E2E Networks Ltd (NSE: E2E) Q4 2025 Earnings Call dated Apr. 17, 2025
Corporate Participants:
Taran Dua — Managing Director
Megha Raheja — Whole-Time Director and Chief Financial Officer
Analysts:
Soumya Chhajed — Analyst
Pritesh Chheda — Analyst
Keshav — Analyst
Kshitij Saraf — Analyst
Aditya Agarwal — Analyst
Bhavya Gandhi — Analyst
Harshi Shah — Analyst
Amar Maurya — Analyst
VP Rajesh — Analyst
Prathamesh Dhiwar — Analyst
Neil Munot — Analyst
Gunjan Kabra — Analyst
Sampat Nayak — Analyst
Presentation:
Operator
Ladies and gentlemen, good day and welcome to E2E Networks Limited Q4FY25 earnings conference call hosted by Goindia Advisors. As a reminder, all participant lines will be in the listen only mode and 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 then zero on a touchstone phone. Please note that this conference is being recorded.
I now hand the conference over to Ms. Soumya Chajan from Goindia Advisors. Thank you. And over to you ma’am.
Soumya Chhajed — Analyst
Thank you Yusuf. And good evening everyone. We welcome you to E2E Networks Limited Q4FY25 results on call. We have with us on call today Mr. Tarun Dua, Managing Director, Ms. Megha Raheja, the whole time Director and CFO Mr. Ronald Gaba, the company secretary.
I must remind you that the discussion on today’s call may include certain forward looking statements and must be viewed in conjunction with the risk that the company may face.
I now request Mr. Tarun to take us to the company’s business and financial highlights subsequent to which we will open the floor for Q. Thank you. And over to you sir.
Taran Dua — Managing Director
Yeah, thanks. Good evening everyone. Thank you for joining us for E2E Networks Q4 financial year 25 earnings call. I hope all of you are doing very well. So let us begin with like this has been a momentous year for E2E Cloud and E2E. So we started doing like GPUs in like 2019 and we have slowly built up our capacity. So like last time I think we spoke like we were practically at like 700 or so 700,000 Hopper series GPUs today like including all our GPUs that we almost have like a GPU capacity of almost like nearly 3700 GPUs. This makes us like practically the very largest of GPU Cloud GPU installations in India.
Amongst the Indian players this year we also successfully did two fundraises aggregating to nearly 1,500 crores. And so just to recap our capability. Like we have been around since 2009 initially as a private company, in 2018 onwards as a public company. Today we have like nearly 100 plus engineers working on our AIML and cloud infrastructure platform From a data center standpoint like we have access to 10 megawatts of data center IoT power capacity spread across majorly like Delhi, NCR and Chennai region.
So what has set us apart is like we are the kind of like we were the first pay as you go cloud in India starting our operations in 2009. Like our claim to fame historically of course has been that like a lot of large customers were born on E2E cloud in the Internet Indian Internet world. So 2014 onwards like we have continuously built our own infrastructure software for running our public cloud. And like 202021 onwards like we started working on the cloud GPUs and like lot of our software capabilities around cloud GPU we started building around that time all of those capabilities kind of like Collins around our tier platform.
Now all of this software that we built is our practically known proprietary software which wraps around a lot of curated open source software to deliver like a complete platform experience. And we have also kind of like positioned this particular software as a sovereign cloud platform for running both CPU workloads, CPU compute workloads as well as sovereign AI workloads. So we’ll talk more about like what’s the sovereignty part in our software? See essentially one is of course like we continue to expand our public cloud software platform where like all of our data is hosted in India. We have the a lot of our customers like they have the ability to build their org boundary on top of our public cloud platform where they are assured of like security and that their data is not being used to train AI belonging to someone else in any way.
So like that produces the effect of like having AI sovereignty over your infrastructure. So essentially it should not be that usage of AI results in loss of competitiveness to the users of AI because they were running. Something that either could not make use of their own curated data sets, their own proprietary data sets because of concerns of sharing those data sets with someone else or alternatively it should not result in like you share the data set and you got like a better result but like that better result also propagated to other users of the same AI platform.
So in our cloud GPU approach basically we are focusing on customers who want to run single tenanted AI solutions and achieve complete sovereignty over their data and how their data is used. And we are also extending the same capabilities on premise or using customers own equipment in the form of sovereign cloud platform. So this is something that like we have kind of like started doing like in the sense that like we have extended our cloud platform to kind of like usage by in any form factor. So the earlier form factors like we were looking at were like private cloud, hosted with us, public cloud on any workload running on public cloud, any hybrid cloud workload.
Now we are also extending those workloads to include completely sovereign usage for on premise customers or co located customers who want to use our platform where the platform can completely air gap the way the data gets transferred within their whole compute infrastructure. Now kind of like we have been a partner, major partner for Nvidia and India. We have partnered with intel and amd, the other chip makers for a very very long time for the CPU world. So like having our own software which we have built like over last 10, 11 years or 10 or 12 years has enabled us to deliver like leading price performance ratio for our solutions.
More recently like our strategic partnership with Larbo gives us like access to the enterprise channels. So like these are the some of the things that like are really setting us apart in terms of like our ability to kind of like help with the AI workload. So we kind of started post like the major fundraise that we did in the last couple of quarters as a company, we shifted our focus to. Make sure that we are servicing the needs of larger customers. Now larger customers come with their own requirements in terms of like typically we were working for a long time with customers who were using like maybe 16 GPUs, 8 GPUs, 64 GPUs. Larger customers typically tend to go between 64 GPUs to even 512 GPU clusters.
So what we felt over the last couple of months was that like, yes, we are presumably larger, but we are not able to do enough trials for those large customers. And that is something that has resulted in us not being able to kind of like gain additional mrr. Like MRR has stayed monthly, recurring revenue has stayed constant from December to the March period. So major reason for that has been that like we have been running like a lot of trials with larger customers, larger startups, both directly as well as through our strategic partnerships. And like these are like long cycle sales. And we kind of like felt the need that like there is a lot more waiting for us in the pipeline.
Based on like all the conversations our sales team has been having with various parts of the AI ecosystem in India and based on the ground level feedback, we decided that we need to really build up our capacity for pursuing bigger goals. So we have built additional capacity of 2048 GPUs, which is kind of like getting rolled out and getting integrated with our platform and kind of like we have been a believer in AI ML infrastructure long before Gen AI was cool from 2020 onwards being ahead of the curve. And our belief that like this is a multi, this is at least like multi decade opportunity. AI is a multi decade opportunity, continues to remain very, very strong and we continue to see ourselves as a supporter of AI, sovereign AI in India, supporting like AI startups in India, supporting enterprises, supporting like educational institutions.
So we see some big opportunities. Like so one opportunity that I described was a sovereign cloud platform. So where we are working in a market which, where it is dominated by two players today, both of them, one of them does like about $10 billion of revenue and the license cost is almost like four to five times higher than. Than what we are charging for our platform. And the other player does nearly like $2 billion of revenue. Again, the license cost is very, very similar on a per CPU core basis. So what we feel is that like a lot of our customers who were running like public cloud workloads with us, they also were running like a lot of workloads where they didn’t want to put those workloads on it.
Public cloud platform where they did not control the entire network completely end to end sovereign cloud platform was our answer to having that for their data. Ability to keep it air gapped, ability to work in any kind of regulated and compliant environment with complete vendor independence. So the second thing that like we are focusing on this year your company got impaneled as one of the major cloud GPU providers in India AI Mission. Now we feel that like next couple of weeks this is going to pick up like a lot of steam.
Like one of the major demands that is coming through coming into India AI Mission from what we have heard is basically again like most of that is like public information on India AI Mission site is that like they are inviting a lot of applications for people who want to build sovereign LLMs for India. And we have spoken to quite a few of those applicants and typically we have seen that the requirements run into hundreds to thousands of GPUs for building sovereign LLMs. And we do hope to get a piece of this action in India AI Mission like government has been. Government has allocated a budget of nearly 4,500 crores for Cloud GPU which is 40% subsidy over the overall. Which means that India AI Mission itself will drive a demand, additional demand of AI infrastructure of about 10,500 crores over next three years.
So essentially you’re looking at average new demand in AI in India crores. So essentially that is also a big part of why we chose to do a rapid capacity expansion as opposed to controlled capacity expansion that like we were doing earlier and with access to investment like we felt that this is the right time to kind of like bet big on our capacity expansion. Now India is home to largest number of engineers in the world or the very largest number of engineers in the world. Like we are practically talking about 3 to 4 million engineers who are currently in engineering colleges or. Equivalent like MCA or BCA or. So what we believe is that like AI being the future, a very large portion of these engineers would want directly or through their institutions to access like AI labs as a service where they are able to kind of like learn and do demos on AI and kind of like build a career in AI. So that has also kind of like, that is also something we have announced very recently.
We built a complete product for any student to be able to learn AI. And we believe that like with a very large student base in the engineering side of the 30 to 40 lakh students, even if we are able to take like 10% of that overall market, like we are looking at like thousands of GPUs being consumed in either of these two things that you mentioned in the AI mission and AI labs as a service. We continue to work very strategically with startups that are working in the AI ecosystem where we are identifying good quality partners for building complete solutions, end to end solutions, where many of the partner startups solutions we will take to the market along with them.
With a joint go to market. And like there we feel that like there are like large vertical industry segments, each of which could be like worth like couple of thousand GPUs or a couple of hundred billion dollars of revenue every year from a market size perspective. And we do intend to pursue a couple of these verticals and kind of like build our software into these verticals to expand into these markets which are developing very, very rapidly. We have recently strengthened our team in all its aspects to kind of like pursue more expansion from a sales point of view where we are building like partnerships very rapidly across the ecosystem like the major hardware OEM and the chip makers. And from a channel perspective like channel which reaches out into multiple sectors.
So like we have the team now to be able to kind of start leveraging that and we will continue to leverage our major strategy partnership from one of our major investors to reach out across the globe for selling our sovereign cloud platform. So we have a couple of sex into multiple ways into the market where each of the ways feels like it’s like at least even in the Indian market alone we are talking about overall demand. Of like tens of thousands of GPUs. Now we have built a capacity where the capacity is enough to take us from monthly revenue of like 11 crores to maybe around 35 to 40 crores where we are targeting this number to be like the by end of March 26th or maybe like a quarter here or there. And we believe that like, so this is the MRR number which translates into ARR.
And from a revenue perspective, we do expect to kind of like improve our showing next year from whatever current number we have to maybe about 1.5 to 1.7 times of that now. So these are aggressive goals. We are setting a very aggressive goal of like three times monthly recurring revenue growth. And we have robust deal pipelines where we are doing like Fairly large size POCs, many of which are waiting for most of our new capacity to kind of like get tested and get deployed for these. And we continue to pursue more contracted business compared to on the hour or monthly business. So like, hopefully we feel that like the time is right for our expansion in the next couple of years.
And now I would like to kind of like hand over our call to our CFO Megha, who will briefly touch upon the financial highlights for the year and the quarter ended March 2025. Over to you, Megha.
Megha Raheja — Whole-Time Director and Chief Financial Officer
Thank you, Taran and good evening everyone. Let me first start by giving you some of the key financial highlights. First I will summarize the performance for Q4FY25. For Q4FY25, the revenue from operations to that 335 million which witnessed a growth of 14% on year on year basis, EBITDA margin for the quarter is 40% and PAT margin for the year is 41%. So for the current quarter as compared to 12% for Q4.24, the diluted EPS is 8 for the current quarter as compared to Q4.24 which was 2.4.
Coming to the financial year ending March 25, the total revenue stood at 1640 million which witnessed a substantial growth of 74% on year on year basis. EBITDA for the year is 967 million which further shows a growth of around 102% on year on year basis. EBITDA margin for financial year 25 is 59% which demonstrate a growth of 820 basis points year. Is reported as 475 million which showcase growth of 117% year on year basis tat margin for the year is 29% and diluted EPS is 27.2% for the year which is around 85% increase by on Y
That concludes the update for the quarter and SY25 and now we can open the floor for question and answer session.
Questions and Answers:
Operator
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 the touched on telephone. If you wish to remove yourself from the question queue you may press star and two participants are requested to use handset while asking a question. Ladies and gentlemen, we’ll wait for a moment while the question queue assembles. First question is from the line of Pritesh Cheddar from Lucky Investments. Please go ahead sir.
Pritesh Chheda
Clarification for the opening comments. So you mentioned the lower utilization with respect to clients so that was not clear if you could tell the the asset being put but the utilization dropping is a function of
Taran Dua
Like. Basically I was talking about doing a lot of trials and POCs. So this means that like like paid revenue basis like we did not like pursue greatly because we were having like very very limited capacity. So our goal over the last couple of months has been to focus on the customers who can grow very large for us. In that sense we prioritized future growth over today’s revenue and we have now kind of like done a lot of POCs. We have a lot of POCs waiting from larger customers for us and we feel that like. So we have built the capacity to cater to those customers that we have trialed with and newer customers like who want to trial with us. And we are seeing like quite a robust pipeline of demand for cloud GPU from us. I hope I am able to answer the question.
Pritesh Chheda
Yeah. So based on that are you saying that your MRI will add 35 to 40 crores as an exit number for March 26?
Taran Dua
Yeah, around that time like we should. We are targeting that we should be able to hit like that kind of mirror.
Pritesh Chheda
Okay. The other question is how much capex you will do in FY26 which is the year ongoing. And how much capex you will do in FY27 based on whatever visibility that you have as on date business visibility that you have.
Taran Dua
I think like it would be a combination of number of factors. So infrastructure like get built in wave. So one is of course like we have built a lot of hopper infrastructure where there would be a lot of hopper related demand at some point of time. Like we’ll continue to sell hopper, we’ll obviously look at like a reasonable percentage of that hopper platform we build. And when we see that reasonable percentage of hopper is built, then we will also look at like adding like a lot more Blackwell into the mix. So essentially from the perspective of like how much capacity to build would be a combination of these two factors where we start seeing the demand for the newer architecture.
So like we’ll aggregate that demand and then like kind of like make a conception and like immediately invest into that. And also we want to have like some reasonable amount of like utilization on the hopper before we kind of like start building out infrastructure on Blackwell. That is not to say that like we will not build any infrastructure on Blackwell. Like we’ll obviously like to build like initially a small portion of Maxwell infrastructure compared to the total we intend to build. And with every major generation that comes through, like we obviously would build additional infrastructure. I know that’s a qualitative answer from a quantitative perspective. Like we are very positive that like the growth that we have started seeing the Indian cloud, GPU and AI N market is sustainable and long term.
So I believe that like we could potentially see like, like any, any large number like in terms of like what’s the number of GPU you want to add like potentially like every generation add at least like 2000 GPUs and each generation is about 30 to 40% higher than the previous generation. So if we look at like say next two, three years like we would be targeting like anywhere close to 10,000 GPUs. Plus overall on our platform today we are at like 3700. So anywhere like next two years like we should be able to add like anywhere between like 6,000 or so GPUs, maybe even more.
Pritesh Chheda
Okay, and my last question is, in your capital work in progress, what actually sits in the capital?
Taran Dua
This is the 2048 GPU is being deployed. So that is the capital working for.
Pritesh Chheda
This is 2000. GPU’s it is to look with, it is in tune with the multi order.
Taran Dua
So we have the delivery, it has already been delivered. Like currently these gpus are being tested on a cluster. These individual gpus will always work. But the testers, before you give them out in production or even for POC to any customers need to be very well tested before you kind of like give them out to anyone. So currently Most of these GPUs are undergoing like very rigorous steps that are designed by our team as well as like the tests which are kind of like recommended by the major hardware OEMs.
Pritesh Chheda
Okay, thank you very much and all the best. Thank you.
Taran Dua
Thank you, thank you,
Operator
Thank you. Next question is from the line of Keshav from Nivesha Investments. Please go ahead.
Keshav
Thanks for the opportunity. As you mentioned, like in the previous quarter you have given the details for try, so your expenses must have gone high. But in the financial statements we are not able to see that. So could you please give some color on that?
Taran Dua
So actually our EBITDA this quarter came down to 40% from 60%. Steady state is of course like we intend to get back to 60%. So more or less the expenses have remained the same. But Obviously like the GPUs on POCs which are non revenue or GPUs on trial which are non revenue, the costs have remained like nearly constant over there.
Keshav
Okay. And if you could highlight some, like what, what kind of revenue we can expect from the sovereign cloud, like how much is the opportunity size that we are seeing?
Taran Dua
Opportunity size for sovereign cloud is very, very large. Like the two major companies working on this one, like between the two of them, they do about $12 billion of revenue. One of them does $10 billion of revenue, the other one does about $2 billion of revenue. So from a sizing perspective, like sovereign cloud and sovereign AI global market runs into tens of billions of dollars. Now to pursue that, like obviously we have to kind of like do a lot of work in terms of popularizing, introducing, trialing, working with a lot of customers for that, working with a lot of partners for that. So this year we are not like making any prediction on like what would be the revenue for sovereign cloud platform.
Primarily for revenue, we are projecting only the cloud revenue, cloud GPU revenue based on the amount of capacity that we have built up. So we are making an estimate, this is that which is what we have predicted our in cloud platform would be over and above that, whatever that number is. So like software sales, like initially when you bring it out where you say that yes, we have a better tested software that we have operated for last 11 years where we have solved hundreds of customer problems, where we have solved hundreds of security performance, reliability problems. Like so yes we do have a better tested platform. But then like people who buy this kind of a software, like they have their own set of tests they want to run, they may have like additional feature requirements. So this will definitely take more time. So it’s a, it’s a long get session period on this but again it’s very, very rewarding.
Keshav
Got it. Thank you. Both customers which were on trial last quarter, so what could be the timeline that you can expect? Like the conversion timeline.
Taran Dua
So we, we are seeing, we are seeing like increased conversion already. So now the quantum of conversions like you cannot predict but like in a quarter. But what you can predict is like what you feel would get done in like next four quarters or quarters. That’s easier to predict than saying like what is going to happen like next quarter.
Keshav
Like the customers were on trial last quarter. So like what could be the timing that we can expect like three months, two months or four months. What is the trial period that they generally take?
Taran Dua
So I would again, once again repeat that. These are like long cycle sales where we are asking for a lot of money from the customers we are working with. So like it’s not, it’s very, very hard to predict like when you do the closures, like you could be doing the closure at the beginning of the, you could be doing the closure at the end of the quarter. A single customer with like 128, 256 GPUs makes a massive difference over the MRR. So these things are hard to predict in the short term. So easier to kind of like say that, okay, over a long cycle with a larger number of trial customers on a larger base of GPUs, eventually it would become more and more predictable. So today like again like we are looking at a very small base to begin with. We had very small capacity all through till the end of March. We have now built the capacity. We now have the sales talent and the other talent in the company at the back end to be able to do larger number of POCs and trials. We should expect like currently the predictability is over a yearly time cycle. We should expect like more predictability as we grow bigger.
Keshav
Got it. One last question.
Taran Dua
Sorry, say that again, I didn’t understand your question.
Keshav
Is the training data printer live or.
Taran Dua
It’s like in almost live state. So where we are still kind of like conducting a few more tests before we go live. Like the go live is still like a couple of weeks away, but like a lot of work has already been done.
Keshav
Okay, got it. Thank you. Thank you for the.
Operator
Thank you. Next question is from the line of Shitej Saraf from Tusk Investments. Please proceed.
Kshitij Saraf
Hi Tarun, Good evening. Question, question on the CapEx deployment. So we understand that, okay, we are sitting on the cache and we are at the PoC. So how would you time this capex in terms of the PoCs you’re running and in terms of the deal. So if, if you could just give a sense of what the average size could be working for?
Taran Dua
Definitely, definitely. So we have today built a capacity of nearly 3700 GPUs out of which about 2300 or so of them are H200. And this is based on our assessment of like what sort of demand cycle we are seeing today based on that. Like basically like the CapEx deployment plan is that like in every GPU SKU family, like the Hopper family, like we will kind of like estimate the demand for that family. We’ll invest into that, we’ll go and invest into the black belt. Then we’ll go and invest into the generation after the black belt. Every time we want to build a capacity of at least like minimally, at least like 2000 GPUs and hopefully like next year we might be talking about like maybe do 4000 GPUs in the next generation.
Kshitij Saraf
Okay. And could you share a flavor on the clientele, the size, what kind of workloads will run, what kind of enterprises are we talking about
Taran Dua
Today? It’s like very, very workload dependent. It’s not like from what industry, where you’re coming from. Typically we are working with everyone. We are working with higher education in research, we are working with startups, we are working with media and entertainment, we are working with enterprises. So enterprises includes like all sorts of enterprises. So like we are practically working with everyone in our pursuit of like AI workloads. Okay. And of course with India AI mission, we will start working in a very big way with the government as well.
Kshitij Saraf
Okay, and just another quick one. Are you doing joint go to markets, users, seeing some events wherein the LNT Cloud Synert team and you guys are lively.
Taran Dua
LNT is our largest partner. So obviously we would be doing a lot of activities jointly with lnt. We also have a lot of other partners. Who we are already working with. So these partners include the AI community partners. So where we are kind of like working together with them, where we are doing events for startups, for students, for industry, for enterprises, for government and higher education ecosystem. So we are present Everywhere where Cloud GPUs for AI are required.
Kshitij Saraf
Thank you so much. All the best.
Taran Dua
Thank you.
Operator
Thank you. Next question is from the line of Aditya Agarwal from Sagun Capital. Please go ahead.
Aditya Agarwal
Thank you. I just wanted to understand the economics of India AI Mission because it looks like we have offered the CPU at the market at a discount of about 80% of the market rate and eating after the subsidies.
Taran Dua
That is not true. That is not actually true. So the street price, the difference between street price and what we have offered to India AI mission, the difference is about like 20, 25%. So like you can’t like really compare the MRP of workloads which are very, very infrequent and compare with, compare them with like where the workload coming in are in a much higher volume and almost assured workloads, like long running workloads. So it’s like two very different platforms. There is no customer acquisition cost, there is no sales and marketing cost. And of course like when you go through one single generation of a gpu, like the beginning of the generation, if the GPU is most expensive, by the end of the generation the GPU becomes cheaper. And the same thing also happens when you are selling the cloud gpu. In the first year you make more money.
Then second year to fifth year the money stabilizes, sixth year again it becomes like kind of like slightly lower utilization. But all in all like you’re looking at like five good years in a gpu. Now India mission is a outlier in terms of like can do. We are Talking about like 3500 crores of new demand every year. And this is a demand which kind of like does not require you to go out, sell to one customer at a time. This is a demand where you don’t need to do trials. So your customer acquisition cost is very, very low. So reflect all of those savings and maybe some bit more restaurant with a view to kind of like win the bid. Like typically that’s how the pricing is done. So it’s not like it’s not like crazy pricing that we have offered. Like that’s not the case. It’s like reasonable commercially viable pricing that we have offered.
Aditya Agarwal
Okay, thanks for explaining that. And when do you expect this mission to take off and can we start some good amount of revenue coming with it?
Taran Dua
It’s already taking off. Like I think like hopefully in next couple of weeks we’ll start seeing the workload and all of this coming to India mission.
Aditya Agarwal
Okay, and you have any expectations about what set of numbers will that be?
Taran Dua
See like we are talking about like some 3,500 crores of revenue every year. Now we would obviously go and target like minimally about like say 10% of this revenue and hopefully like we could even like top that 10% to maybe even do like maybe potentially like 20%. So let’s see how it goes.
Aditya Agarwal
Thank you.
Operator
Thank you. Thanks. Next question is from the line of Bhavya Gandhi from Dalal Crocha stockbroking. Please go ahead.
Bhavya Gandhi
Yeah, hi, thanks for the opportunity. My question is regarding the peak revenue that you can generate from the existing set of GPUs. Assuming that you don’t add any other GPUs, what sort of revenue can you generate?
Taran Dua
That would be close to 40 crores. So that’s why like the number of like 35 with you that like there would be some percentage which remains like unutilized for USPS etc or for hourly users to be able to use at any given point of time. So peak should be like all the way up to 40 and reasonable utilization rate of like somewhere between 85 to 90 should result in about 35 or so.
Bhavya Gandhi
Okay, so okay. And so for these incremental capex of 6000 GPUs, basically we are at 3700 GPUs. And you mentioned about a target of 10,000 GPUs for additional 6300 GPUs. Even if you assume a 45 lakh per GPU cost that comes to closer to 2500 to 3000 crores of incremental capex. So how do we intend to fund this?
Taran Dua
So like obviously like once you start generating cash flows it opens up the possibilities of like using a lot of debt. So we have used it in the past. Like we have done very well with that and like we’ll continue to be very, very aggressive in terms of like being able to arrange whatever capital is required for our extension. Either I’ll get or.
Bhavya Gandhi
Yeah, Okay. And with respect to AI and Sofin Cloud overall, if you were to, you know, calculate or separate these businesses, what would be the rough EBITDA margin? I understand there will be some cost savings, but if you were to just separate the entire two verticals, AI Mission and Sovereign Cloud, what would be the margin range?
Taran Dua
It would not be very, very far. Like, I think, like, the difference would not be very high.
Bhavya Gandhi
Okay, so we can. So with AI Mission and Sovereign Cloud also will be able to maintain the 60% sort of EBITDA margins.
Taran Dua
Yes, yes, yes.
Bhavya Gandhi
Okay. And is there any seasonality in Q4? Because I see other expenses, even last year were high in Q4.
Taran Dua
I tried to explain that, like, look, we had very, very small amount of capacity all through, like last couple of months. So yes, some of that capacity actually went idle, which is what we were able to utilize for doing trials with our larger customers. And the thought process was that like, immediate revenue is not that super important. What is more important is to build for the future. So based on like all the conversations and all the trials and the engagements that we have had with a lot of customers, we decided to build the capacity of like overall, like additional 2048 GPU. And that, that, that is definitely the goal. Now, obviously the cost do not come down because, like you were at that cost structure, but the revenue was not there to kind of like. Yeah, yeah, let me just.
Bhavya Gandhi
So in Q1, Q1, Q2, Q3, there was an expense of closer to 2.3 to 2.9 cr. And in the fourth quarter it has directly jumped to 5 crores. And the same period last year it was 4 crores. So just trying to understand what is this variability or volatility in other expenses?
Taran Dua
Okay, so see, we had a very, very small company and a very consistent feedback. We got when we met people that, oh, you guys are doing great. Like, you guys have great technology, you are able to deliver great services, but how come we never heard about you? So basically we had to up our game on the marketing to kind of like obviously correct this. So this is what is showing up in terms of like higher end on the marketing is even marketing and marketing efforts from our side.
Bhavya Gandhi
Okay, and just last thing, on the long cycle customers, what would be the cycle range? Is it like a two year, three year commitment? And what sort of margins is it going to be lesser than the historical margins or how are we going to sign this up?
Taran Dua
I think the. Price and margins would be very very similar to what we have been doing historically. And like typically like we look for yearly renewable contracts, six months to one year renewable contracts and mostly like depending on the customer typically like the six month or one year contracts do get renewed for the same period again and we hope for a reasonable lifetime value of like 2 to 3 years with more stable and larger customers.
Bhavya Gandhi
Okay. And so on the overall demand environment, are we seeing reduced infra spending across India or globe or what is the environment out there?
Taran Dua
No, no, no. Infra trend keeps on improving like so we have seen that in the cloud in last 20 years the like amount of compute you can buy with the money like that can increase especially with newer software and newer SKUs. But the demand continues to remain for everything that has gotten built into the cloud like on a continuous basis. So like the cloud always gets utilized. That demand is never a challenge on the cloud.
Bhavya Gandhi
Fair enough. All the best sir. All the best to you and the team. Yeah. Thank you,
Taran Dua
Thank you,
Operator
Thank you. Next question is from the line of Harshisha from Bees Capital. Please go ahead.
Harshi Shah
Hi Tarun, just wanted to a broad view on how the update on this L and T partnership is going. The impact of you know, gaining enterprise clients and you mentioned marketing. So you know all along those lines how are we place to get more of enterprise client and maybe the next one two years. How do you see the mix changing towards these bigger clients?
Taran Dua
Obviously the LNT partnership is very, very strategic to us. So like I think like like LNT brings like a lot of capabilities to us like so which we have not yet fully started leveraging which we intend to change over like next couple of months and couple of quarters. Now it’s a very large ecosystem, extends across the globe. We have been limited to India. So I think one of the major changes with LNT is automatically being able to kind of like pursue more international business especially for our sovereign cloud platform. Dream big and think big. So think bigger deals, bigger customers, be able to put more towards customer acquisition costs by having a well qualified sales and presales team and strengthening of the overall thought process towards longer and bigger vision. So I think that’s all. All kind of like some of the things that have changed for us in terms of having partnered with someone as capable as ND
Harshi Shah
Understand. And how’s the progress on the Chennai data center?
Taran Dua
Yes, it’s almost ready so we should kind of like go into production mode very very soon. So it’s like up and running already but like we have to kind of like flip the switch and say now we can accept like production customers. So that hasn’t yet happened but like that will happen very very soon.
Harshi Shah
Okay, great. Thanks a lot.
Taran Dua
Yeah, thank you.
Operator
Thank you. Next question is from the line of Amar Maurya from Lucky Investments. Please proceed.
Amar Maurya
Yeah hi sir, am I audible?
Taran Dua
Yeah, hi. Yes we can hear you.
Amar Maurya
I just wanted to understand. Now you said you have how many H200, how many H100, how many N5s and B5s? Now
Taran Dua
They are all mentioned in the presentation. Approximately we have 700H100 and nearly 2300H200. So practically the near latest generation like we have nearly like 3000 also deep use today in the hopper generation.
Amar Maurya
And secondly this 626 crores of capital work in progress
Taran Dua
That is majorly the 2048 H200 that we procured around end of March and which are going live into production very very soon. So they are already up and running in Delhi NCR and they would be kind of like flipped over to production soon in Chennai data center as well.
Amar Maurya
Okay. Okay. So basically in six, in three, four months in Q1 everything will be capitalized, right?
Taran Dua
Yes, in Q1 like everything will go live.
Amar Maurya
And second is that this India mission, right? We, we aspire to what you’re Saying is targeting a 20 market share?
Taran Dua
Yes, absolutely. So over the long term like over a three year period, like we are obviously trying to be a major player in the India AI mission. So we would definitely try and target like somewhere between like 2025% of the market share within India mission.
Amar Maurya
So then we are saying sir, 4,500 crore is the spend of the government and we are saying that
Taran Dua
That is 40% subsidy. Overall spend over a three year period would be 10,500 crores.
Amar Maurya
Yeah, but those are ultimately subsidies. Whatever customer is going to pay that. Right? So for you.
Taran Dua
Yes, for us the whole thing is revenue. Absolutely.
Amar Maurya
Practically you are talking about a 900 code of a revenue just from India.
Taran Dua
It could be possible, yes.
Amar Maurya
Okay. And you mentioned that sovereign cloud will also have a similar kind of margin profile. What your GPU have
Taran Dua
Sovereign cloud is practically like which is like a gap cloud for customers who are running it typically on premise or on their own controlled colocation environment. So you are technically looking at like apart from support sales, marketing cost. So essentially you’re looking at like a 80 to 75% kind of a gross margin profile on power and cloud platform.
Amar Maurya
Okay, so basically you are saying majority of that is going to flow to ebitda. That is what is your understanding here?
Taran Dua
Yes, absolutely. Absolutely. Absolutely.
Amar Maurya
And according to you how much scale up, let’s say in a three year period of sovereign cloud business can happen to you? Because I believe you are targeting here the government business as well as the global business with the NNT partnership. So if you can give us a long term vision about this particular business.
Taran Dua
Long term of course like we are like this is a very, very large market. So even as a player in a very large market like we should be, we are, we are practically looking at like, even if we capture like couple of percentage points of that market that could be worth like hundreds of millions of dollars. But practically like we can’t like really start predicting it today. So we would like to give it some time before we kind of like make some sort of projections around like how big it can be.
Amar Maurya
Okay, but, but early, early indications like what are the kind of RFPs, what is the range of RFPS you are bidding with LNT in government business?
Taran Dua
So like they are all, all sorts of ranges like typically of course like any, any requirements or multi year requirements. So like essentially they could range for very range from very small numbers, sub 100 crores to 2000 crores. Like that’s the intent that like where our software should go along with our partners in larger RSPs. That’s the goal that we have for us. We’ll not talk about like what we are doing currently over there.
Amar Maurya
Got it. Thank you.
Operator
Thank you ladies and gentlemen. In order to ensure that the management is able to address questions from all the participants in the question queue, please restrict your questions to one per participant. If you have any follow up questions, please rejoin the queue. Next question is from the line of VP Rajesh from Bunyan Capital. Please go ahead.
VP Rajesh
Hi, thanks for the opportunity. And I just wanted to check what was the revenue in March 25th MRI
Taran Dua
That is mentioned in the presentation. It’s like nearly about 11 crores or so.
VP Rajesh
And did I hear it right that from 11 crore it will go to 40 crores in the next March? March 26th.
Taran Dua
That’s the goal.
VP Rajesh
Got it. Okay, thank you. I just wanted to clarify that. Appreciate it.
Operator
Thank you. Next question is from the line of Prathamesh Diwar from Tiger Asset. Please proceed.
Prathamesh Dhiwar
Yes, I’m audible.
Operator
Yes, go ahead.
Prathamesh Dhiwar
Yeah, hello. Am I audible, sir?
Taran Dua
Yes, yes please.
Prathamesh Dhiwar
Yeah. So just if I missed it earlier. So just wanted to know about our TIR platform more so as it optics what sort of incremental MRR it can generate for us.
Taran Dua
The team is a integrated part of our cloud platform itself. So it’s like in a way like bundled along with when someone is using like any kind of cloud sku, they have the ability to use the platform with that. So it does not like represent additional revenue on the public cloud platform, but it increases the stickiness of the public cloud platform. It increases the value being produced for a customer on the public cloud platform and it allows the customer on the public cloud platform to kind of like do more with less amount of time spent and less amount of effort spent.
Prathamesh Dhiwar
Okay, got it. So it will kind of increase our contract timeline with the customer.
Taran Dua
So it increases the for the customers. So once they are used to a particular platform. So it’s like any major software, like when you get used to a particular software, you want to continue using that software rather than learning something new and you build your workflows like the software becomes a part of your workflows. So to make making your life easy so broadly that’s the value software, like any software features that we build on our cloud platform.
Prathamesh Dhiwar
Got it, got it. And so the 900 crore revenue we are expecting from AI mission from when are we looking at to
Taran Dua
Over next three years. What’s the distribution in year one, year two, year three? We don’t know that yet. Today.
Prathamesh Dhiwar
Okay, got it sir. Thank you so much. All the best.
Operator
Thank you. Next question is from the line of Neil Munoth from Pico Capital. Please proceed.
Neil Munot
Hi sir, good evening.
Taran Dua
Hello. Hi. Yes please.
Neil Munot
Am I audible?
Taran Dua
Yes, yes.
Neil Munot
Hi sir. My first question is with respect to what is our current capacity transition on the existing GPU excluding the 2048 ensure it.
Taran Dua
Yes, it’s total 3700 GPU. So 3700 minus 248 is the current capacity on GPU that includes multiple generations and multiple SKUs on the GPU starting right from V100 T4s, a 140GB, 880GB, then L40, L4, L40s, a 40 and so on and 830s. Lot of 830s as well.
Neil Munot
Yeah. So what’s the capacity utilization there?
Taran Dua
Capacity utilization like typically like we track like mod on a GPU on number of GPU basis. So we track the capacity utilization overall from a perspective of like what it can generate. Like. So if you subtract the 2048 GPUs then we are looking at like maybe sort of like MRR capacity of around say somewhere close to 16 to 18 crores per month overall. Including all the existing CPU storage and GPU capacity and including this one, the overall MRR capacity comes close to around 40 crores. Obviously subject to variations in terms of what SEUs get used by the customers. Whether these are hourly, whether these are yearly, whether we are monthly and so on.
Neil Munot
Understood sir. This was helpful also
Operator
Sir, if maybe please request you to
Taran Dua
Let him update the question, please. Yeah, yeah.
Neil Munot
Just to interpret this correctly. So the remaining 2048 GPUs would get us an MRR of 22 odd cr. Is that correct?
Taran Dua
Yeah, roughly. Yeah, that would be correct,
Neil Munot
Yeah. And one last question with respect to the early pricing on the AI mission, if you put some color on it.
Taran Dua
Yeah. So like I said that like this is reasonably close to the street price. It’s not way off. And hourly is a representation of like monthly and 6 monthly and yearly pricing. Most customers, they don’t tend to use hourly pricing at all. Most customers are on monthly or 6 monthly or 1 yearly pricing. Basically this is by and large very similar to the street pricing. It’s not very far off.
Neil Munot
So isn’t the current revenue dispatch. The last two quarters with respect to pricing or the mix seems to be a pricing issue rather than a utilization issue.
Taran Dua
See we focused all of our inventory to the future where we focused not on converting that inventory to immediate revenue with smaller customers. We prioritized doing trials and display our capabilities to the customer, customers who could grow potentially very large. And that is where like we spent a lot of our capacity and so that decreased the paid revenue basically utilization on our inventory. So that is the cause of the tears. Practically from a price point perspective, I don’t see like a major challenge over there.
Neil Munot
Okay sir, thank you for taking my questions.
Operator
Thank you. Next question is from the line of Gunjan Kabra from Nivesha Investments. Please proceed.
Gunjan Kabra
Hi Karan, thank you so much for the opportunity. So from the industry dynamics and adoption I wanted to understand that see like the longer term trend of adoption look pretty interesting and intact that way. But the last quarter there were trials and this quarter also we witnessed a lot of trials that day. So where we wanted to convert the customers in a big way. Right you mentioned. So just wanted to understand from this first question what is the percentage growth in terms of new inquiries and trials that you saw basically quarter on quarter. And the second is from E2E perspective and from the industry perspective in India also, approximately considering the customer base also that we have right now in terms of trials that we are witnessing, where do you think this learning curve will translate? From our customer base also and from the industry perspective also will translate into an exponential growth kind of a scenario from overall E2E perspective.
Taran Dua
It’s a multi part question. I don’t understand it fully but I’ll try to answer it as best as I can. So I think like broadly you’re asking like whether there is demand for gpus. Has a demand grown or has the demand come down?
Gunjan Kabra
Perspective. I wanted to understand what the number of sizes.
Taran Dua
I am talking from a 2A perspective so I’m not talking from an overall industry perspective over here. I think like we’ve seen increased demand in terms of like number of inquiries and number of conversations that we have been having. The pipeline has been very, very robust and that is the main reason why we decided to build pesticide. Very, very rapidly. If you remember like we have been talking about like why don’t we build like maybe 256 or 512 GPU clusters like every quarter. And then we changed our strategy and said that like look, there is enough demand. We don’t need to wait that long. We can go ahead and build like 2048 GPU cluster like right away. So that is our view that like the demand is quite robust. And let’s see, long sales cycle with the conversions, you cannot time them. So there are certain objectives of every trial that you need to meet. And then customers, of course, like larger customers have their own processes. They have like multiple stakeholders.
So these conversations, these conversions in the enterprise side, larger customer side, like they obviously take time. So you cannot predict like okay, what’s the sales cycle? So like this is like practically like we started working with customers much larger than us or customers who have much larger scale than we have seen in the past like over last two quarters only. So like we’ll obviously we are also gaining a lot of experience over there. So like I said that we do not have immediate predictability but we are very, very confident that like over a yearly cycle or maybe four or five quarter cycle, like we are targeting that overall like 35 crore to 40 crore kind of like MRR. So that’s, that’s what we are more confident about like to be able to say that.
Okay, do we know what is the measurement of the sales cycle today? The answer is no. Do we know that like when a particular customer will grow from say 8 GPUs to 1000 GPUs? We don’t know the answer to that. Like so many of the customers are also growing very fast in terms of their utilization of cloud GPUs. And sometimes we ourselves are surprised that, okay, oh this customer was like doing like a couple of lakhs of mrr. And now that is like already doing say five times of that or ten times of that. So like these are not like super predictable items but over a longer term on a larger customer base for the capacity that we have built, we are confident that like we’ll kind of like get on with our journey to where we intend to reach.
Gunjan Kabra
Okay. Okay, I understood it. Okay, thank you so much.
Taran Dua
All right, thanks. Yeah,
Gunjan Kabra
Thanks.
Operator
Thank you. Next question is from the line of Sampat Nayak from ZTO Capital. Please go ahead.
Sampat Nayak
Hi Tarun, how are you doing?
Taran Dua
Hi. Yeah, please go ahead.
Sampat Nayak
Two questions. So you said MRR of 35 to 40 cr. So it’s exclusive of India air emission. Right.
Taran Dua
All inclusive.
Sampat Nayak
Okay. And second question on covering cloud platform. So in the agreement with the lnt, You had mentioned number of 30 crore per year. You know you’ll be licensing it for 30 crores and the reseller agreement is for 35 crore. Can you tell me more about it?
Taran Dua
So these are the limits. These are the limits for which we have taken the shareholder approval for related party transaction. Now this is the maximum number for which we have taken the approval. Of course if we exceed those numbers we’ll obviously go and take larger limit approval from the shareholders. So but these are not like, these are not the numbers which represent revenue today.
Sampat Nayak
Right, right. So I mean are they are these numbers per megawatt? I mean like 30 crore license. So you can prevent.
Taran Dua
No, no. So like the licensing metrics kind of like are dependent on the amount of compute capacity. So power is one measure of those. There are other measures also. So like the measure that we have with LNT could be different from what we have with other customers.
Sampat Nayak
Right, right. So also Karun, this SAP platform, so this we lost because are we seeing increased trend of, you know, repatriation? I mean the companies are preparing.
Taran Dua
Absolutely, absolutely. Absolutely. So there is a very broad and increasing trend of cloud repatriation that we are seeing. And we are seeing like a lot of conversations around our sovereign cloud platform.
Sampat Nayak
Okay, okay. So say, I mean India comes up with, you know, 4 gigawatt of capacity in next two, three years. Right. So are they in a position
Taran Dua
Every, every megawatt. Every megawatt of compute from a software licensing perspective, although the license terms would not be in terms of power. So every megawatt of like compute capacity that gets built on the private cloud or sovereign cloud side, that represents about, let’s say anywhere close to 2 to 3 million dollars of revenue on the software side.
Sampat Nayak
Right, right. So coming to that, so if India comes up with say 4 gigawatt of, you know, data center capacity,
Taran Dua
So let’s say 2 gigawatts of that is not amongst the public cloud. 2 gigawatt is being used by say private cloud. Then that represents about like $5 billion of software market for infrastructure software.
Sampat Nayak
Right, right. So are we in a position to you know, capture at least 10% of the market?
Taran Dua
See like we have already said that like this is like early days for sovereign S platform. So like let us, let us wait for a. Couple of quarters before we are able to kind of like, have predictability around the sovereign cloud platform.
Sampat Nayak
All right. All right. So. Okay. Okay. Thank you so much. And all the best.
Taran Dua
Yeah. Thank you.
Operator
Thank you. Ladies and gentlemen, due to time constraint, we will take that as a last question for the day. I would now like to hand the conference over to the management for the closing comments.
Taran Dua
Yeah. So thank you everyone for attending our conference call. Like, we hope we have answered all your questions. In case, like, you have additional questions, please do reach out to us on email id investors networks.com and we will try our best to answer the questions. And thank you everyone, once again and have a good day.
Operator
Thank you on behalf of Goindia Advisors. That concludes this conference. Thank you all for joining us. And you may now disconnect your lines.