AI Capex Boom: Who Pays for India’s Compute Infrastructure?

Quick View

Current Context

India’s digital economy is being shaped by public digital rails, AI infrastructure, open networks and payment interoperability. ONDC’s official portal reports 616+ live cities and 7.64 lakh sellers or service providers, with the portal’s order statistic dated May 2025.

How It Works

• GPU clusters require power, cooling, networking and specialised operations
• demand can grow quickly but workload mix and price competition affect utilisation
• public subsidy can expand access while shifting part of investment risk to taxpayers

Economic Logic

The central question is who bears the capital, utilisation and obsolescence risk of building large-scale AI compute in India. Digital businesses often appear asset-light because customers see software rather than infrastructure. Economically, however, the model can carry heavy compute, data, distribution, compliance and switching costs.

The first channel is that gpu clusters require power, cooling, networking and specialised operations. This means a technical improvement is not automatically a financial improvement. The relevant unit must be tied to a transaction, task, customer or outcome for which someone is willing to pay.

The second channel is that demand can grow quickly but workload mix and price competition affect utilisation. Scale can reduce average cost, but it can also magnify concentration, outage and governance risk. The design should therefore include both normal operating economics and a stressed scenario.

The third channel is that public subsidy can expand access while shifting part of investment risk to taxpayers. For investors and managers, this shifts attention from headline adoption to durable gross margin, customer retention, data rights and control over distribution.

Digital economics should be analysed layer by layer. Infrastructure includes compute, power, storage and network. The model or software layer transforms inputs. The workflow layer determines whether the tool changes actual work. The distribution layer acquires and retains users. Governance covers privacy, security, accountability and legal rights. A weakness at any layer can absorb the value created elsewhere.

Many digital products have low marginal distribution cost but high fixed and semi-variable cost. Inference, support, fraud, refunds and compliance can rise with usage. A business should therefore calculate contribution per task or transaction rather than assuming that more users always improve economics.

Data is useful only when the business has lawful rights, adequate quality and a repeatable method for turning it into decisions. Data cleaning, consent, storage, security and deletion all cost money. A model that depends on unavailable or restricted data may have impressive technical tests but weak commercial durability.

Distribution is often the scarce asset. Platform rules, app stores, advertising auctions and network effects influence customer access and pricing. A technically strong product can still lose money if customer acquisition cost rises faster than gross profit or if one intermediary controls discovery.

Governance should be treated as an operating system rather than a final legal review. Access limits, logs, approvals, incident response and human accountability reduce expected loss. For high-value finance, identity or payment actions, a small amount of deliberate friction can be economically rational.

Finally, digital investment needs a staged evidence plan. Begin with a narrow use case, baseline cost and error rate, cap authority, measure realised outcomes, and expand only when the economics survive normal and stressed demand.

Calculation Framework

The formula is a decision aid rather than an accounting standard. Define every input consistently, use cash amounts where possible and run a downside case. A short payback can still be unattractive when the benefit is uncertain, while a longer payback may be acceptable when it removes a major operational risk.

Worked Example

Decision Scenarios

Metrics to Track

Cash Flow Lens

Translate the plan into actual collection and payment dates. Include deposits, taxes, implementation cost, financing, maintenance, refunds, penalties and contingency. An attractive margin can still create a funding crisis when cash arrives after unavoidable outflows.

Use incremental economics. Costs that continue without the decision are not incremental. New supervision, support, compliance, working capital and failure risk are incremental even when they do not appear in the vendor proposal or headline business case.

Risk Signals

90-Day Action Plan

1. Assign one owner to GPU utilisation and define a monthly threshold.
2. Create a baseline using at least three recent operating periods.
3. Model a downside case with slower collections, lower utilisation or higher failure cost.
4. Document authority, exception and escalation rules before scaling.
5. Review the decision after 30, 60 and 90 days using realised cash and operating data.

Evidence Checklist

• Source contracts, invoices and transaction-level records
• Bank statements, ageing reports and reconciliation support
• Operating logs, usage records and exception reports
• Approval trail, access register and management review notes
• Assumptions, calculation workbook and downside scenario

Finin2min Takeaway