Indian startups being supplied last-gen, end-of-life GPUs in IndiaAI Mission: Report – Firstpost

Indian startups have raised issues over the standard of GPUs being supplied below the IndiaAI Mission, claiming that a number of the {hardware} included within the programme has already been marked as end-of-life (EOL) by producers.

As per a report by the Financial Instances, sure cloud service suppliers (CSPs) are supplying outdated NVIDIA A100 GPUs, regardless of the mannequin being declared EOL or Finish Of Life in February 2024.

The IndiaAI Mission, which goals to spice up the nation’s AI capabilities by offering inexpensive entry to high-performance GPUs, has listed 176 NVIDIA A100 GPUs among the many 18,693 models in its procurement plan.  

With entry to those GPUs set to be facilitated by way of a web based portal launching quickly, some bidders are questioning why older-generation GPUs are being included when they’re not in manufacturing or supported by the producer.

EOL GPUs appropriate for some duties, not all

NVIDIA’s official discontinuation of the A100 collection and labelling them as EOL implies that these GPUs will not obtain software program updates, safety patches, or technical assist. This has raised issues about their long-term viability, particularly for mission-critical AI purposes.

Nevertheless, some trade consultants instructed ET, that EOL standing doesn’t essentially render a GPU ineffective. Tarun Dua, CEO of E2E Networks, famous that whereas the A100 is not being bought by authentic gear producers (OEMs), it’s nonetheless extensively accessible by way of cloud suppliers and stays a strong choice for AI workloads. He identified that AI processing doesn’t at all times require the most recent {hardware} and that older GPUs can nonetheless ship aggressive efficiency for sure duties.

That mentioned, there are dangers concerned in counting on discontinued {hardware}. If efficiency points come up or a GPU malfunctions, there could also be no producer assist or replacements accessible. Whereas older GPUs may nonetheless be helpful for analysis and testing environments and are completely superb to be used by college students and researchers, they aren’t preferrred for production-level AI purposes the place 24/7 uptime is vital, and may even be a possible level of failure in mission-critical purposes.  

Value and authorities subsidy issues

Past technical points, startups are additionally elevating questions concerning the monetary logic behind the procurement. As per the ET report, some bidders argue that the Ministry of Electronics and Data Know-how (MeitY) is offering a bigger subsidy for the A100 in comparison with superior-performing GPUs.

For instance, the NVIDIA A100 40GB GPU prices Rs. 136 on-demand and Rs. 89 per 30 days for a reservation.  

MeitY is providing a subsidy of Rs. 54 per unit for the A100, whereas newer, extra highly effective GPUs are being subsidised at simply Rs. 28. Over longer-term reservations, the A100 is being subsidised at Rs. 34 per unit, whereas higher-performing fashions obtain solely Rs. 23.6, as per the ET report.  

Some bidders argue that this allocation ought to be re-evaluated to prioritise cost-effective and high-performance options for Indian AI startups.

Regardless of the issues, some trade voices consider that utilizing a mixture of older and newer GPUs may help steadiness value and efficiency, particularly in an trade going through world provide shortages and commerce restrictions.  

Whereas EOL GPUs can nonetheless be precious in particular situations, their position in a nationwide AI mission stays a subject of debate. Whether or not the IndiaAI Mission will rethink its GPU procurement technique stays to be seen, however the situation has actually sparked a dialog about balancing value, efficiency, and long-term sustainability in India’s AI ecosystem.