E2E Networks Ltd Q2 FY2027 Concall Summary

Key Takeaways

• Revenue inflected, driven by Blackwell: Revenue hit ₹223.8 crore, up 411% YoY and 43% QoQ, as the new Blackwell GPU cluster contributed, utilization improved, and average realizations rose.
• A ₹1,000 crores anchor deal locked in: A multiyear contract with a Sovereign AI company converted the flagship Blackwell capacity from pay-as-you-go to a commitment through June 2029, giving long-term revenue visibility.
• Margins surged to 76.6% EBITDA: EBITDA margin expanded 141 bps QoQ on the same Blackwell and utilization drivers, and management views E2E as a 70% plus EBITDA margin business for the medium and long term.
• Platform strategy is the bet, not just GPU resale: Management frames GPUs as a means to an end, with the real long-term value in building globally adoptable technology platforms, though the jury on success is still out.
• What to watch next quarter: Deployment of the next lot of B200 GPUs, the official launch of inference products Blaze, Endpoint and Jarvis Lab, and any further conversion of uncontracted capacity into longer-term commitments.

The Blackwell Inflection & Sovereign AI Deal

• Blackwell drove the quarter: The deployment of the Blackwell GPU cluster was the primary driver of the 43% QoQ revenue jump, better utilization, and higher average realization across the GPU fleet.
• Anchor tenant secured: A nearly ₹1,000 crores deal with a Sovereign AI company converted the existing B200 cluster from a pay-as-you-go model to a contract lasting through June 2029, improving long-term revenue visibility.
• Next deployment imminent: The next lot of B200 GPUs, already announced, is expected to go live this quarter (Q3 FY27), with 1,024 plus B200 GPUs targeted for deployment this quarter.
• Beyond B200 is unscripted: Nothing has been decided yet around the next lot of B200 beyond the current deployment, and no contracts are set; plans for B300 or Vera Rubin will be shared only as they fructify.
• Capex philosophy is aggressive but opaque: Management will continue to be both aggressive and prudent in capex deployment but gives no speculative guidance, informing only when deployment plans are sufficiently solidified.

Tarun Dua: "This is still day zero of AI for us, as far as we can see."

Platform & Product Strategy

• GPUs are a means, not the end: Management explicitly rejects viewing GPUs as rental assets, framing them instead as the substrate for building technology platforms that can be adopted domestically and globally.
• Platform adoption is already the norm: The majority of GPUs today run on top of E2E's platforms, with very little bare metal sold, indicating the transition from raw compute to platform is already underway.
• Inference products near launch: Inference-based products Blaze, Endpoint and Jarvis Lab have been soft-released, with an official launch "a few weeks or so away," though substantial adoption will be discussed only when it shows up in volume.
• Platform investment consumes own capacity: A big part of the platform build-out requires using E2E's own GPUs for product development rather than selling all capacity, which management frames as an investment into the future.
• Success is a medium-term question: Management asks that the jury on platform success be held till the medium term, acknowledging that success shows up over time and the proof of the pudding lies in eating.

Tarun Dua: "The long-term strategy of the company is not dependent on whether we can make a few cents more or few percentage points more from same set of GPUs. The long-term strategy is to make money by building more technology, and then the resulting growth is a result of our advances in technology and customer adoption of our technology."

Margin Structure & Profitability

• EBITDA margin hit 76.6%: The 141 bps QoQ expansion was driven by Blackwell GPU contribution, better utilization, and higher average realization across the GPU fleet.
• Gross margin reached 88%: Management declined to attribute this to a single factor, noting many moving parts including capex timing and capacity built in advance that is not yet live, and gave no specific gross margin target.
• Profitability flipped decisively: H1 FY27 delivered PBT of ₹155.8 crore and PAT of ₹116.6 crore, compared to losses in H1 FY26 (PBT loss of ₹22.2 crore, PAT loss of ₹16.3 crore).
• 70% plus EBITDA is the medium-term anchor: Management views E2E as a 70% plus EBITDA margin business for the medium and long term, with no change in that overall target.
• Earnings upside from uncontracted capacity: For capacity that is currently uncontracted or on short-term contracts, management expects earnings to grow from the same assets going forward as those assets get better contracted or utilized.

Competitive Landscape & Market Opportunity

• Vast blue ocean, not a red ocean: Management sees the AI infrastructure market as large enough for 150 to 250-plus players globally, with competition not a concern for the next couple of years.
• International revenue is material: Roughly 20% to 25% of revenue for the quarter came from international customers, indicating the platform is already finding adoption beyond India.
• Technology strategy is flexible by design: Management's stated approach is to keep building technology people want to adopt, keep pushing efficiency to lower total cost of ownership, and remain agile enough to make whatever works, work.
• Partnership openness is absolute: E2E is open to all possibilities of working with all partners in the ecosystem, with a "never say never" stance on funding, distribution, or any arrangement, though specific partnerships like using L&T's idle capacity "happen when they happen."

Tarun Dua: "It's a vast blue ocean. There is no red ocean today that we need to worry about competition today... For next couple of years, probably this is not even a question."

What Management Will Not Disclose

• No forward-looking predictions, period: Management explicitly declines to give any forward-looking guidance on capex, GPU deployments, MRR, or exit run-rates, asking investors to look back at what has been done rather than forward.
• No point-in-time operating metrics: Utilization levels, contracted revenue percentage, and inference versus training revenue breakdowns are all withheld, with management arguing the company is too small for point-in-time disclosures that would change dramatically.
• No specific MRR contribution from the anchor deal: When asked how much of current MRR the ₹1,000 crores contract contributes, management deflected, suggesting analysts calculate "close enough by looking at the ₹1,000 crores number and the time period."
• No GPU allotment visibility: On allotment visibility for B300s or Vera Rubin in FY28 and FY29, management declined to predict, stating "it is hard to predict the future" and they will take the future as it comes.
• No base case for annual GPU additions: An analyst's suggestion of 2,000 GPUs per annum as a base case for the next three years was rejected, with management reiterating they never make forward-looking predictions.

Tarun Dua: "I think somewhere in a very large world, in a very large TAM, you have to be having the humility of saying that we don't know what is going to work, but be agile enough to say that, Yes, let's make it work together, whatever is going to work."

Guidance & Commitments

Key Metrics