Anthropic's $321bn lease book spells opportunity for Indian IT

By Tom Whitfield

Anthropic reported $16.3 billion in revenue for the first half of calendar 2026, a 14-fold year-on-year jump, and still sits on roughly $321 billion in leases that have yet to commence. That gap between income and obligation is, according to a new Anand Rathi report, precisely where Indian IT services firms should be looking.

The brokerage argues that the financial risk of the AI boom is concentrating in the infrastructure layer, not the services layer. Companies building and operating computing capacity face rising capital intensity, heavy funding requirements and utilisation risk. Indian IT companies, by contrast, can ride the enterprise deployment cycle without writing those cheques.

"Indian IT offers relatively asset-light exposure to the subsequent enterprise deployment cycle, supported by net-cash balance sheets, low capex intensity & strong cash conversion," the report said.

What does the balance sheet asymmetry mean?

The numbers make the point bluntly. Anthropic's adjusted operating profit turned positive at $1.6 billion in the first half of 2026 — a genuine milestone. Against that sits a computing-commitment pile of some $321 billion in not-yet-started leases. Anand Rathi reads this as evidence that the capital-intensive part of the AI business stays with the infrastructure builders, while downstream service providers collect fees without replicating the investment.

The model provider's own spending pattern supports the thesis. Anthropic's sales, marketing and partnership expenses rose roughly 27 times year-on-year to $3.2 billion in the first half of 2026, driven largely by a $2.7 billion year-on-year increase in channel-partner fees. Somebody is being paid to move these models into enterprises. Increasingly, that somebody is a systems integrator.

"Underscoring the growing role of SIs in enterprise AI adoption, creating downstream opportunities for model-agnostic systems integrators," the report said.

Where is the enterprise demand coming from?

Adoption is measurable, not speculative. Anthropic counted approximately 11,900 customers generating more than $100,000 in annual revenue as of September 2026. Around 80 per cent of its revenue in the first half of the year was consumption-based — enterprises paying for what they actually use, which tends to favour deployment work over licensing deals.

The opportunity, in Anand Rathi's framing, sits with model-agnostic integrators: firms that help enterprises select, integrate and deploy AI models across existing systems, applications and workflows. Indian providers, with net-cash balance sheets and low capex intensity, fit that description better than most.

What could go wrong?

Two risks, both named. First, pricing. "Microsoft's workforce rationalisation signals AI-led pricing deflation," the report said — if AI compresses what clients will pay, service margins follow. Second, budget crowding-out: rising compute capex at client companies could constrain the discretionary IT spending that traditionally funds integration work.

The report's verdict is therefore conditional. AI adoption should create new revenue for Indian IT, but the pace and profitability of that growth depend on enterprise demand, the sector's ability to deliver AI integration services, and how technology budgets reshuffle as compute bills grow.

For CIOs watching from the buyer side, the practical question is simpler: if integrators are winning channel fees at 27-times growth rates, deployment demand is real. The report at least suggests the work exists — and that somebody else is holding the $321 billion lease.