E2E Networks Ltd Q2 FY2025 Concall Summary

Key Takeaways

• Revenue doubled, GPU demand robust: Total revenue grew ~120% YoY to ₹48.4 crore, driven by strong cloud GPU demand, though utilization dipped temporarily with new capacity additions.
• ₹405.6 crore fundraise closed: The capital is earmarked primarily for GPU infrastructure expansion, with 75% directed to capacity and 25% to engineering and corporate purposes.
• Margins expanded sharply: EBITDA margin reached 66.1%, up 1,440 bps YoY, and management expects further expansion as volume scales.
• Enterprise adoption is the next leg: AI startups lead today, but enterprises are beginning to allocate budgets for AI/GenAI proof-of-concepts, signaling a multi-year production deployment opportunity.
• What to watch next quarter: Utilization rates on the newly deployed 512 GPUs (256 in September, 256 in October), any quantification of the partner program or GCC pipeline, and the pace of the next hardware purchase cycle.

GPU Demand & Competitive Positioning

• Demand remains robust: Cloud GPU demand continues to be fairly strong, with domestic and international demand split approximately 65/35 over the last two quarters. International interest comes mainly from startups and higher education/research institutions.
• Utilization dip is mechanical: The last 256 H100 GPUs deployed in October are very recent and mostly unutilized from a revenue perspective as of the call date, so the lower utilization rate reflects new capacity that has not yet been absorbed.
• Cloud repatriation tailwind: Cloud GPU has accelerated repatriation from hyperscalers, as customers find hyperscaler costs rise rapidly while E2E offers predictable pricing with fewer variables.
• Middle-ground positioning: E2E positions itself between on-prem (low cost, very high capability required) and hyperscalers (low capability required, very high cost), a niche management believes is defensible.
• Software is the moat: Competitive advantage comes not just from sticker price but from platform software that prevents GPU idleness, enables rapid cluster deployment, and reduces idle time. The TIR platform, released in the last year, delivers software matching global majors in AI/ML/GenAI computing.
• Market size absorbs competition: No major impact of competition on yields has been seen yet, and management believes the market is large enough for 10 successful players in India alone, each with its own niche.

Tarun Dua: "Competitive advantage just doesn't come from like being able to offer at a lower per hour sticker price, but also the efficiency of the efficiency and the usable-ness of the software that you build on the platform."

Capacity Expansion & Fund Deployment

• 256 H100 GPUs secured in October: This adds to the existing GPU fleet and brings the GPU-CPU mix to approximately 90-10, which management considers the steady state.
• ₹405.6 crore fundraise closed: 75% of proceeds will go toward augmenting infrastructure capacity, majorly GPUs; the remaining 25% sits in the general corporate corpus with optionality for engineering and technology capabilities on the software side.
• Hardware purchase cycles: Each cycle involves approximately 255 GPUs plus associated storage, memory, peripherals, and some CPU compute, costing between ₹90-125 crore. The company tries to consume the majority of capacity and see visibility for the next set before placing another order.
• Capex ambition over two years: Management stated the company should definitely do more than ₹800 crore of capex over the next couple of years, though no annual breakdown was provided.
• Supply chain improving: India is not a top buyer of GPUs and is slightly behind in queue for first versions, which go to hyperscalers buying 100,000+ GPUs per year. Once a new SKU stabilizes, supply is very fast, and things have improved compared to a year ago or six months ago in terms of delivery speed.
• Blackwell GPU intent: For the next GPU version, the company intends to place orders much in advance to be first in queue as soon as new supply comes in, with trillion-parameter models expected to be a good match for Blackwell.

Tarun Dua: "India is definitely not a top buyer of GPUs... we do see kind of like India slightly behind in the queue from people who are able to kind of like get the very first versions of the GPUs that come out."

Enterprise Adoption & Customer Pipeline

• Startups lead, enterprises follow: AI startups are leading adoption today, but enterprises are increasingly allocating budgets to identify core problems solvable by AI/GenAI. Small language models running on previous-gen GPUs are seeing production deployments, indicating a multi-year push toward more cloud GPU usage in production.
• Partner program still nascent: Revenue from the partner program is still a fairly small number compared to overall revenue, though the company is working on building partnerships across the ecosystem and expects more revenue in the future.
• Dell MOU signed: Dell is a vendor supplying GPU servers and other hardware, and both companies are exploring jointly targeting solutions for customers from a sales perspective. No specific terms, timeline, or expected impact were disclosed.
• People+ai collaboration: The company is working with People+ai and other organizations to explore what makes sense for cloud GPU players and CPU cloud players in India to do jointly. This is still in very early days with no current revenue impact.
• GCC pipeline unquantified: When asked about Global Capability Center trends and demand, management acknowledged a few in the pipeline but could not quantify the opportunity.
• Government and enterprise push: Key strategic initiatives include increasing diversity of the customer base by incrementally adding government and enterprise customers, and establishing at least one new major cloud zone south of India including GPUs.

Financial Model & Margins

• EBITDA margin at 66.1%: Up 1,440 bps YoY, with management hopeful of further increasing EBITDA margin in coming years as volume grows and operating leverage kicks in.
• Asset turn at ~0.5 for incremental deployment: Management expects this to grow as software continues to be built, but the pace is hard to predict. The company recommends looking back at historical trends rather than projecting forward.
• GPU pricing lifecycle: For any GPU SKU, demand is initially low as people figure out utilization with newer software libraries, then demand and supply increase, eventually reaching a steady-state price point.
• No balance-sheet leverage target: When asked about comfortable leverage levels, management deflected, saying it depends on MRR growth and declined to set a number.

Tarun Dua: "Currently, it's like broadly about 0.5 for incremental deployment. So, like we of course like as we continue to build our software like we expect it to grow. Now, what phase it grows is very hard to predict. So, which is why we say that, like, yes, like generally like let's look at what is currently happening. And when it changes like, for example, like when it changes then we know that it has changed."

Management Philosophy & Disclosure Gaps

• Reactive, not predictive: Management repeatedly declined to provide forward-looking numbers on capex, gross block, asset turn, or margin trajectory, emphasizing a philosophy of reacting nimbly to market conditions rather than committing to hardbound plans.
• Utilization data not provided: When asked for the utilization rate of the 700 GPUs deployed in September and 256 in October, management said they did not have it handy.
• Internal KPIs are strategic, not numeric: When asked about primary internal KPIs, management answered with strategic initiatives (new cloud zone, customer diversity, engineering additions) rather than specific measurable metrics.
• Three-year perspective declined: An analyst asked for a three-year perspective on asset creation and what asset turn and margin capital would be invested at; management declined, saying it is very hard to predict and recommended looking back into the past.

Tarun Dua: "We always recommend that in our cloud business like look back rather than look ahead."

Guidance & Commitments

Key Metrics