Nvidia is investing $2 billion (approximately 3 trillion won) in Marvell Technology, a Silicon Valley fabless semiconductor design company. The move aims to run Marvell's custom artificial intelligence chips on Nvidia's infrastructure, lowering the barrier to Nvidia's expensive graphics processing units (GPUs). The partnership is also expected to accelerate the development of silicon photonics, a key technology for the future of AI.
Nvidia announced the strategic partnership and investment in Marvell on March 31, according to the Financial Times.
Marvell specializes in data storage, server-to-server connectivity and high-speed transmission technology, and custom AI semiconductor (ASIC) design. The company has supported hyperscalers such as Amazon in developing their own AI chips. Demand for such custom chips has surged recently, driving Marvell's fiscal year 2026 (February 2025 to January 2026) revenue up 42.0% year-on-year in a sharp growth trajectory.
Through the partnership, the two companies plan to combine Nvidia's chip interconnect networking technology, NVLink Fusion, with Marvell's AI semiconductors to develop custom AI systems. Marvell will provide custom chip solutions compatible with NVLink Fusion, while Nvidia will supply core infrastructure technology including central processing units (CPUs) and connectivity components to ensure optimal performance.
The companies will also collaborate on silicon photonics to accelerate data flow within data centers. Silicon photonics uses light instead of electrical signals to transmit data, boosting transmission speeds while reducing power consumption. The technology is considered essential for scaling large-scale AI infrastructure.
Nvidia's investment in and partnership with Marvell is seen as an extension of its strategy to expand its ecosystem outward. The collaboration is expected to make it easier for big tech companies to integrate their self-designed AI chips into Nvidia's infrastructure ecosystem. Nvidia CEO Jensen Huang said in May last year that the company would broaden access for multiple partners to offer customer-customized chips using Nvidia's infrastructure.
The deal is also interpreted as a strategic move by Nvidia to further solidify its position at the center of AI computing. As some big tech companies have recently shifted toward adopting self-designed custom chips instead of expensive GPUs, analysts say Nvidia aims to maintain its influence by bringing even these players into its ecosystem. Nvidia is also pursuing mergers and acquisitions of promising companies to strengthen its position in the AI market. Reuters described the approach as "Nvidia seeking to broaden the scope of the overall AI semiconductor industry while maintaining its dominant position."
