E2E Networks Ltd Q1 FY2024 Concall Summary

Key Takeaways

• First public quarter, ambitious medium-term target: Management guided to 30–40% revenue growth over the next couple of years, while cautioning that a small base makes that growth inherently lumpy.
• Self-service pivot is complete and now a moat: With 80% of customers on the self-service public cloud, the painful transition that once caused customer losses now differentiates E2E from smaller domestic rivals and aligns its capabilities with global cloud platforms.
• AI/ML demand is the growth engine, but early: The AI/ML workload environment is nascent and picking up, driven by data localization needs; these workloads cost nearly 10x a typical CPU workload, making competitive pricing a material edge.
• CAPEX ambition is large but deliberately vague: Management intends to invest a couple of hundred crores over the next couple of years, mostly into GPUs, and wants to raise significantly more debt to fund it, but declined to pin down a precise timeline or capacity figure.
• What to watch next quarter: Whether the H100 deployment plan materializes as supply constraints ease, any quantification of GPU-driven revenue, and early traction signals from the TIR AI platform.

The AI/ML Demand Thesis

• Nascent but accelerating: Demand for AI/ML and deep learning workloads is significantly picking up, pulled by higher education, research, mature enterprises and innovative startups that need data localization and must comply with sectoral regulations.
• Cost advantage magnified: Running an AI/ML workload costs almost 10x a typical CPU workload, so E2E's competitive pricing, built on being small, nimble and focused, provides a sharper edge in this segment than in traditional compute.
• GPU and CPU are inseparable: Management refused to split revenue between GPU and CPU compute, arguing that any customer using a GPU also consumes CPU, storage and other cloud services, making a clean segregation impossible. The primary driver for new customers, however, is majorly AI/ML on Cloud GPU.
• TIR platform still in incubation: The advanced AI/ML workload environment has very few customers in production and has not yet become a major driver of numbers; management expects to go deeper on its progress over the next couple of quarters.
• H100 deployment imminent: A plan exists to deploy H100 capacity, with requirements expected to come online in the next couple of months as temporary supply chain issues resolve. Management declined to disclose capacity numbers until deployment is complete.

Dua: "So, we have existed since 2009 and last 10 years we have competed with the Hyperscalers and of course, we have the pricing advantage due to our thing small and nimble and being very focused on building whatever customers need than building everything under the sun that Hyperscalers typically do."

Competitive Positioning & Customer Dynamics

• Hyperscalers are the constant rival: For every deal, management sees Hyperscalers as the major competitor, with the customer typically working simultaneously with both E2E and the Hyperscaler.
• Self-service moat: The completed shift to a self-service model, now covering 80% of the customer base, positions the company apart from smaller domestic competition and closer to global cloud platforms in capabilities, after a transition that previously caused customer losses and revenue replacement challenges.
• Customer migration is not inevitable: Management pushed back on the idea that growing customers must migrate to a Hyperscaler, noting that E2E's compute capacity is of the order of 20x to 25x that of its largest customers, so it believes it can continue to service them as they scale.
• Churn is hard to measure, but not alarming: In a dynamic environment where the same customer may run multiple organizations and AI/ML workloads are project-driven, management finds churn very difficult to calculate and declined to confirm or deny any specific rate, while stating the target medium-sized customer does not show very significant churn.
• Overseas revenue is negligible: E2E has started serving customers in the US, UK and other geographies, but revenue from overseas customers is "not much as of now," and management declined to quantify it.

Dua: "I believe that the customers will grow big and potentially we will be able to continue to service them."

CAPEX & Capacity Build-Out

• Ambition framed, not sized: Over the next couple of years, management intends to invest a couple of hundred crores to ramp up capacity, covering GPU, CPU, storage, networking gear and other hardware, but declined to provide a specific figure or a GPU-only breakdown when pressed.
• Timeline deliberately loose: When an analyst tried to pin the "couple of hundred crores" to a two-to-three-year window, management corrected that it takes a couple of years but would not commit to exactly two to three years.
• Debt-fueled: Management definitely wants to raise a lot more debt than currently held to finance its CAPEX aspirations.
• Pace governed by absorption: The company wants to deploy as much additional capacity as it can generate, as much as the system allows without getting destabilized.
• Data center model is asset-light: E2E does not own physical data centers; it takes racks on co-location from much larger data center players, which keeps the model capital-efficient relative to owning real estate.

Dua: "So, as much as our system allows without getting destabilized, we would to make that much CAPEX."

Margins, Efficiency & Accounting

• Operating leverage improving: Enhanced margins were attributed to improving operating efficiency, with R&D and platform costs declining as a percentage of sales.
• Depreciation policy changed: The company shifted from the WDV method to SLM and revised asset life from three years to six years, in compliance with Schedule II of the Companies Act and reflecting the actual life of assets at six to seven years.
• Hiring to grow slower than revenue: Management will continue to hire across R&D, sales and operational support, but the growth in the base platform headcount will not be in exact proportion to revenue growth, implying further operating leverage.

Product & Platform Evolution

• Feature velocity: Recent launches include the TIR advanced AI/ML workload environment with security groups, lifecycle management for database-as-a-service, advanced load balancing, improvements to the cloud native platform, a simple queue service, and a replication feature for the object storage platform.
• Certifications expand the addressable market: Having basic ISO certifications increases the overall target market because some customers require them as a precondition.
• Long-term vision is infrastructure-plus: Over a five-to-seven-year horizon, management intends to remain nimble and fast-moving, invest early in broad technology trends, be an early mover, and continue servicing infrastructure-as-a-service needs while expanding the portfolio based on customer requests.

Guidance & Commitments

Key Metrics