India is subsidising access to GPUs for researchers and startups, bringing the per-hour cost to under $1, according to officials. The government expects to add nearly 25,000 graphics processing units (GPUs) in the coming weeks to the existing commitments of 40,535 units under the IndiaAI Mission, a senior government official told The Economic Times. The fourth round of GPU procurement is underway and the technical bid evaluation is in progress.
India is subsidising access to GPUs for researchers and startups, bringing the per-hour cost to under $1, according to officials. Of the 40,535 GPUs committed so far, 22,787 units, or 56.21 per cent, have been allocated through the IndiaAI Compute Portal.
At least four of the 14 empanelled cloud service providers have confirmed participation in the fourth round: Yotta Data Services, E2E Networks, NxtGen Datacenter & Cloud Technologies and Cyfuture India. One of the bidders has offered to supply at least 17,000 Nvidia B300 liquid-cooled GPUs.
Deployment timelines and vendor commitments
E2E Networks is deploying 1,024 Nvidia B200 GPUs, while NxtGen Datacenter & Cloud Technologies has committed 4,096 B200 GPUs. Cyfuture India is offering 1,024 GPUs, although the model details were not disclosed.
Yotta Data Services will deploy 8,192 Nvidia B200 GPUs that it had committed in earlier rounds.
E2E’s GPUs are expected to go live for users later this month. Yotta’s deployment is likely in March, while NxtGen’s rollout may take several more months. Cyfuture’s deployment timeline remains unclear.
Among other empanelled providers, CMS, CtrlS, Ishan, Orient and Vensysco said they are not offering additional GPUs in the fourth round. Jio, Locuz, NTT-Neysa, Sify and Tata Communications did not respond to queries from ET.
Supply constraints and demand pressures
One company, speaking on condition of anonymity, said only 15,114 GPUs are currently in use. “This is because there is a global issue of memory shortage. Prices have gone up by 650 per cent and there is a global NAND shortage. NVMe disk prices have increased fivefold,” the executive said.E2E Networks chief revenue officer Kesava Reddy said the expansion of common compute capacity signals the government’s intent to scale domestic AI infrastructure. He added that E2E is commissioning India’s first B200 mega cluster using Nvidia Blackwell GPUs to support AI training and inference.
Cyfuture India chief executive Anuj Bairathi said demand for GPUs through the IndiaAI platform remains strong. “There is a great demand for GPUs coming from various users via IndiaAI,” he said.
AS Rajagopal, chief executive of NxtGen Datacenter & Cloud Technologies, confirmed the company’s commitment to deploy 4,094 B200 GPUs.
Leave a Reply