India's Neoclouds Face Twin Tests as US AI Selloff Deepens

Thin anchor demand at home

US neoclouds lean on large customers such as Meta and OpenAI, which gives lenders some visibility. Indian players have no comparable base. International customers account for 75-80% of Yotta's customers, partly because they pay more. At home, the IndiaAI Mission had onboarded over 45,000 GPUs by June 2026 and approved 237 projects with 93.18 lakh GPU hours by August. But sanctioned hours do not guarantee usage: the mission used less than half of its ₹10,000 Cr FY26 outlay, which affected hardware procurement and deployment timelines, as Neysa noted.

The market also looks mismatched. Some GPUs sit underused while startups and researchers struggle with fragmented capacity, allocation delays and unsuitable hardware. Juspay, which buys GPUs with its own cash, has been told it will wait three to six months for delivery, according to co-founder and COO Sheetal Lalwani.

Leverage and exposure

Yotta plans to spend over $7 Bn on AI infrastructure by the end of FY27 and is targeting a $1.5 Bn IPO, partly to repay debt. ESDS went to market after the cost of the GPU servers it planned to buy rose from ₹25 Lakh to ₹1.8 Cr. E2E Networks reported a 411% year-on-year jump in Q2 revenue. NxtGen, NeevCloud and Neysa are also building capacity.

CoreWeave is entering India as the sole tenant of a planned 240 MW AdaniConneX campus, so pressure on its financing could affect that project. The open question is whether Indian neoclouds can secure enough institutional funding and long-term customer commitments to ride out a global correction, or whether they will stay dependent on overseas demand.