Lenovo to design, manufacture, and export AI servers from India

Las Vegas: Global technology major Lenovo is planning to transform India into a key export hub for its infrastructure business, with plans to design and manufacture artificial intelligence (AI) servers in the country for global markets, a top company executive has said.

Speaking to PTI on the sidelines of CES 2026, Scott Tease, VP and General Manager, Infrastructure Solutions Group at Lenovo, said the company will utilise its Bengaluru development lab to design AI server systems that will subsequently be manufactured at its Pondicherry facility for both domestic consumption and export.

"We are going to be designing a lot of our one- and two-socket systems... think of those as the workhorses of AI in the future. We are going to be designing them in India. Once we have designed and engineered them, we are going to be manufacturing them there as well. It is going to be an important part of Lenovo's value chain...

"...our initial focus in India is India for India, but given geographic proximity, the quality of the workforce, the fact that we've already had such great success building phones and PCs in India, there is no reason at all that the future is going to hold us building servers in India for the rest of the world. Absolutely no reason whatsoever," Tease said.

Lenovo India is among the companies that have been selected for the Rs 17,000-crore IT hardware production-linked incentive scheme.

With the high cost of massive data centres often proving prohibitive for Micro, Small and Medium Enterprises (MSMEs), Tease advocated for a hybrid AI model. He explained that Indian businesses do not necessarily need heavy investments to incorporate AI in their operations.

"Building a model might require something heavy..., but we can outsource that to cloud providers or GPU-as-a-service providers in India… who can help build the model," he said.

Once the model is built, Tease explained, running it (inferencing) requires much lighter hardware, often as small as a laptop or edge device.

"It makes it very easy for them (MSMEs) to adopt real, powerful AI... It is not power-intensive. It is not expensive," he noted.

On accessibility, he noted that the common language for programming AI in the future is likely to be natural language, such as English, which democratises access to technology.

Commenting on the regulatory landscape, the Lenovo executive praised the Indian government's approach to "Sovereign AI" -- the concept of a nation building its own domestic compute capacity.

"The domestic consumption in India is so large that even if you just tackle that alone, we're going to see massive, massive growth there. I visited several GPU-as-a-service and hosted sites that are doing AI, both for local India and for international cloud providers as well.

"You've got people that are really interested in building out sovereign capabilities in the country. You've got a very AI-friendly, AI-centric government, and that's going to be a big, big help for that kind of growth in the future," Tease said.

Addressing the environmental concerns surrounding the massive energy consumption of AI data centres, Tease highlighted that the industry must move away from traditional air cooling.

He pointed out that in typical data centres, air conditioning can add 40 per cent more to costs, adding that Lenovo is pushing for liquid cooling technologies (Lenovo Neptune), which reduce energy consumption by roughly 40 per cent and allow waste heat to be recycled for other purposes.

Asked about the key differentiator for nations by 2030, Tease asserted that workforce readiness, rather than access to silicon or energy, would determine the winners of the AI era.

"It is access to talent, for sure... never before have we seen the possibility for countries or workforces to catch up to more mature workforces, as we see right now with AI… the winners are going to be those regions, those companies, those governments that help their people embrace AI as part of what they do," he said.