Architecting the AI Future from India: How Marvell Is Driving Silicon Ownership, 2nm Breakthroughs, and Next-Gen Connectivity
Bengaluru, India — From pioneering 2nm design nodes to revolutionizing cloud-scale AI fabrics, Marvell Technology is turning India into a strategic powerhouse for end-to-end silicon execution. In an interview with electronica India, Vaishali Umredkar, Editor of Semiconductor For You, spoke with Rajneesh Rathi, Associate Vice President, Products at Marvell Technology on how the company is driving $250 million in local R&D investments, solving critical data center bottlenecks and cultivating top-tier engineering talent.
▶ Watch the Interview
1. Deepening End-to-End Ownership in India
The Marvell presence in India has evolved far beyond transactional engineering support into a core driver of global R&D. Operating across Bengaluru, Pune, and Hyderabad, India is the second-largest site globally for Marvell.
“We demonstrate end-to-end ownership all the way from India, which essentially fuels India’s aspiration to become a semiconductor powerhouse in the future.”
— Rajneesh Rathi, AVP – Products, Marvell Technology
Rathi emphasized that local teams drive complete product lifecycles—from early architecture definition and customer engagement to silicon design, tape-outs, bring-up, and full production support. Backed by a newly announced $250 million investment over three years, Marvell plans to double its India workforce while scaling R&D infrastructure to support global hyperscalers and Tier-1 OEMs.
2. Elevating Ecosystem Capabilities with 2nm Tech & University Labs
As a fabless semiconductor leader, Marvell is bringing frontier process nodes and advanced packaging capabilities directly to India.
- Advanced Node Design: Teams in India are currently working on cutting-edge 2nm silicon designs and complex advanced packaging architectures.
- Academic Labs: Marvell has established dedicated research labs inside top institutions like IIT Hyderabad and BITS Pilani, equipping students with high-end DPU silicon and development software.
- Talent Pipeline: Through initiatives like the Marvell STEM sponsorship which saw over 7,000 applicants this year—and robust internship programs, Marvell is actively expanding the country’s highly skilled semiconductor talent pool.
3. Solving the Next AI Bottleneck: Connectivity at Ultra-Low Power
Discussing the rapid evolution of AI infrastructure—from training and inference to complex agentic AI—Rathi noted that system bottlenecks have shifted from compute to memory, and now decisively to connectivity.
To keep pace with increasingly complex AI workloads, Marvell has released several key architectural innovations:
- Industry-First 100Tbps Switching: Flagship scale-up and scale-out switches delivering high-bandwidth connectivity across dense compute clusters.
- Market-Leading Optical Interconnects: High-speed optical links engineered for maximum performance per watt across AI data center fabrics.
- Strategic Ecosystem Collaborations: Deep partnerships with industry leaders, including custom AI infrastructure development with Google, silicon photonics integration with NVIDIA, and an industry-first cloud-scale payment security deployment with Microsoft Azure using the Marvell LiquidSecurity modules.