E2E Networks remains an operating-leverage thesis, but the Q2 FY27 evidence shifts the proof point from capacity installation alone to contracted utilisation of installed Blackwell capacity. The company owns NVIDIA GPU clusters and sells them through its own cloud and TIR software layer; management says most GPUs run on its platform and very little is sold as bare metal. Q2 FY27 revenue was ₹223.8 crore, EBITDA ₹171.5 crore at 76.6%, and PAT ₹72.7 crore; H1 revenue was ₹380.5 crore and EBITDA ₹289.4 crore. That margin is exceptional against management's 70%+ medium- and long-term EBITDA guide, so the key question is not whether the first Blackwell cluster can earn high incremental margins, but whether the next capacity is contracted and absorbed before depreciation and interest build.
The roughly ₹1,000 crore sovereign-AI Blackwell contract to June 2029 materially improves visibility. It converted existing capacity from pay-as-you-go to a multi-year commitment, turning deployed GPUs into a longer revenue stream and supporting the high reported margin. But the annual contribution is not disclosed and depends on the ₹1,000 crore amount and time period; it is not a revenue guide. The next 1,024 B200 lot was still expected to go live only in the current quarter at the Q2 call, and management has not decided whether to contract it long term. Q1 FY27 reported about 5,100 live GPUs, excluding the next 1,024 B200s; Q2 did not update the live count. The earnings path therefore depends on commissioning the next lot and converting it into revenue before depreciation on that lot becomes material.
The business economics are attractive only while utilisation and contract duration hold. The first 1,024 B200s cost roughly ₹600-650 crore, partly funded by Axis and HDFC term debt; debt was about ₹450 crore at Q1 FY27. Q2 did not update debt, but management said there is no mental upper limit and it is comfortable while cash flow supports repayment and covenants. GPUs are depreciated over six years for accounting, while management argues 7-8 years is a reasonable economic life. The spread between those numbers matters: if older Hopper or early Blackwell capacity loses pricing before the depreciation schedule runs down, reported profitability can lag EBITDA. The roughly ₹1,000 crore contract reduces that risk for existing capacity but does not remove it for the uncontracted next lot or older fleet.
Management delivery is mixed but has improved. The March 2026 MRR target was met, the margin path was exceeded, and Chennai eventually went live on 1 August 2025. But the first Blackwell cluster slipped from Q4 FY26 to Q1 FY27, and the next B200 lot is still not live. The company has a pattern of delivering the financial destination after infrastructure dates slip. Q2 added a new Chief Revenue Officer, a founder's office team, a field studio and strategic advisor Anand Roy; these are early and not yet evidence of revenue acceleration. Software licensing remains an open promise: management expected some action in FY27, but no software revenue or new date was disclosed in Q2. The platform strategy is real, with some GPU capacity reserved for internal platform work, but it is not yet a separately monetised software line.
The current mix reduces one concentration while adding another. IndiaAI was roughly 13-15% of Q2 revenue and international 20-25%, both lower than Q1. The roughly ₹1,000 crore sovereign-AI contract adds a large single-customer commitment for existing capacity. Management gives no MRR, exit revenue, contracted-mix or speculative capex guidance. It says pricing uplift is mainly available on uncontracted or short-term capacity and that it honors contracted and verbal commitments. That is good practice, but it means a larger contracted book can lag spot price increases while spot capacity remains exposed to swings.
The thesis needs three things to keep working. First, the next 1,024 B200 lot must commission in the current quarter and be sold, whether on a long-term contract or high-utilisation short-term terms. Second, the 76.6% Q2 EBITDA margin must stay comfortably above the 70%+ medium-term guide as new capacity and internal GPU use rise. Third, debt and interest must remain serviceable from operating cash flow without a dilutive equity raise at an unfavorable time. The main falsifiers are a further B200 delay, a weak ramp on the roughly ₹1,000 crore contract, a fall in EBITDA margin toward 70% or below before new revenue arrives, or a decision to add capex without contracted demand. The platform and software licensing optionality would strengthen the thesis, but Q2 did not prove it; it remains an execution milestone rather than a current earnings driver.
companyname: E2E Networks Limited ticker: E2E sector: Cloud Infrastructure / AI / GPU-as-a-Service E2E Networks is an Indian cloud infrastructure company that sells access to NVIDIA GPUs for AI and machine learning workloads. Founded in 2009, it runs data centers in Noida, Chennai, and Mumbai, and reported about 5,100 live GPUs at Q1 FY27. At the Q2 FY27 call in October 2026, management said the next B200 lot was expected to go live in the current quarter; no updated live GPU count was disclose...
Read the full report →margin expansion
Explicit forward guidance is limited: management guides a 70%+ EBITDA margin over the medium and long term, expects the next B200 lot to go live in the current quarter, and plans B300/Vera Rubin without quantities or dates. It gives no MRR, exit revenue, contracted-mix or speculative capex guidance, and says a capex upgrade will be disclosed when justified.
Guidance maintainedmixed
Get valuation models, detailed research reports, thematic primers, one-pagers, risk analysis, growth triggers, bear case, capex tracker, walk the talk, and more for E2E Networks Limited and 4,900+ companies.
5-day free pass. No card required.