Mega Networks: Mega Networks plans to make AI servers in Maharashtra
The factory will be in Maharashtra and the company plans to invest Rs 350-400 crore in two to three years, Amrish Pipada told ET.
“The initial investment for setting up the factory will be funded through internal accruals and debt of Rs 100-120 crore, with the remainder raised from long-term debt and private equity,” Pipada said.
Mega Networks is one of two Indian companies to be selected for the production-linked incentive (PLI) scheme 2.0 for IT hardware, under which the government provides an incentive of around 5% on net incremental sales of goods manufactured in India.
It has a technology partnership with US chipmaker Intel. The AI servers, launched in February, are optimised by Intel’s Habana Gaudi 1 and 2 series and 4th and 5th generation scalable processors, Pipada said. The modular servers can be used for generative AI, high-performance computing and in data centres.
“The funnel we have right now for AI servers is huge. If we are able to convert even 20%, our business will grow by 100-200%,” Pipada said.
Discover the stories of your interest
Mega Networks has identified potential orders of close to Rs 1,000 crore for the servers, with confirmed order bookings of Rs 150-200 crore at this time, he said.To support growth, Mega Networks plans to increase its employee count from 110 now to around 400 over the coming year, in roles including R&D, finance, marketing and sales, he added.
The company recorded revenue of Rs 300 crore in FY24, growing 30% year-on-year, Pipada said. It expects AI server manufacturing to significantly boost revenue, and is looking to hit Rs 1,000 crore in two-three years, Pipada said.
“The AI ecosystem in India is a growing market,” Pipada said, pointing to the government’s investment with the India AI Mission, state governments investing in data centres, as well as demand from major data centre players and conglomerates like Reliance and Tata that are looking to build large language models.
Given the huge demand, more players could enter the AI server space, Pipada said. “But this product requires a lot of support also.”
Government support is needed to make the ecosystem of parts available faster for manufacturers and also to get allocations from major global manufacturers which supply AI chips, microprocessors and other mission-critical parts, he said. Domestic design capabilities also need to be boosted.
Supply chain constraints pose a major challenge. “If you order today, deliveries take 30-50 weeks,” Pipada said. “The ecosystem of parts is yet to be built in India. That was a major concern we had in our meeting with the IT minister.”
Electronics parts for servers currently need to be imported, while mechanical requirements can be met locally, Pipada said. He expects that in the next three-five years, half of the electronics parts could be sourced locally as the domestic ecosystem grows.
