Bay Area/ San Jose

Nvidia’s $2 Billion Silicon Power Play Locks In Marvell On Santa Clara’s AI Front Line

AI Assisted Icon
Published on March 31, 2026
Nvidia’s $2 Billion Silicon Power Play Locks In Marvell On Santa Clara’s AI Front LineSource: Daniel J. Prostak; Crocodiletiger~commonswiki Crocodiletiger~commonswiki used courtesy of Daniel Prostak, CC BY-SA 4.0, via Wikimedia Commons

NVIDIA is dropping $2 billion on Marvell and pulling the Santa Clara chipmaker directly into its NVLink Fusion rack-scale program, a high-stakes move aimed at accelerating the adoption of silicon photonics links in AI data centers. The partnership pairs Marvell’s custom XPU, networking and optical tech with Nvidia’s Vera CPU, ConnectX NICs and Spectrum-X switches to build tighter, rack-scale systems for AI training and inference. For Bay Area engineers and equipment suppliers, it effectively stakes out another piece of next-generation AI infrastructure in local fabs and packaging shops.

What Was Announced And Who Broke The Story

As reported by Bloomberg, the companies said the agreement covers integration with NVLink Fusion and joint work on silicon photonics, and that Nvidia has taken a $2 billion position to deepen the relationship. The news landed through company statements and a shared press release spelling out the technology scope and commercial intent.

What The Companies Said

In today's statement, Marvell framed the deal as connecting the company “to the NVIDIA AI factory” and confirmed that “NVIDIA has invested $2 billion in Marvell,” according to Marvell. NVIDIA CEO Jensen Huang called the moment an “inference inflection,” while Marvell chairman and CEO Matt Murphy cast the partnership as a way to accelerate high-speed connectivity and optical interconnects for large AI systems.

Why Silicon Photonics Matters

Silicon photonics, which puts light-based transmitters and receivers directly onto silicon wafers, promises far higher bandwidth and much lower power per bit than traditional copper interconnects. That is why researchers and vendors see it as critical for hyperscale AI clusters. Peer-reviewed surveys and recent experimental work back up the energy and bandwidth advantages for data centers, as outlined in npj Nanophotonics. Industry watchers also point to Nvidia’s broader optics push, including major photonics investments announced this month, as part of a strategy to lock in supply and capacity for photonic components. Those recent optics commitments were summarized by Tom’s Hardware.

Local Stakes And Marvell’s Optics Buildout

Headquartered in Santa Clara, Marvell has been steadily assembling an optical-interconnect stack. In December, the company agreed to acquire photonics startup Celestial AI in a deal valuing the business at roughly $3.25 billion, according to an earlier Marvell filing. Marvell says that the acquisition adds a "Photonic Fabric" technology intended to be co-packaged with XPUs and switches to cut power and boost bandwidth, positioning local suppliers and packaging partners to step in if volume production ramps. Analysts and semiconductor research groups have flagged Marvell’s moves as part of a broader turn toward optical scale-up fabrics for large AI deployments.

Market Reaction And What To Watch Next

Traders did not wait around. Marvell shares jumped after the announcement, with market reports citing a double-digit intraday gain as investors priced in the deal’s strategic upside, according to Investing.com. Next up, keep an eye on investor filings and any SEC disclosures that spell out the investment mechanics, the timetable for pulling the products into NVLink Fusion, and how fast the partners plan to scale silicon-photonics manufacturing capacity. If they follow the pattern of earlier optics rollouts, commercial modules and co-packaged optics demos could start surfacing in product briefs and customer pilots over the next 12 to 24 months.