Govt eyeing 25,000 more GPUs as AI Mission's fourth tender closes
The government expects to add close to 25,000 graphics processing units (GPUs) in the next few weeks to the existing commitments for 40,535 GPUs, a senior government official told ET, as the fourth round of GPU procurement is underway and is expected to close soon.
India is subsidising GPUs for researchers and startups, bringing the per-hour cost to under $1, the cheapest in the world.
Technical bid evaluation for the fourth round of IndiaAI Mission's GPU tender is ongoing. Out of the existing 14 empanelled cloud service companies providing GPUs, at least four confirmed having bid in the fourth round. One of them has offered to provide at least 17,000 Nvidia B300 liquid cooled GPUs, among the most sought-after AI hardware.
The four confirmed participants are Yotta Data Services, E2E Networks, NxtGen Datacenter & Cloud Technologies and Cyfuture India.
E2E is deploying 1,024 GPUs and NxtGen 4,096 GPUs, both Nvidia B200, in the fourth round. Cyfuture is offering 1,024 GPUs, but details of the model are not available.
Yotta will be deploying 8,192 Nvidia B200 GPUs that it had committed in the previous rounds.
Out of 40,535 GPU commitments so far, 22,787 GPUs (56.21%) have been allocated, according to the IndiaAI Compute Portal.
E2E's GPUs are expected to go live for end users this month, while for Yotta, it is likely in March and NxtGen may take a few more months. The timing is unclear on Cyfuture’s deployment.
The IndiaAI Mission’s decision to significantly increase common compute capacity is a powerful signal of India’s commitment to becoming a global AI leader, said E2E Networks chief revenue officer Kesava Reddy.
In parallel, E2E Networks is commissioning India’s first B200 mega cluster with Nvidia Blackwell GPUs, bringing next-generation AI training and inference capabilities to Indian enterprises and startups, he said.
Together, these efforts strengthen India’s sovereign AI infrastructure and reduce dependence on overseas hyperscalers, he added.
Anuj Bairathi, chief executive of Cyfuture India, said the India AI Mission has been quite aggressive in securing GPUs. “There is a great demand for GPUs coming from various users via India AI," Bairathi told ET.
AS Rajagopal, chief executive of NxtGen Datacenter & Cloud Technologies, confirmed the commitment to deploy 4,094 B200 GPUs.
Yotta did not respond to a request for comment.
One of the companies requesting anonymity said: "Only 15,114 GPUs are being used as of now. This is because there is a global issue of memory shortage, the price has gone up by 650%, and there is a global NAND (flash memory) shortage. Non-volatile memory express (NVMe) disk prices have increased by five times."
Among other empanelled companies, CMS, CtrlS, Ishan, Orient and Vensysco said they are not offering any new GPUs in the fourth round. Jio, Locuz, NTT-Neysa, Sify and Tata Communications did not respond to ET's request for comment.
India is subsidising GPUs for researchers and startups, bringing the per-hour cost to under $1, the cheapest in the world.
Technical bid evaluation for the fourth round of IndiaAI Mission's GPU tender is ongoing. Out of the existing 14 empanelled cloud service companies providing GPUs, at least four confirmed having bid in the fourth round. One of them has offered to provide at least 17,000 Nvidia B300 liquid cooled GPUs, among the most sought-after AI hardware.
The four confirmed participants are Yotta Data Services, E2E Networks, NxtGen Datacenter & Cloud Technologies and Cyfuture India.
E2E is deploying 1,024 GPUs and NxtGen 4,096 GPUs, both Nvidia B200, in the fourth round. Cyfuture is offering 1,024 GPUs, but details of the model are not available.
Yotta will be deploying 8,192 Nvidia B200 GPUs that it had committed in the previous rounds.
Out of 40,535 GPU commitments so far, 22,787 GPUs (56.21%) have been allocated, according to the IndiaAI Compute Portal.
E2E's GPUs are expected to go live for end users this month, while for Yotta, it is likely in March and NxtGen may take a few more months. The timing is unclear on Cyfuture’s deployment.
The IndiaAI Mission’s decision to significantly increase common compute capacity is a powerful signal of India’s commitment to becoming a global AI leader, said E2E Networks chief revenue officer Kesava Reddy.
In parallel, E2E Networks is commissioning India’s first B200 mega cluster with Nvidia Blackwell GPUs, bringing next-generation AI training and inference capabilities to Indian enterprises and startups, he said.
Together, these efforts strengthen India’s sovereign AI infrastructure and reduce dependence on overseas hyperscalers, he added.
Anuj Bairathi, chief executive of Cyfuture India, said the India AI Mission has been quite aggressive in securing GPUs. “There is a great demand for GPUs coming from various users via India AI," Bairathi told ET.
AS Rajagopal, chief executive of NxtGen Datacenter & Cloud Technologies, confirmed the commitment to deploy 4,094 B200 GPUs.
Yotta did not respond to a request for comment.
One of the companies requesting anonymity said: "Only 15,114 GPUs are being used as of now. This is because there is a global issue of memory shortage, the price has gone up by 650%, and there is a global NAND (flash memory) shortage. Non-volatile memory express (NVMe) disk prices have increased by five times."
Among other empanelled companies, CMS, CtrlS, Ishan, Orient and Vensysco said they are not offering any new GPUs in the fourth round. Jio, Locuz, NTT-Neysa, Sify and Tata Communications did not respond to ET's request for comment.
Next Story