E2E Networks Ltd Q4 FY2025 Concall Summary

Key Takeaways

• MRR stalled as trials consumed capacity: Monthly recurring revenue was flat from December to March because the company prioritized unpaid trials for large customers (64 to 512 GPU clusters) over converting inventory to paid revenue with smaller clients.
• Massive GPU buildout positions for sovereign AI demand: E2E now operates nearly 3,700 GPUs, making it the largest GPU cloud installation in India, with a further 2,048 H200s procured and going live in Q1.
• IndiaAI Mission is a material, near-term revenue opportunity: Management targets 10% to 20% of the ₹3,500 crores annual IndiaAI Mission spend, with workloads expected to start flowing in the next couple of weeks.
• Margins compressed sharply as investment cycle ramped: EBITDA margin fell to 40% from 60% as non-revenue GPU trials and a jump in marketing spend hit the P&L; management intends a return to 60% steady-state margins.
• What to watch next quarter: Whether the 2,048 H200s going live convert the trial pipeline into paid MRR, the pace of margin recovery, and any initial revenue contribution from the IndiaAI Mission.

The Trial Capacity Trade-Off

• MRR flatlined by design: Monthly recurring revenue held constant from December to March because management deliberately allocated GPU inventory to unpaid proofs of concept for larger customers rather than converting it to immediate paid revenue from smaller clients.
• Customer profile is shifting upmarket: The company is now pursuing customers that require 64 to 512 GPU clusters, a step change from the previous base of 8 to 16 GPU customers, and these larger prospects have longer, multi-stakeholder sales cycles that management cannot yet predict.
• Revenue dipped on utilization, not pricing: The top-line softness over the last two quarters reflects lower paid utilization as capacity went to trials, not any deterioration in pricing power.
• Early signs of customer ramp: Some customers that were doing a couple of lakhs in MRR are now running at 5x to 10x that level, suggesting the trial-to-production flywheel can work once capacity frees up.
• Pipeline is robust but conversion timing is opaque: Inquiries and conversations have increased materially, and many large POCs are waiting for the new 2,048 GPU cluster to come online, but management explicitly states it cannot predict near-term sales cycles.

Tarun Dua: "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 customers who could grow potentially very large."

Capacity Buildout & Infrastructure

• Largest GPU cloud in India: E2E now operates nearly 3,700 GPUs, up from the 700 to 1,000 Hopper-series GPUs previously available among Indian players, making it practically the largest GPU cloud installation in the country.
• 2,048 H200s procured and going live: The ₹626 crores in capital work in progress is predominantly the 2,048 H200s acquired around end-March; they are already running in Delhi NCR and will flip to production in the Chennai data center in Q1.
• Chennai facility nearly ready: The Chennai data center is up and running but not yet accepting production customers; go-live is expected in a couple of weeks.
• 10,000 GPU ambition: Management targets close to 10,000 GPUs on the platform in the next 2 to 3 years, planning to add at least 2,000 GPUs with every major chip generation and initially building a small Blackwell infrastructure tranche.
• Funding posture is aggressive: Once cash flows begin, the company will consider debt and intends to remain very aggressive in arranging whatever capital is required for expansion.

IndiaAI Mission & Sovereign AI

• Paneled as a major Cloud GPU provider: E2E has been selected as one of the major GPU providers under the IndiaAI Mission, with demand from applicants building sovereign LLMs running into hundreds to thousands of GPUs.
• Revenue target is explicit: Management targets minimally 10% of the ₹3,500 crores annual IndiaAI Mission revenue, with an aspiration to reach 20%, and a three-year market share target of 20% to 25%.
• Pricing discount is 20-25%, not 80%: Management pushed back on an analyst's suggestion that GPUs were offered at an 80% discount, stating the difference between street price and the IndiaAI Mission offer is about 20% to 25%, and that high-volume assured workloads cannot be compared to infrequent-use MRP.
• Workloads expected imminently: The IndiaAI Mission is already taking off, and management expects to start seeing workloads flow through in the next couple of weeks.
• Sovereign Cloud Platform is a longer-duration bet: The platform, built over 10 to 12 years, enables fully sovereign CPU and AI workloads with air-gapped data transfer for on-premise and co-located customers; management is not making any revenue prediction this year and asks for a couple of quarters before predictability emerges.

Tarun Dua: "We are talking about some INR3,500 crores of revenue every year. Now, we would obviously go and target minimally about 10% of this revenue. And hopefully, we could even top that 10% to maybe even do maybe potentially 20%."

Partnerships & Channel

• L&T is the flagship strategic partner: The partnership gives E2E access to enterprise channels and, critically, a path to international business, especially for the sovereign cloud platform; management intends to start leveraging L&T's capabilities more fully over the next couple of months and quarters.
• L&T agreement has defined limits: Shareholder approval covers up to ₹30 crores in licensing and ₹35 crores in reseller arrangements; if these are exceeded, the company will seek approval for a larger limit.
• NVIDIA, Intel, AMD relationships: E2E is a major partner for NVIDIA in India and maintains long-standing partnerships with Intel and AMD for CPU infrastructure.
• Global reach via strategic investor: The company plans to leverage its partnership with a major investor to sell the Sovereign Cloud Platform across the globe.

Tarun Dua: "We have been limited to India. So, I think one of the major changes with L&T is automatically being able to kind of pursue more international business, especially for our sovereign cloud platform. Dream big and think big."

Margins & Costs

• EBITDA margin halved on idle capacity: EBITDA fell to 40% from 60% because GPUs allocated to trials generated no revenue while costs stayed nearly constant, and some capacity sat completely idle during the trial phase for larger customers.
• Marketing spend jumped in Q4: Other expenses rose to ₹5 crores from a ₹2.3 to 2.9 crores range in prior quarters, driven by a deliberate step-up in marketing after consistent feedback that the market was unaware of E2E's technology and services.
• Steady-state margin target is 60%: Management intends to return to 60% EBITDA margins once utilization normalizes, and believes both IndiaAI Mission and sovereign cloud workloads can sustain that margin profile.
• Sovereign cloud carries 75-80% gross margins: Stripping out support, sales and marketing costs, the sovereign cloud platform is expected to deliver a 75% to 80% gross margin profile.

Product & Market Expansion

• AI Lab as a Service launched: A complete product for students to learn AI targets India's 30 to 40 lakh engineering students; even 10% penetration would consume thousands of GPUs.
• Cloud repatriation is a tailwind: Management reports a broad and increasing trend of cloud repatriation, driving conversations around the sovereign cloud platform and the recently launched SAP platform.
• Vertical industry push planned: The company intends to pursue a couple of verticals and build software into them, with each vertical potentially worth a couple of thousand GPUs or a couple of hundred billion dollars of revenue annually.
• Client base is workload-driven, not industry-specific: E2E works across higher education, research, startups, media and entertainment, and enterprises of all types, with demand tied to specific workloads rather than vertical concentration.

Guidance & Commitments

Key Metrics