E2E Networks Ltd Q3 FY2026 Concall Summary

Key Takeaways

• MRR target on track, but capex and guidance withheld: Management says it is 70–80% of the way to a ₹35–40 crore MRR run-rate by March 2026, yet declined to provide capex or next-year MRR guidance, asking investors to wait.
• India AI mission is the near-term catalyst: Contribution was muted in Q3 but is expected to ramp massively in Q4 as customers won through the mission scale up, with a second round of bidding for ~15,000 Blackwell GPUs already underway.
• Blackwell deployment is live, revenue timing is vague: 1,024 B200 GPUs have been received and deployment is underway, with a go-live expected before Q4 end; management is “keeping fingers crossed” on order wins.
• PAT loss is a deliberate build-out cost, not an operating miss: The loss stems from ₹47.6 crore in depreciation on newly commissioned GPU clusters and finance costs on term loans, while EBITDA margin expanded QoQ toward a stated 70% target.
• What to watch next quarter: Whether India AI mission revenue actually materializes at scale, the pace of Blackwell order conversion, and any concrete capex or MRR guidance management finally provides.

India AI Mission & Sovereign GPU Demand

• Q3 was a soft start, Q4 is the ramp: India AI mission contribution was muted because December was the most valuable month; management expects Q4 to be “massively different” as customers scale up in January.
• Training workloads dominate, inference is next: The majority of India AI mission workloads are focused on building sovereign LLMs via training; the company is already working with LLM builders to run inference workloads and expects this trend to accelerate.
• Second round of bidding is live: The government is planning another round for ~15,000 Blackwell GPUs, with a bid submission date of January 18; management interprets the ₹10,000 crore allocation as an initial outlay, not a cap.
• Payment cycle improved: India AI mission has recalibrated its payment cycle from quarterly to monthly, a working-capital positive.
• Sovereign AI positioning: The company continues to deliver on a sovereign AI promise, positioning itself as a unique Indian business with built infrastructure software.

Tarun Dua: “I think going out on a limb over here, I think Indian government intends to continue to support the AI mission in a big way.”

Blackwell Deployment & Pricing

• Hardware received, deployment underway: 1,024 B200 GPUs have been received at the Chennai location and the deployment process is in progress, with a go-live expected before the end of Q4.
• Pricing benchmark set at \$3–\$4 per hour: Management is targeting this range for slightly longer-term contracts, citing international market rates; order win timing remains uncertain.
• Majority of Blackwell capacity earmarked for India AI mission: The company continues to build capacity beyond the mission as well, but the bulk of these GPUs will serve sovereign AI workloads.
• Next batch is an ongoing process: Management is working on expanding capacity across multiple types of Blackwell GPUs and other SKUs, but gave no timeline or volume commitment.

Revenue Momentum & Demand Environment

• Non-India AI revenue dominated the quarter: The majority of Q3 revenue came from outside the India AI mission, reflecting an improved ability to capture short-term demand compared to the past.
• Short-cycle demand filled the gap: The company sold capacity for shorter projects in the quarter while India AI mission only began contributing in December.
• Enterprise traction via L&T partnership: Actual conversions are happening among enterprise customers through the L&T partnership, with more expected in the near and medium term; management declined to quantify the pipeline.
• Voice agents emerging as a use case: The first major voice agent deployment in India is going online, and management sees interest only increasing.
• Inference marketplace ambition: The company expects to set up a marketplace of inference providers on top of its platform.
• Demand outlook is positive: Management cited increased demand for capacity recently, with India AI mission demand expected to rise further.

Tarun Dua: “Very, very likely. We are already working with the LLM builders to run their inference workloads with us.”

Margins, Costs & the PAT Loss

• PAT loss is depreciation-led, not operational: The loss was primarily driven by ₹47.6 crore in higher depreciation from large GPU deployments commissioned in FY25 and FY26, plus finance costs from term loans drawn for infrastructure expansion.
• EBITDA margin expanded QoQ: Management stated it is a matter of establishing scale to hit the expected number closer to 70%, up from the current 55–60% range.
• Cost control is tight: Total expenses of ₹30.4 crore were in line with business expansion; the business model allows incremental revenue to flow through at higher margins.
• Asset useful life assumption: Management expects 7–8 years of revenue-generating life from GPU assets, versus the 10–11 years used by domestic and traditional providers.
• Hardware cycle risk is acknowledged but dismissed: Memory price increases on GPUs are expected to balance out; management has seen these cycles repeatedly over 16–17 years and expects supply/demand irrationality to fade.

Tarun Dua: “I think quarter-on-quarter now, we have expanded the EBITDA margins. And I think it's a matter of establishing the scale at which point we will be able to hit the expected number as per the business, which is closer to 70 than closer to 55 to 60 today.”

Jarvis Labs Acquisition & Global Ambition

• Acquisition closed this quarter: The asset purchase from Jarvis Lab is a technology-centric move to rapidly position E2E to scale in the global market; substantial Q4 revenue contribution is not expected.
• Jarvis brings a self-service platform: Jarvis built a highly integrated platform for lower-ARPU, higher-margin, fully self-service customers, with a majority of its customers operating from outside India.
• Scaling expected in the medium term: Management expects Jarvis to get scaled up outside India and add to overall capabilities over time.
• R&D investment continues: The company will keep investing heavily in software and AI capabilities, creating standards-based infrastructure with software layers for quick utilization and ROI for customers.

Mumbai Outage & Infrastructure Resilience

• Mumbai was already a sunset location: The Mumbai facility was substantially smaller, and the plan was always to migrate customers to Delhi NCR and Chennai locations; a migration plan to newer, stabler infrastructure is underway.
• No material medium- or long-term revenue impact expected: Management does not foresee a material revenue hit over the medium and long term from the outage.
• Remediation is underway: Investments into site reliability engineering have increased; the company is working with back-end vendors to strengthen overall infrastructure and helping customers build architectures that withstand single-zone failures.

Tarun Dua: “We don't see -- we don't expect to see a material revenue impact over the medium term and the long term. But that being said, we are very, very conscious of working very hard towards making sure that these kind of incidents are completely avoided.”

Guidance & Commitments

Key Metrics