01 — At a Glance
The Loss-Making AI Cloud Company That’s Worth ₹4,500 Crores
- 52-Week High / Low₹3,895 / ₹1,732
- Q3 FY26 Revenue₹700 Mn
- Q3 FY26 PAT-₹57 Mn
- Full-Year FY25 Revenue₹164 Cr
- TTM (Trailing 12M) Revenue₹183 Cr
- Book Value per Share₹783
- Price to Book2.81x
- Operating Margin (OPM)56.6%
- Debt / Equity0.10x
- EV/EBITDA31.5x
Auditor’s Opening Note: E2E posted ₹700 Mn revenue in Q3 FY26, up 68.3% YoY and 59.8% QoQ — one of the snappiest revenue accelerations in Indian cloud history. Operating margin hit 56.6%. PAT was negative at -₹57 Mn, purely because depreciation and finance costs murdered the bottom line (hello, ₹100+ crore GPU capex binge). The real story: EBITDA of ₹396 Mn up 60.9% YoY. This is not a cash-destruction machine. This is a depreciation haircut while the core infrastructure prints money. Exactly what you’d expect from a company buying 1,024 Blackwell GPUs. The stock is up 125% in 5 years. Patience pays.
02 — Introduction
The Company That Sells Supercomputer Rental Cards as Monthly Subscriptions
E2E Networks is India’s leading AI/ML cloud infrastructure provider. Not a “platform play.” Not a “software SaaS business.” Not a fintech aggregator. Just: Give us your ₹35,000 a month, get a NVIDIA H100 GPU in a data centre in Delhi or Chennai, train your AI models, come back next month. Rinse, repeat. Scaling achieved.
Since FY21, the company has scaled revenue from ₹35 crore to ₹164 crore in FY25 — a 46% CAGR. In one quarter (Q3 FY26), they did ₹70 crore of the ₹164 crore annual FY25 revenue. Do the maths. We’re looking at an annualized run-rate north of ₹280+ crore by March 2026. Growth is not slowing. It’s exponentiating.
Founded by Tarun Dua (a founder in cloud infrastructure since the mid-2000s), E2E has attracted ₹1,500+ crore in funding over FY25 alone. L&T (19% stake), government contracts (₹265 crore IndiaAI orders), and a cohort of 2,500+ active customers across e-commerce, fintech, and AI startups. Three data centres in India. More than 5,000 GPUs deployed or committed. And a December 2025 outage in Mumbai that nobody will remember because the money tap hasn’t stopped flowing.
This is not Acme Corporation selling nothing to nobody. This is a real infrastructure play in a space where compute is the new oil, and India suddenly owns the pump. The challenge: PAT is negative because accounting. The opportunity: EBITDA is positive and accelerating.
Concall Candour (Jan 2026): “Monthly revenue run-rate reached INR 280m by December 2025.” — Management. Translation: The company is now selling ₹280 crore a month in GPU time. If they hit the March guidance of ₹35–40 crore monthly recurring revenue, FY26 full-year could hit ₹280–300 crore in revenue, a near-doubling from FY25. They said it casually in a concall, like mentioning the weather.
03 — Business Model: We Rent You Expensive Computer Cards
NVIDIA GPUs. Indian Data Centers. Monthly Subscriptions. That’s It.
The model is refreshingly simple: E2E acquires NVIDIA GPUs (H100, H200, L40S, etc.) at ~₹40–50 lakh per unit, installs them in air-conditioned data centres in Noida, Delhi, Chennai (and soon Mumbai + Bangalore), and rents hourly access to AI researchers, startups, and enterprises at global rates. H100 pricing globally: $2–4 per hour. E2E targets the same. Do the maths: 3,900 GPUs × 30 days × 24 hours × $3/hour = over $10 million monthly gross if fully utilized at that rate. Add the IndiaAI contracts. Add government workloads. The ARR potential is staggering.
Revenue mix is twofold: (a) Enterprise Subscription — Zomato, Nykaa, CarS24, etc. using GPUs for AI inference and training. (b) Government Contracts — India AI Mission orders for sovereign LLM development and inference. Both cohorts sticky, high-margin, and growing in tandem.
Utilization is the key metric management obsesses over. By end-Dec, utilization was 60–65% (low for a new infra play). Management targets 80–90% by FY27. At 80% utilization with 5,000 GPUs, the company will print money at a scale that’ll make ₹700 Mn quarterly revenue look like appetizers. The only constraint: can they keep filling the capacity? Answer from the concall: 1,024 B200 Blackwell GPUs arriving in Chennai in Q4. That’s the answer.
Current GPUs~5,000Deployed+Committed
Utilization60–65%Target: 80–90%
Q3 OPM56.6%Operating Profit
Capex Planned₹600–650 CrFor B200 series
Blackwell Moment: Management disclosed they’ve already received 1,024 B200 GPUs (NVIDIA’s latest, 3x faster than H100) and deployment is live in Chennai. ARR per B200 batch: ₹245–250 crore on “bottom price.” Do that 5 more times, and you’re looking at a ₹1,500+ crore revenue run rate by FY27 on just B200s. The depreciation hit will be real. The TAM (total addressable market) is astronomical.
💬 Do you think 80–90% utilization is achievable, or will over-capacity doom margins? Drop your take in the comments!
04 — Financials Overview
Q3 FY26: The Plot Twist
Result type: Quarterly Results | Q3 FY26 Revenue: ₹70 Mn | Annualised Revenue (Q3×4): ₹280 Mn | Full-year FY25 Revenue: ₹164 Cr
| Metric (₹ Mn) |
Q3 FY26
Dec 2025 |
Q3 FY25
Dec 2024 |
Q2 FY26
Sep 2025 |
YoY % |
QoQ % |
| Revenue | 700 | 416 | 438 | +68.3% | +59.8% |
| EBITDA | 396 | 245 | 180 | +60.9% | +120.2% |
| EBITDA Margin % | 56.6% | 58.9% | 41.1% | -230 bps | +1,550 bps |
| Operating Profit | 396 | 245 | 180 | +60.9% | +120.2% |
| Depreciation | 476 | 178 | 428 | +168% | -11.2% |
| PAT | -57 | -2 | -135 | -2,750% | +57.8% |
The Depreciation Assassination: E2E’s PAT is negative not because the business is dying, but because GAAP accounting says every GPU depreciates over 5–7 years. The company loaded ₹600 crore in capex in FY25 and another ₹400+ crore estimated for FY26. That’s nearly ₹1,000 crore in GPU acquisition translating into ₹400–500 crore annual depreciation. Operating profit? ₹396 Mn in Q3 alone. EBITDA? Growing fast. But slap depreciation on top, add finance costs for the debt used to buy the GPUs, and the bottom line goes negative. This is not Amazon Web Services (AWS) losing money in the early 2000s because they were spending on infrastructure. This is E2E doing exactly the same thing — and the market is rewarding it like a unicorn. The concall made clear: management targets a long-term ROE “north of 20%,” implying profitability swings once utilization stabilizes and capex tapers.
05 — Valuation: The 31.5x EV/EBITDA Question
Is ₹4,500 Crore Fair for a Loss-Making Startup?
