blackwell. Utilization of 80%-90%. We do not expect that soon. Overall, we believe that whatever number of Hoppers we have acquired, like H100 and H200 will be anywhere between 2x to 3x that number. When we run through the entire sector, we will know. Like I mentioned that, we are both prudent as well as aggressive. We will continue to operate both prudently as well as aggressively to acquire more selectively, obviously, particularly. As we are immediately planning to acquire about 2,048 Blackwell.
As soon as they are acquired or while we are in the process of acquiring them, expecting that there is plenty more capacity that can be utilized based on the interest that we see coming from overall demand side. We might decide to kind of like double the sector on the way to 2,048 before March. We believe that with that order book, we may be acquiring each Blackwell until the next generation is ready to, when it becomes available for receipt in the Indian market. We believe that there is a long run for us to be able to answer that question on an immediate baseline , I think anywhere between 2,048 to 4,096 should be there.
You are mentioning before March. This March, you are saying March 2026?
Yeah, I am not talking about requirement, but in terms of pipeline and what is already underway under offer. Essentially, another 2,048, even if it goes up to cluster. Whether the requirement will happen in say March or later, that is pretty hard to say unless we take full orders with vendors.
Can we assume 4,096 if average prices for one Blackwell GPU is closer to INR 50 lakh, can we assume that CapEx for Blackwell should be closer to INR 56 lakh?
It is hard to kind of like that. Like I said, look, the pricing is dependent on a lot of things. Let us not make things very hard compared to detail. But we are building sufficient excess capacity, which is enough to lose any opportunities for accidental shifts coming our way.
The second question is regarding the purchase of service and consumables . I believe that is largely with the Chennai facility and largely done with. Can we assume that purchase orders would be recurrent going forward in future quarters as well?
Purchase orders will be there both in that capacity and that side in terms of when our utilization goes high, the cost obviously goes higher. As we continue to build, let's say, a lot of Blackwell GPUs for training for clusters and many other things. That being said, broadly it's a range that does not change drastically. Unless you add a few more clusters of GPUs. At that time, we don't expect it to change drastically. That number obviously would change, but not to any significant extent.
And sir, just regarding the data center capacity, how much megawatts do you currently have, and how much do you require with Blackwell capacity coming on stream? If you can throw some light on this.
We have some of the older products such that we specify close to 1 MW or so. The newer capacity that have been built in the recent past, like past two years or so. If you look at the Chennai one with newer and infrastructure, about 9 MW plus 1 MW will be for capacity. So that's an overall of about 10 MW capacit. Broadly, if you look at 10 MW, we are looking at somewhere between 8,000-10,000 cloud GPUs can be made available over there. That being said, our Chennai facility, we have the ability to get more capacity over there at a reasonably short time.
Got it. Just a follow-up on the depreciation part. Is the depreciation charge for the newer new gross block already capitalized? I mean, if you look at P&L right now, we are up INR 42 crore run rate. That would be largely depreciation number for following fiscal as well.
I will let Nitin answer this question.
Yeah. I think the number would be broadly similar, but only one point that Chennai cluster will be in active of August 1. So one month depreciation would be an additional number.
Got it. We can assume INR 50 crore depreciation run rate for the entire year. Sorry, for the next coming quarters?
Yeah. Should be on the same lines.
Got it, sir. Thank you so much. I will get back in the queue.
Thank you. With each transfer then, in order to ensure that the management is ready to address questions from all participants, we request that please limit your questions to a single question. If you have a follow-up question, you may rejoin the queue. The next question is from the line of Anand Bhaskaran from KSEMA Wealth Pvt Ltd. Please go ahead.
Good evening, sir. Can you hear me?
Yes.
Yeah. I just want to get this thing, one thing. The data centers which you are offering, this is of third party, is it not all of them?
Right. Yes. We do not own any data centers, so these are all third party data centers.
Okay. Recently, CtrlS announced that they will be installing a 1 GW data center in Andhra. They said that the main customers will be for accelerators, for cloud GPU accelerators. Will E2E Networks be a client or going to be participating in that data center?
I think this is a question which is a bit premature. I think we like to see a data center where we are planning to execute. Once we build those data centers and we have the capacity available within, we would be happy to participate along with all the other players. We essentially go by merits of any player before making a decision. If CtrlS is in the market, we are happy to work with them.
Are talks underway or something or like?
No, not currently.
Okay. My second question is, you said the MRR for March 2026 would be the same as you mentioned, which was INR 54 crores?
We have integrated our guidance based on the conditional allocation we have received from IndiaAI Mission, where if we get the two orders together, then that takes us to the four-plus MRR coming, and then we have another company asked to build up more runway. We are reasonably confident of the same guidance as we will earlier.
Okay. Would you say that in next quarter, would you reach that MRR?
You see, the guidance we have given is for March. We would like to stick to that. Actually, it should be earlier than that. It is because these two contracts, these are imminent. As and when they go online, obviously we will book the revenue.
Okay. Just also last question, because I just want to get some information right. Currently, you have 3,900 + GPUs available, right? I think that is your sort of capacity. Previously, you mentioned 8,000-10,000 cloud GPUs. That is apart from the AI GPUs that are available? Or there is a potential that you can use those-
I think earlier you asked a question about what the data center capacity. I was specifically talking about the data center capacity can sustain that many GPUs. That doesn't mean that we have to have that many GPUs.
Okay. Thank you so much. I'll get back with you.
Thank you. The next question is from the line of Keshav Sureka from Niveshaay. Please go ahead.
Thank you so much for the opportunity. Hi, sir. Regarding the recent order that you won through IndiaAI Mission. We can see that realization for this order appears to be on the lower side. It potentially extends the interest period. Could you provide some color on the margin profile for this project and what margin should we expect going forward?
See, I think, overall, once we actually go into implementation and then we actually look at all the requirements that might come through, whether through AI or through the people who have received these allocations. Only then we will be able to comment on margin profile with impact. Currently, we are not able to say whether it would be too low or too high or whatever. I think we'll have to do a look back to whether our- How is the overall margin comparison for Q3?
Are we in line with the margin guidance that we earlier guided?
The EBITDA guidance I have said, also that I have the E2E rate somewhere close to 26%± something. We can try that in the overall company.
Got it. Is there any visibility that we have got in the adoption on the TIR front?
No, not currently. There is no immediate visibility of revenue on that.
Got it. I will come back. Thank you so much.
Thank you. The next question is from the line of Shubham Agrawal from BS Securities. Please go ahead.
Hello. Am I audible?
Yeah. Hi, Shubham. Yes, you are.
Yeah. Hello. Hi. I just have one question. What is the useful life that you take for your GPUs for the depreciation configuration?
For the depreciation, I believe it is five years, which is also the life of the GPU. Mostly useful life means we have for GPUs or CPUs, we take five to easily about seven, eight years.
Seven years. Okay. That was the question I had. Rest of it for the numbers side.
Thank you. The next question is from the line of Varun Gandhi from Fident Asset Management. Please go ahead.
Thanks for the opportunity. Tarun, my question to you is, on the AI development side in India, as far as my understanding is, if someone is using an LLM API, the computing happens on the LLM's end. The computing and heavy data is on foreign waters, foreign land. What I wanted to understand is, in India, do you see that significant of a compute requirement? Because if AI development is happening through APIs, and I am not sure how much is happening through APIs, if it is happening, then it is not much of a big market for us.
There are so many functions over there in terms of what people are trying to do with AI. If you look at key use cases where people need to have a control over the outcome, where people need to have security, where people need to have the sovereignty of how they are processing the data. Let us take the case of a company with a significant number of users. Eventually, with any significant number of users who are doing financial transactions here, then I believe that a lot of the data actually is in India itself. The thing is that these are not the people who could use a one-size-fits-all generic inference API to do their business or to run their business.
They would obviously look for an edge, which can only be achieved by having a very different architecture or inference APIs, where you are able to make changes to the underlying platform, with the help of retraining, fine-tuning, and being able to teach the AI models. We see a lot more development in the open-source AI, which enables all of these things that we are talking about. While we have seen that proprietary AI definitely comes first and says that, Yeah, we are doing something that nobody else can do. What we have seen is that the rate of change in open-source is very rapid today. Mostly we see that there is a one-quarter, two-quarter, or three-quarter gap between open-source with respect to the basic things that proprietary AI models are able to do.
The way to get derived from open-source models is by running your own on-premise around those models and being able to take definite and direct benefits of the model. Cloud tools are a major usage alone for AI. AI is present everywhere. Wherever developers are using AI is not the only question which is happening. There is a need and want for sovereignty over AI, which is very important. That, I believe
Sorry to interrupt. The implication is that we are largely dependent on open-source architecture and developments around that architecture.
Okay. Sure. That is one way of putting it.
Understood. Just a follow-up to this. OpenAI has opened an office in Delhi and still very premature.
We will not be commenting on OpenAI over here in this call. Yes, OpenAI is a large company, but we are not talking about that over here.
Is there a chance this could be fulfilling their compute resource? Or you won't be commenting on that?
Too early to say at this point in time.
Got you. The last question would be, although the economic life of a GPU, you said was seven to eight years, do you think going forward, since technology is advancing so rapidly, do you think seven to eight years of economic life would be an appropriate assumption?
Yeah, it has continued to be an appropriate assumption so far. What we have seen is that where each system of what software has been trained on top of what AI has helped track. All of these things put together continue to create that regular type of demand over a particular model of GPU. That will change very rapidly without investing rapidly into our development in data science and AI/ ML effort. Net, what we have seen in the past is that it continues to be the future where any GPU model that is existing, that will have a feature or certain use cases, which kind of came and go during a peak prime time of a particular issue.
For instance, currently we see all three big models, like V100, like H100, A100 series, A30 and related series, and there is a demand for all three of them. Essentially we are looking at almost a clear life cycle of three of them. That will actually enter into a life cycle. That being said, we believe that our assumption that GPUs will continue to be useful is evidenced with real time.
Understood. Thank you very much.
Next question is from the line of Amina Shen, an individual investor. Please go ahead.
Yeah, hi. Thank you for my opportunity. I just want to understand that other current assets in the balance sheet in H1 have gone up significantly. What I understand is that it's a GST input receivable item. I assume that it was for our GPUs that we had in CWIP. I heard it was fully registered. If you can address that concern. Why has it almost doubled in amount?
Nitin, can you address that question?
Yeah. The primary amount includes the GST balances and the fixed deposits, other current assets, advance to vendors that we have given across procurement of the new GPUs. That is the other reason along with the previous balances of GST receivable.
Sir, the current expense that has happened from the previous amount which has doubled, it is not GST input has not gone up. Some other items have gone up there.
Yeah. GST input was already captured in the March financials because the capitalization was at the CWIP and GST input was already captured. During the current quarter, it has gone up primarily because of advances given for the procurement of new GPUs.
Okay, sir. Got it. The next question is regarding our debts that we have taken about INR 100 crores. Can you guide what our tenure debt towards the end, what our debt utilizations would be for next year?
The facility we have taken is INR 450 crore, which would be utilized as per the terms of agreement with the vendor during the payment cycle. By end of March, that completely should be utilized across. By end of March or early April.
Got it, sir. Got it. Just the last question. Regarding, sir, our initial two contracts that we won, it has been more on the basis of GPUs being given out on an hourly basis. Going forward, if we have any enterprise wins, will this continue to be our model or will we move towards a more traditional cloud-based model, which is our pay-as-you-go model, which will actually avail our platform benefits? If you can throw some light on how our model is going to shape up going forward.
Sorry, I was on mute. We continue to work on all aspects of how people want to utilize the GPU. There would always be customers who want to use GPUs for a short period of time where they would want to hourly, weekly, monthly, bi-monthly. There would always be customers who would look for longer term engagements in terms of how they want to utilize their GPU.
Just to follow up on that.
Sorry to interrupt you, Mr. Amina Shen . We have a request, if you please rejoin the queue.
Yeah. Thank you.
Thank you. Next question is from the line of Kartikeyan Vaidyanathan from Raj Investments. Please go ahead.
Yeah. Thanks for the opportunity. I think I have two questions. First is, why exactly is the mode that E2E enjoys compared to other players and how you're using the L&T buying some of the listing in your portfolio? How are you sort of leveraging L&T to your full potential? Can you throw some light on this? I just want to understand when you compare to the players, what is the unique position that E2E offers?
We are a cloud player, not a data center player. We have a good track record. We have been serving you all the way since around 2019- 2020 kind of timeframe. We have been working on our software platform for last so many years. We have the ability to provide the solution of the kind customers need. That's what we would like to believe is our moat, that our customer orientation toward providing the support, providing the solution, and providing the software platform on which to run either training or inference or model inference deployment. The ability to integrate with our digital platform that we have built up over last many years. All of these are a part of our moat.
Of course, becoming a cloud involves doing a significant number of activities that all need to bring us on board together. Essentially, we have a team structure doing that for a long time. The team itself is also a moat.
How are you leveraging the L&T infrastructure? Is it helping you with physical infrastructure in any way that gives you some advantage?
I mentioned that in the past that L&T is one of the best partners for an India-centric company. What I believe is L&T is a very large group which has the ability to open a lot of doors, a lot of conversations. And we continue to work with them to explore the various inside and outside use cases where we can work together.
Great. Thank you. So I think my second question is, I know you mentioned about Blackwell and all that, right? But there are also other vendors like AMD also coming to the table. And I do not know whether you are looking at supplying other GPUs or primarily the NVIDIA GPUs. Again, I think you, part of the question, after the customer asked his first question, the other investor asked this question. You started looking at NVIDIA GPUs earlier in your calls. In the previous call, you mentioned about training. So I want to understand who are these customers? What kind of inference serving are they looking for on Blackwell, NVIDIA GPUs as well as other GPUs?
I think I understand the question. So you are talking about-
Are you expanding your portfolio to other vendors like AMD GPU in addition to the NVIDIA GPUs or is it only focused on the cloud?
Currently, we are very focused on where the customer demand is coming from. Primarily we see actually customer demand today with NVIDIA GPU, specifically H100. That is one. What was the other question you had?
Specifically, in your investor presentation, on one of the slides you talked about vLLM and frameworks , all of that. These are being-
All of that is a part of our TIR platform that we have been doing for a while now. There are customers who are running many of the use cases that we have built.
Okay. These customers are what? Are they AI startups, enterprises or are they government educational institutes?
All sorts.
Okay. But I thought the training side of the story is probably you had government and education institutes. I was expecting more inside story on the AI startups and enterprises. But having all of it, there's no difference here, right?
Yeah. There is no difference, right.
Okay. Thank you.
Perfect. Yeah.
Thank you. The next question is from Zainab Akbar, an individual investor. Please go ahead.
Yeah. Thanks for the opportunity. Sir, my first question is, I think we have practices of association we have with broker authorized by around INR 1,000 crore.
Just wanted to understand, are we looking for M&A activities in the near term? If yes, how would you be doing it, like through a QIP?
We will obviously keep everyone informed about our plan.
Sir,
We will make an announcement as and when we.
Since you are looking to increase capital, I believe most likely it will be an equity fundraiser. If that happens, what kind of a dilution the promoter are we looking at? What will be the impact of price in respect to the equity raising? This is basically what I wanted to understand.
Like I said, that we will obviously keep everyone informed about our exact plan. That's what I would like to say for now.
Nothing for this year? 2026 is not near term?
Once again, the fact is I would like to keep everyone informed about whatever we are doing. If we do anything, we will obviously make an announcement and you will learn from that.
Understood, sir. Thank you so much. Yeah.
Thank you.
Thank you. The next question is from the line of [Akash Rawat] from Ridhanta Vision Private Limited. Please go ahead.
Hi, am I audible?
Yes, you are. Please go ahead.
Okay, so my question is, could you please explain current utilization level across all your business like H200, H100 and other business you are doing with Blackwell?
See, I think like in the previous quarter which has gone by, the overall utilization for all revenue was close to 35%-40% or so. There is obviously like an element of a lot of non-revenue utilization as well. So these were the numbers for like the previous quarter. We are obviously targeting somewhere between 80%-90% utilization for the current infrastructure that we have. It's just we have the confidence that we are able to achieve those numbers based on the IndiaAI Mission that have been allocated to us. So we are very hopeful that by March of the current infrastructure that is already there, that utilization will get to about 80%-90%.
Okay. And, sir, for best utilization range, utilization target for each enterprise which is given on the newly deployed-
I'm sorry, I was not able to hear your question clearly.
Yeah. I was saying, what utilization range you are expecting for the enterprise tags given on newly deployed capacity which will deploy.
Sorry about that. The line is not clear.
[Akash Rawat], I am sorry to interrupt. Your voice is something muffled, sir.
Okay. Is my voice clear?
Yes, please go ahead.
Yes, that's clear.
Yeah. I was asking what utilization range is required for different tags given on the newly deployed capacity.
Let me hand over to Nitin, actually.
Could you please repeat what exact range we require?
Which range of tag given on the newly deployed capacity is required. Like what utilization range is required, what kind of tags is given on the newly deployed capacity.
I think it could be difficult to say for a new capacity because the cost structure that we have is broadly fixed because we would not be able to allocate the cost of employees, et cetera, and the admin expenses to that. From a GP margin perspective, what we foresee is across our GP margin should be ranging from 80%-85% across on that newly deployed capacity.
Understood. Okay.
[Akash Rawat], may we request you to please rejoin the queue. We have participants waiting for their turn.
Okay.
Thank you. The next question is from the line of [ Vaidyanathan ] from RK Investment. Please go ahead.
Yeah. Good evening. Can you hear me?
Yes, sir. We can. Please go ahead.
Yeah. Two questions. The first question is, what is the return on equity that you are looking at in the medium term, say about two to three years, sir? I know that currently you are not making money at the bottom line. Just want to know what is the medium term targets?
See, it will always be, like as we keep growing, it will always be a work in progress. But ultimately, what we are targeting from an EBITDA perspective is to be somewhere in the range of 70% + or something. Over the medium and long term, as the volume grows, I think it will somewhere stabilize around [2%].
70%, is that what you said?
Yes, sir.
I see. Got it. The second question, if you mentioned the MRR, the monthly run rate of INR 40 crore you want to achieve by March 2026.
We are trying to.
What would be the incremental cost? The cost line will be more as what we are facing.
I do not think there could be any major cost difference. Major cost difference in the sense that we have already accounted for a lot of costs that we have already incurred for the current infrastructure. I do not think it should drastically change. Having said, that was the question, or is there another question about it? Infrastructure we already have things for the current phase.
Yeah. Okay. Just to extend the question, I think your business kind of requires a lot of capacity to be deployed upfront. Do you think you will be able to win this race in the longer run, given that, of course, you have the backing of L&T , but I'm just saying that what if you need to deploy another, say, INR 10,000 crore of cash, if the AI really gets very big in India. Do you think you'll have the financial headroom to invest such a large amount of money?
I think it would be very speculative to answer that specifically. Obviously, we are trying to build up our capacity as best as we can.
But how do you address the risk of obsolescence here? That is one big risk that you're running. What is your view on that?
We have been in the compute business for almost the last 15 years. We have seen multiple technologies. Essentially, there are so many moving parts to take care to not have to worry about obsolescence as long as you are expanding in an incremental manner.
Sorry to interrupt, Mr. [Vaidyanathan]. We will request you to please repeat the question. Thank you. Ladies and gentlemen, we will request you to please limit your questions to one participant. The next question is from the line of Neil Munot from PICO Capital . Please go ahead.
Hello. Hi, Tarun. Am I audible?
Yes, you are. Go ahead.
Sir, I wanted to understand our last June quarter exit MRR was INR 14.5 crore hence and we did INR 43.5 crores this quarter. What is our exit MRR that published for this quarter is INR 16 crores. Have you seen some dip in three months and then we have come back to INR 16 crores?
See, broadly, as soon as we receive the long-term order from the IndiaAI Mission, we have to clear the capacity from longer-term users or taking some business from the longer-term users and work with only the capacity which will be required to implement. In that sense, we are a bit short for our capacity. We are trying to utilize it, but we are trying to utilize it for the long term because we obviously want to give priority to see what more things are coming from the IndiaAI Mission. We also look as to how to utilize it, not keeping it like any idle. The balance between the two, I think that is something we are trying to establish.
Thank you. The next question is from the line of Keshav from Niveshaay. Please go ahead.
Yeah, thanks for the opportunity again. My question is, when do we expect the release to commence for the large language model from the IndiaAI Mission?
Can you say that again, please?
When do we expect the release to commence for the two large language model one from the AI engine? Any timeline that you can share?
I think it should be sooner than later. We can't say immediately that. But we are very hopeful that it should be very soon.
Thank you. The next question is from the line of Srinivasu Kedarasetti from Tamilnadu Investors Association. Please go ahead.
Hi, sir. Thanks for the opportunity. Have you done any benchmarks comparing E2E versus AWS or Azure or GCP on training time and cost for the same model and same dataset? What was the difference in terms of training cost and cost on E2E versus one of these hyperscalers? Thank you, sir.
There are certain standard perf detail that our team is doing internally, where they seem to find the compute performance, and we believe that, obviously, we are not trying to believe that we are the best. I quickly believe is that the MLPerf that's been out here amongst the neocloud operators. It's very, very close to everyone else, including the independent software itself.
Thank you. The next question is from the line of Anand Bhaskaran from KSEMA Wealth Private Limited. Please go ahead.
Thanks for the opportunity, Tarun. I just want to ask one question. In the beginning of this month, Trump did try to attempt to restrict NVIDIA's software, Blackwell chips to leave U.S. that easily. He's barring it from China, but he's also made, tried to attempt from other countries as well. Would you have any issues or any threats to that?
I think that question is like solved by itself, right? Let me like we do our level best to follow all our laws India and laws of countries of origin of where we are buying our equipment. We ensure fully that none of our services are utilized by anyone who is not allowed to itself also.
Thank you. The next question is from the line of Bhavya Gandhi from Dalal & Broacha. Go ahead.
Yeah. I just wanted to know if you have seen any convergence on the policy side with respect to enterprise. Because longer run that has a huge business, right? How is the AI demand environment from IndiaAI Mission, if you can spend some time over that as well.
See, the enterprises will continue to adopt AI. We will continue to work with enterprises like through partnerships and convergence like we keep happening. That being said, the overall demand for the IndiaAI Mission, we are very positive about that. We believe that the goal of IndiaAI Mission is to make India self-sufficient and the cloud service provider, I think we are a very big part of that mission, and we will continue to contribute to this mission.
Thank you. Ladies and gentlemen, this concludes the question and answer session . That was the last question for today. I now hand the conference over to management for closing comments.
Yeah. So thank you everyone for listening to us. I would also like to thank our entire team, all our customers, all our shareholders, and all our ecosystem partners, including the advisors and everyone. Moving further along, we hope that we continue to have these conversations.
Thank you. That is a wrap . Because of management, with that, we conclude today's conference call. On behalf of Go India Advisors, that concludes this conference. Thank you for joining us, and you may now disconnect the line.