Santa Clara, California — April 1, 2026, 04:08 PDT
NVIDIA has announced a significant investment of $2 billion in Marvell Technology, further enhancing their partnership to integrate Marvell’s custom AI chips and networking hardware with NVIDIA’s systems. This strategic move resulted in a notable increase in Marvell’s stock, which experienced a rise of approximately 7% during the trading session, while NVIDIA’s shares also saw a gain of 2.7%.
This investment highlights NVIDIA’s commitment to maintaining a pivotal role in AI data-center budgets, particularly as some consumers opt for semi-custom chips over traditional GPUs. In instances where NVIDIA is unable to sell every chip in the server rack, the company aims to provide essential interconnects, CPUs, and the accompanying software stack.
The timing of this investment is particularly relevant, as major players in the tech industry, including Alphabet, Microsoft, Amazon, and Meta, are expected to invest over $630 billion into AI infrastructure this year. Moreover, Broadcom has projected that AI-chip revenue could exceed $100 billion by 2027, driven by an increasing demand for custom silicon.
Under the terms of the agreement, Marvell will supply custom XPUs—processors specifically designed for targeted workloads—along with networking equipment that is compatible with NVIDIA’s NVLink Fusion platform. The two companies will also collaborate on advancements in silicon photonics, a technology that utilises light to transfer data between chips, which aims to enhance speed and reduce power consumption.
Jensen Huang, Chief Executive of NVIDIA, stated that the demand for AI inference is rapidly increasing. He remarked, “The inference inflection has arrived. Token generation demand is surging, and the world is racing to build AI factories.” Huang further emphasised that the collaboration will assist customers in developing specialised AI computation.
Matt Murphy, CEO of Marvell, described the expanded partnership as a response to the growing need for high-speed connectivity, optical interconnects, and accelerated infrastructure to support AI advancements. He noted that integrating Marvell’s custom silicon and photonics with NVIDIA’s ecosystem will facilitate the construction of scalable and efficient systems for customers.
Jacob Bourne, an analyst at eMarketer, indicated that this partnership grants NVIDIA access to Marvell’s semi-custom silicon and advanced optical interconnects, both of which are essential as bandwidth demands increase and power efficiency becomes increasingly critical for large AI systems. Bourne remarked that investors are likely to perceive this collaboration as a means of alleviating some obstacles, enabling non-NVIDIA AI chips to operate within data centres typically dominated by NVIDIA hardware.
This partnership aligns with NVIDIA CEO Jensen Huang’s emphasis during last month’s GTC, where he focused on inference—the stage at which an AI responds to user queries. This area sees custom processors and CPUs competing directly with NVIDIA’s graphics chips. It is noteworthy that Meta also introduced a range of new proprietary AI chips around the same time, indicating that major clients are eager to manage costs and energy requirements.
However, challenges remain. According to IDC data, Chinese suppliers are projected to account for approximately 41% of the AI accelerator server market in China by 2025, with Huawei leading the charge. Meanwhile, NVIDIA’s market share has decreased to 55%, a decline attributed to U.S. export restrictions prompting customers to consider local alternatives. Both NVIDIA and Marvell have acknowledged the potential risks posed by shifting supply and demand dynamics, as well as legal or regulatory developments and broader market conditions.
NVIDIA remains steadfast in its strategy: despite the potential fragmentation of the AI chip landscape, the expectation is that customers will continue to rely on its platform. By integrating Marvell into its operations, NVIDIA expands its offerings to include semi-custom chips and optical connectivity, complementing its existing portfolio of GPUs, CPUs, networking equipment, and software solutions.