MC EXCLUSIVE Micromax-Phison JV Mi-Phi eyes storage chip exports, govt incentive schemes

While production of SSDs and custom NAND storage chips is underway, Mi-Phi is simultaneously looking for global partners by ensuring its products are ready for commercial deployment meeting international standards.

“There are large opportunities where we are working on the qualifications…Of course, there will definitely be partners, because I don't believe that you can satisfy everyone directly,” Balakrishnan said.

Tapping India’s semiconductor push

Mi-Phi is eyeing incentives under the upcoming design-linked incentive (DLI) scheme, which is expected to have a broader scope under the revised India Semiconductor Mission 2.0.

“The current DLI does not have pure play memory as a component... but the new DLI that is going to come up, we will have a space,” Balakrishnan said. “We are in talks with a few government agencies that we are working very closely with.”

Mi-Phi also plans to apply for benefits under the electronics component-manufacturing scheme, as it looks to localise critical parts of the memory-chip supply chain.

Filling India’s memory design gap

Mi-Phi’s goal is to bring memory design and development to India, which has, so far, been limited to import-and-assemble models for storage products.

“When we started Mi-Phi, it was to address one unique gap in the Indian subcontinent… Memory design never happened in India. It was always brought in as a CKD or SKD format, assembled, and delivered to customers. That needed to change,” Balakrishnan said.

The company’s focus now spans consumer, enterprise, AI and embedded storage segments, with its R&D team building solutions tailored for Indian and global markets.

Mi-Phi aims to slash GPU costs by up to 90 percent by delivering the lowest per-token AI compute cost globally, the CEO said. Its memory extension solution expands GPU RAM onto dedicated memory devices, removing bottlenecks and enabling large AI models to run with far fewer GPUs. This not only cuts server and energy costs but also boosts data centre efficiency and makes high-performance AI more accessible.

“A typical 7TB model might currently require around 20 GPUs of a specific class. With our solution, the same workload can be run using just 7 or 8 GPUs, without compromising performance. This reduction translates into fewer servers, lower hardware costs, reduced power consumption, and overall data centre efficiency. This technology doesn’t just save cost, it democratizes AI,” he added.

Traditionally, to deal with memory and compute constraints, AI workloads are quantized — essentially reducing model precision to fit within GPU limits. “But with our approach, you can run full-precision models directly while using just one-third of the GPU resources. That’s a game-changer for the industry. It reshapes the entire CapEx and OpEx equation for enterprises and AI infrastructure providers,” Balakrishnan claimed.

Enterprise and hiring plans

In India, Mi-Phi is targeting enterprise-grade SSDs, working with OEMs who manufacture servers, laptops, and storage solutions. It is also running POCs and tech qualifications with large cloud service providers and enterprises, while developing a customer base with repeat business.

To support its product roadmap, Mi-Phi plans expand its 45-member team in Delhi, Mumbai, Pune, Bengaluru, and Chennai. “Mi-Phi is now an eight-month-old company in India. We have scaled up at a very rapid pace... We will hire 100 more engineers by year-end,” Balakrishnan said.