Several companies involved in India’s AI Mission, including Yotta, have raised concerns regarding the Nvidia A100 Tensor Core GPU, which has reached its end-of-life (EOL) status. Nvidia officially announced the EOL in January 2024, confirming that it would discontinue all A100 products, including PCIe and SXM models.
Of the 18,000 GPUs of various models planned for procurement, 176 units are the discontinued Nvidia A100 GPUs. Some bidders in the tender process have accused cloud service providers (CSPs) of attempting to offload obsolete technology.
“Cloud service providers are trying to palm off EOL GPUs. IndiaAI should bring in experts to evaluate the bid,” one bidder told The Economic Times.
A section of bidders has also challenged the rationale behind being required to procure outdated GPUs while still being expected to match the L1 price — the lowest quoted bid in the tendering process. They argue that such a move could compromise the quality and longevity of AI infrastructure.
When a product reaches EOL status, it signifies that the original equipment manufacturer (OEM) has ceased its production and sale. While CSPs may continue selling such products, OEM support — such as technical assistance, software updates, and hardware replacements — could be significantly limited or unavailable.
Sunil Gupta, CEO of Yotta Data Services, has highlighted potential risks associated with using discontinued GPUs. While they may still function, lack of technical support and unavailability of replacements in case of failure could pose significant challenges, particularly in high-performance computing environments.
However, some CEOs hold quite the opposite views. Anuj Bairathi, CEO of Cyfuture, notes that using legacy GPUs can be a cost-effective solution as the GPUs are working fine.
India is almost on the verge of losing the AI race. Krutrim AI, developed by OLA and considered India’s first AI tool, is nowhere near DeepSeek, the AI developed by China.
Given that India is looking to develop an in-house AI within ten months and the government has approved 18 proposals for this purpose, the use of EOL GPUs has certainly raised questions regarding the direction of the country’s AI program.
In the News: Two Estonians found guilty of $577 million crypto ponzi schemeIn the News:
