India's AI Mission Advances: GPU Bidders Set to Receive AI Workloads

AI Mission GPU L1 bidders set to get workloads assigned by April second week

Reliance Jio, E2E, NxtGen, Locuz, and CtrlS, as L1 bidders in IndiaAI Mission's GPU tender, will receive AI workloads. They'll integrate with the AI Compute portal by April 14, with further evaluations and fresh applications in April.Companies such as Reliance Jio, E2E, NxtGen, Locuz and CtrlS, which emerged as the lowest price (L1) bidders in the IndiaAI Mission tender for sourcing graphics processing units (GPU), will get artificial intelligence (AI) workloads assigned by the second week of April, a senior government official told ET. Ten technically qualified empanelled bidders have offered 14,517 GPUs at L1 rates against the tender for 10,000 GPUs. IndiaAI Mission chief executive Abhishek Singh told ET that some of the L1 bidders have already integrated with the AI Compute portal launched on March 6. IndiaAI would renew the empanelment every quarter inviting fresh proposals for discovering any revised rates, the Request for Empanelment (RFE) document dated February 21 had said. Empanelled agencies can submit a revised financial proposal which can be the same or lower than the existing L1 rates. The last date for filing the next round of applications under this continuous empanelment process is April 30, with evaluations set to begin next month. A pre-bid meeting for the second round of the GPU tender held on March 26 had received very good response, Singh said. "IndiaAI is expecting a lot of companies which missed out in the first round to come in. None of the current L1 bidders offer Nvidia's B200 GPUs right now. The Blackwells (Nvidia's B200 GPUs) may come in," Singh said. "Our portal integration will be done by April 14," said AS Rajgopal, chief executive of NxtGen Datacenter & Cloud Technologies, adding that the company will be offering H200 GPUs along with other GPU models. Sources from E2E Networks said it would take at least a month to assess the market uptake of a particular GPU model. Yotta Data Services, which has offered as many as 9,216 GPUs but has not emerged as an L1 bidder for any of the 30 GPU models, may bid for more models in coming quarters, chief executive Sunil Gupta had told ET in February. He said on Tuesday: "Once L1 bidders exhaust their capacities as committed to India AI, the next AI workloads shall be assigned to L2 bidder (who matched the L1 rates and is empanelled). This is as per RFE guidelines and as confirmed by the India AI Mission." A large number of startups that have submitted proposals to develop foundation AI models have sought a wide range of GPUs along with funding support, the government informed the Lok Sabha last month. Umesh Shah, director, Orient Technologies, one of the ten technically qualified bidders that had proposed to offer 1,000 GPUs, told ET: "Given the rapidly growing demand for AI-driven solutions, we anticipate early indicators of market uptake within the next few months." Adoption will depend on factors such as ease of access, cost-effectiveness, and the ability of enterprises to integrate these resources into their AI workloads, he said. "A reasonable timeframe to assess broader industry impact would be within two quarters, as businesses experiment, benchmark performance, and scale their AI initiatives," he said. Anuj Bairathi, chief executive of Cyfuture India that did not qualify technically in the first round of the GPU tender, said he had raised objections with respect to concentrating on one OEM, Nvidia, when the market has a lot of options including ASIC-based Sambanova and Cerebras which Cyfuture is considering offering in the next round. ASIC stands for application-specific integrated circuit, a custom-designed integrated circuit tailored for a specific task or application, unlike general-purpose chips. The company has also raised a question about pricing wars. Now that the L1 pricing is public, any new player may offer cheaper prices than existing players. "Once the new player becomes L1, the capacity of this new vendor will be consumed first, sparking price wars which is not good for the industry. We have asked IndiaAI to find some solution for this problem," he said. Nvidia H100 GPUs comprise 69% of the total GPUs offered. For now, in H100, E2E has emerged as the L1 bidder and Yotta the L2 bidder. So, as soon as E2E exhausts their 716 committed H100 GPUs, Yotta gets to offer its 8,192 GPUs. The IndiaAI Application Development Initiative has received as many as 900 applications for support for developing AI solutions. Of these, 30 AI solutions have been shortlisted for the next stage in three stages of maturity: idea, prototype and existing solutions. Teams shortlisted in the initial stage were awarded Rs 5 lakh each on March 6 to develop their solutions further. Some teams received Rs 25 lakh to further hone their prototypes. The winners of this stage will get the chance to deploy their solutions in the government and receive funding of up to Rs 1 crore. "We are working with the concerned departments to scale them up. Ultimately, they will have to adopt it, and the solution will be further developed, deployed and taken to maturity. And those solutions which reach that level, will get a reward of Rs 1 crore. It will take three to six months for this process," Singh said.

Reliance Jio, Other GPU L1 Bidders In IndiaAI Mission To Get Workloads Soon Reliance Jio, Other GPU L1 Bidders In IndiaAI Mission To Get Workloads Soon

The IndiaAI mission covers initiatives like IndiaAI Compute Capacity, IndiaAI Innovation Centre (IAIC), IndiaAI Datasets Platform among others With the Centre stepping up its efforts to develop public AI cloud infrastructure under the IndiaAI Mission, companies such as Reliance Jio and CtrlS, which have emerged as as the lowest price (L1) bidders in the tender for sourcing graphics processing units (GPUs), will reportedly get AI workloads by the second week of April. The ten tech companies that include E2E, NxtGen and Locuz other than the aforementioned companies have offered 14,517 GPUs at L1 rates against the tender for 10,000 GPUs. GPU is a specialised chip that renders graphics and images by performing rapid mathematical calculations that are needed in processing AI. An ET report, citing IndiaAI Mission CEO Abhishek Singh, said that some of the L1 bidders have already integrated with the AI Compute portal launched last month. The AI Compute portal is aimed at providing researchers, startups, and government agencies access to high-powered computing resources such as advanced GPUs for AI applications. The report also mentioned that the pre-bid meeting for the second round for GPU tender held last month got a good response and IndiaAI is expecting a lot of companies in the upcoming round. Singh told ET that none of the L1 bidders have offered Nvidia's B200 GPUs which are part of Nvidia's Blackwell architecture but it may come in the second round. It is pertinent to note that the Ministry of Electronics and Information Technology (MeitY) launched a continuous empanelment process for AI cloud service providers that allows companies to apply on an ongoing basis to supply GPUs, MoneyControl reported in February. "IndiaAI would renew the empanelment every quarter inviting fresh proposals from the empaneled agencies for discovering any revised rates. Empaneled agencies shall submit a revised financial proposal which can be same or lower than the existing L1 rates," the Request for Empanelment (RFE) document said. Inside The INR 10,372 Cr IndiaAI Mission In March last year, the Union Cabinet approved the IndiaAI Mission with an allocation of INR 10,372 Cr in the next five years, aiming to foster innovation in the homegrown AI ecosystem and implementing the Mission's vision via a public-private partnership (PPP) model. It covers initiatives such as IndiaAI Compute Capacity, IndiaAI Innovation Centre (IAIC), IndiaAI Datasets Platform, IndiaAI Application Development Initiative, IndiaAI FutureSkills, IndiaAI Startup Financing, and Safe & Trusted AI. IndiaAI Compute Capacity envisions building a cutting-edge, scalable AI computing infrastructure by deploying more than 10,000 GPUs through PPP model. With IndiaAI Innovation Centre (IAIC), the Centre is looking to create a leading academic institution to retain top research talent, develop and deploy indigenous large multimodal models (LMMs) and domain-specific models While IndiaAI Datasets platform looks into accessibility, quality, and utility of public sector datasets, the IndiaAI Future Skills program focuses on expanding the accessibility of graduate and post-graduate AI programs as well establishing AI Labs to impart foundational-level courses Government's Focus On GPUs And AI Infra The Centre is looking to build affordable AI infrastructure to support startups, academics, and research agencies. "The government is actively shaping an AI ecosystem where computing power, GPUs, and research opportunities are accessible at an affordable cost," said MeitY in a statement last year. Hailing IndiaAI's vision, the ministry said that the key focus of this mission is the development of a high-end common computing facility equipped with 18,693 GPUs. The ministry underlined that this capacity is nearly nine times that of the open-source AI model DeepSeek and about two-thirds of what ChatGPT operates on. Under the mission, the government also eased the eligibility criteria for the procurement of GPUs under the IndiaAI mission to allow more participation of smaller companies and startups. It was reported in January that MeitY received offers for 20,000 GPUs which was twice of what was initially proposed for empanelling companies. After that, the Centre selected ten tech giants for the final bidding process to procure 10,000 GPUs including Jio Platforms and Tata Communications. Other companies bidding for the tender are CMS Computers India Pvt Ltd, Ctrls Datacenters Ltd, E2E Networks Limited, Locuz Enterprise Solutions Ltd, Orient Technologies Ltd, NxtGen Datacenter and Cloud Technologies Private Ltd, Vensysco Technologies Limited and Yotta Data Services Limited. Besides, the IndiaAI Mission last month received around 120 applications to build AI foundation models in the second round of bidding, which closed on March 15.