E2E Networks Ltd Q1 FY2026 Concall Summary

Key Takeaways

• Revenue hit by Chennai delay: Revenue fell 12.6% YoY to ₹36.1 crore and the company posted a PAT loss of ₹2.8 crore, as costs for the new Chennai facility began in April but revenue from that capacity was delayed by technical issues.
• MRR exit rate jumped to ₹14.5 crore: End-of-quarter MRR rose from ~₹11.2 crore to ₹14.5 crore, driven by new POCs and a larger pipeline for large GPU clusters, even without Chennai contributing.
• Chennai resolution imminent, demand waiting: The 1,024 GPUs in Chennai are expected online by end of July or early August, and management says the sales pipeline was not impacted by the delay: demand exists and will convert as soon as the capacity is live.
• Exit MRR target of ~₹35 crore reaffirmed: Management remains on track for a March 2026 exit MRR of ₹32–38 crore, assuming 75–90% utilization of current capacity, with the majority of revenue coming from Noida.
• What to watch next quarter: Whether Chennai goes live on the promised timeline and begins converting its pipeline into revenue, the pace of MRR ramp toward the ₹35 crore exit target, and any clarity on INDIAai Mission revenue contribution.

Chennai Delay: What Broke and When It Fixes

• Timing mismatch crushed the P&L: Fixed OPEX nearly doubled as the company invested in people, technology and a second major location in Chennai, but the revenue those costs were meant to support did not arrive in the quarter.
• Deployment stretched to 10–14 weeks: Setting up a large GPU cluster outside Delhi NCR for the first time extended deployment from the normal 4–6 weeks, as the company worked through GPU and network security hiccups with multiple technology partners.
• Complexity, not demand, was the bottleneck: The delay stemmed from inherent complexity in running large GPU clusters, requiring deep technical effort from in-house teams and major hardware partners; Noida went online mostly without issue, but Chennai hit snags at multiple points.
• Resolution expected by mid-August at the latest: Issues are expected to be resolved early this quarter, by end of July or latest before the 15th; the 1,024 GPUs currently in CWIP should go online by end of July, worst case first week of August.
• Management declined to quantify the MRR impact of the Chennai delay, stating that what matters is the exit MRR/ARR, not the journey.

Dua: "Now cost we started incurring for the new facility starting from April, now that should not super drastically change. Obviously, with higher utilization, some costs will change, but not too drastically over the next couple of quarters. But yes, that has hit our numbers hard"

MRR Trajectory and the Path to ₹35 Crore

• MRR jumped ~30% in the quarter: End-of-quarter MRR reached ₹14.5 crore, up from ~₹11.2 crore at the end of the previous quarter, driven by new POCs and a larger pipeline for large GPU clusters ranging from 32 to 1,024 GPUs.
• Growth came from both existing and new large customers: Major traction is on training workloads, though model-as-a-service has also gone online on the platform.
• Training workloads are lumpy but diversifying: The majority of new large GPU cluster workloads are training, which is inherently lumpy; management is working to build more inference workload customers with year-or-longer contracts, and as the customer cohort grows, lumpiness reduces with a larger base.
• Exit MRR target assumes 75–90% utilization: The ₹35 crore (+/-) March 2026 exit MRR target is based on current capacity; management expects to reach that utilization band by Q4 this year.
• No quarter-on-quarter roadmap provided: Management declined to give a quarterly breakdown of the path to the exit MRR target.

Dua: "We are very, very confident of this pipeline that we will have the conversion over here, over next many coming months."

Capacity Expansion: Chennai, Noida and Blackwell

• Chennai can scale to 20,000 GPUs: The facility can host close to 20,000 GPUs very easily once the initial deployment issues are resolved, and management says additive capacity should not face significant delays going forward.
• Noida can reach at least 6,000 GPUs: Existing Noida capacity can expand to at least 6,000 GPUs if required, possibly more.
• Blackwell investment is definite, timing is fluid: Management intends to invest in Blackwell GPUs and does not expect acquisition challenges given its early NVIDIA partnership; if capacity addition starts by December, it may go online by March; if added in October, capacity expansion could appear in the last quarter of FY26.
• Capacity addition is gated by utilization or anchored demand: The company wants to generate utilization on existing capacity before adding more, or see independent demand anchored by a few customers; it remains aggressive on capacity addition overall.
• Non-flagship GPUs and AMD remain on the table: The company will continue adding non-flagship GPU capacity as required, and is open to evaluating AMD or other LPU providers if customer demand or price-performance dynamics shift away from NVIDIA.

Dua: "we are one of the earliest partners of NVIDIA, we have a very great relationship with NVIDIA. So, I do not think we would have a challenge in terms of acquisition of Blackwell."

INDIAai Mission and Software Revenue: Early, Unquantified

• INDIAai revenue is small and not separately disclosed: The company received some small revenue from INDIAai Mission in Q1, but management declined to provide a ballpark number or forward target, saying it needs one or two quarters for clarity on delivery and consumption.
• More significant revenue expected soon: Management expects more significant revenues from INDIAai Mission this quarter or next quarter, with a tailwind from being an L1 provider as more workloads come through over the next couple of quarters.
• INDIAai margins will contract but not hurt overall margins: Management acknowledged some margin contraction on INDIAai work, but said it does not expect that to impact overall company margins.
• Software licensing revenue is a next-FY story: No major revenue number is budgeted this year for software licensing; the company expects that to come online in the next financial year only.
• Model-as-a-service is live but early: Certain models have been made available as model-as-a-service on the platform, but software revenue overall remains in very early days with no near-term promises.

Dua: "It's very early days. We are working towards things. We do not want to promise something today, but hopefully over the next many quarters, we will make our full efforts to kind of come to a place, where we are able to start counting software revenue."

Margins: Compressed Now, Recovery Expected

• EBIT margin contracted on structural investment: The impact came from building up a team for future growth and building capacity at a second major location, which nearly doubled fixed OPEX.
• Cost structure should not change drastically from here: With higher utilization, some costs will change but not too drastically over the next couple of quarters; no significant major changes to the cost structure are expected.
• EBITDA margin of 65–75% seen as doable at exit: Management expects 65% to 75% EBITDA margin to be achievable with the capacity expected to convert to revenue by the March exit month, though it would not commit to 70% specifically.
• Operating margins should improve as revenue from deployed capacity comes online without a corresponding step-up in costs.

Competitive Position and GPU Economics

• Early-mover lead in GPU cloud software: Management believes it maintains a long lead over competitors through its software platform built around GenAI and GPUs, particularly its TIR platform, which has been running as a self-service portal since 2014.
• GPU lifecycle supports long revenue tails: GPUs have a six-to-eight-year lifecycle from development to production to maturity to expansion, and revenue-earning capability does not necessarily decline over time; the company depreciates GPUs over six years.
• L&T and QpiAI partnership is aligned but undefined: The quantum computing angle is not currently well understood by E2E, but management sees quantum users as future cloud GPU users and is fully aligned to work with L&T on joint solutions.
• Management declined to comment on how L&T is thinking about its 22.5 MW DC capacity or whether another player is filling that gap.

Dua: "we were early to the game. We continue to maintain a lead in terms of a software built around gen AI and GPUs, especially our tier platform. So, we do believe that we have got a long lead compared to many of our competitors."

Guidance & Commitments

Key Metrics