MiPhi Semiconductors Pvt. Ltd., a Mumbai-based semiconductor startup, has unveiled a solution that allows small-scale AI model training without the use of graphics processing units (GPUs), in collaboration with Intel Corp. and SUSE. The development aims to make AI training more accessible, efficient, and cost-effective by leveraging direct CPU upgrades.
The initiative addresses a key bottleneck in the AI ecosystem: the high cost and limited availability of GPUs, which have traditionally been central to AI training workloads. By optimizing CPU performance through advanced hardware-software integration, the collaboration enables developers and enterprises to train models directly on CPUs, reducing infrastructure costs and simplifying deployment.
“Our collaboration with Intel and SUSE reflects MiPhi’s mission to democratize technology,” said Prasad Balakrishnan, CEO of MiPhi Semiconductors. “By reducing dependency on GPUs, we are opening the door for developers and organizations—particularly in emerging markets—to adopt AI faster and more affordably. AI workloads are advancing at an unprecedented pace, and the infrastructure that powers them must evolve in parallel.”
The solution combines MiPhi’s semiconductor expertise with Intel’s CPU architecture and SUSE’s enterprise-grade Linux platform, targeting a range of AI workloads, including natural language processing, computer vision, and recommendation systems. According to company executives, the integrated approach allows CPUs to handle tasks that previously required GPU acceleration, potentially transforming how small-scale AI projects are deployed.
“Intel is committed to delivering compute technologies that are accessible, efficient, and powerful,” said Santhosh Viswanathan, Vice President and Managing Director, India Region, Intel. “This collaboration with MiPhi and SUSE unlocks the potential of Intel CPU-based architectures for AI workloads, reducing barriers to adoption and supporting developers and organizations in innovating at scale.”
SUSE highlighted the enterprise software perspective, with Sachin Vig, Director Strategic Channels & Alliances, Indian Subcontinent, stating: “Through this collaboration, we are enabling developers to accelerate their innovation cycles by training AI models on secure and scalable Linux infrastructure without expensive GPUs. Our aim is to democratize AI and empower industries to advance their digital transformation journeys with greater agility.”
The partnership underscores a broader industry trend of expanding AI capabilities beyond GPU-centric frameworks, while also positioning India as a growing hub for semiconductor and AI innovation. Founded in 2024 as a joint venture between Micromax Informatics and Phison Electronics, MiPhi Semiconductors focuses on AI-enabled storage solutions and semiconductor design, aiming to strengthen India’s presence in global AI and storage markets.
By reducing reliance on specialized hardware, the collaboration promises to make AI more inclusive and scalable, particularly for startups, researchers, and enterprises in cost-sensitive markets.
