Hiranandani Group-led Yotta, which has been the frontrunner in creating compute infrastructure in the country with its Nvidia portfolio of graphic processing units (GPUs) is looking to replicate the Jio model in the AI GPU space – high volumes at lower prices. Yotta co-founder and CEO Sunil Gupta in an interview with Jatin Grover, said the company’s AI platform services on top of GPUs will be driven by the India market and help reduce its dependency on a few global clients. He expects India’s GPU capacity to increase to 100,000 over the next few years. Excerpts:
Where does Yotta stand in terms of compute infra for the Indian market?
About a year back, GPUs (graphic processing units) in India were very few and they were not good enough for AI training or inferencing purposes, but only for gaming and content creation. From that point of view, Yotta took a good bet on the future. We have not only been able to serve India’s GPU and AI market, but also grow the market. Startups would not have thought of training a 10 billion parameter model in India, which they are doing now just because Yotta has been able to give them, let’s say 1,000 GPUs just for two weeks which otherwise would have cost them millions of dollars. With 16,000 GPUs, which Yotta placed the order for, it is possibly the ninth or tenth largest cluster in the world, barring GPUs procured by the big tech companies.
India accounts for 25% of Yotta’s capacity. Do you see the share growing?
In the foreseeable future, based on the trends of customer demands, Yotta will continue to have atleast 60-80% or bulk of its capacity getting into this wholesale type of customers, who want raw GPUs, mostly from outside India. But the exciting market, which will grow much bigger in future will be India. This is because inferencing and AI use cases have got no limitation, and the demand will be coming from Indian startups, IT companies who are actually creating use cases for end customers. We not only provide GPUs but also give platform-as-a-service on top of that.
Do you think offering low prices when investments are huge, is sustainable for Yotta?
The thought process is simple that Indian market always works on volumes. We have example of the Jio effect and Yotta is also thinking on the same lines. The company is making money even at these prices (around $2.5 per GPU power) and these are not marketing gimmicks. Instead of trying to sell to too few customers at high margins, we can possibly serve a much larger market, but having smaller margins, and spreading my cost over a much larger base of customers. Worldwide pricing for GPUs (as a service) have already come down to the levels, which Yotta is giving. In India, given its population, if we have to deploy AI on scale, there is a need to bring prices further down for affordability.
The company is into losses. What is the plan to achieve profits?
First, we are Ebitda positive. If we stop investing cash back into the business today, we will be profitable. However, we cannot stop investing. We expect to become profitable in Q4 of FY26.
Out of 16,000 GPUs procured, Yotta has deployed 4,000. What is the plan on further procurements?
First 4,000 GPUs went live in June. Next, 8,000 GPUs will go live between December and January. Remaining 4,000 may not remain so, it may become bigger because as we are now serving both the Indian and global market we may start ordering more also. We are looking forward to ramping up this capacity maybe to 50,000 or 100,000 basis the demand.
Where do you think India’s GPU capacity can increase in the near term?
The expectation is that GPUs will go 100,000 plus within two years. As of today, the Indian market is small and it is waiting for a trigger. There is a need to get good commercial applications and price needs to be reduced in the hands of end customers. If the US, for a much smaller population, today has got more than 300,000 GPUs, and given that India is much more digitally-enabled than the US, it is very logical that whatever GPUs the US is going to have for its citizens India is supposed to have 3-4 times of that.
What are the growth levers you see from the IndiaAI mission?
IndiaAI mission will be a game changer because it is not just one customer. It is the government’s push to the whole industry and to fund the end user. IndiaAI mission, giving a subsidy on top of Yotta’s already lower prices, will be the biggest trigger and make business models of many startups feasible.
Is Yotta also looking to increase data centres capacity in India?
Our outlook is to serve the market as a sovereign cloud operator and being a GPU and AI services operator, we are building larger data centre capacity for our own GPU requirements. Yotta is now scaling Mumbai campus to 1 GW capacity from 52 MW and Delhi campus will scale to 250 MW from 28.8 MW capacity.
Why is the company looking to list on Nasdaq and not in India?
US comprises around 80% of the AI market, if we don’t factor in China. This whole industry is too US centric and we wanted to be seen on a global scale. For us, it is a huge branding exercise. Many similar companies are listed on Nasdaq, so this makes it possible for Yotta to get right returns and right valuations.