Shares of chipmaker Marvell Technology saw a significant climb in March, tacking on a 21.3% gain. The buzz around the office—the hum of the AC unit, the constant tapping of keyboards—reflected the wider market sentiment. According to Misryoum analysis, this move wasn’t just random luck; it was driven by a solid fourth-quarter earnings beat and, more importantly, that massive $2 billion investment and partnership announcement with Nvidia.
Looking at the raw numbers, Marvell’s fiscal fourth quarter was impressive. Revenue hit $2.2 billion, a 22.1% jump, and adjusted earnings per share rose 33.3% to $0.80. They’re projecting a 9% sequential revenue hike for the first quarter, which—actually, let me double-check that—yes, the guide sits at $0.79 per share. These figures handily beat the expectations that everyone had floating around.
There was this lingering worry lately that Marvell might be losing its grip on the Amazon business for those custom Trainium chips. But the guidance suggests the XPU-attach business is doing just fine. Plus, they’ve brought Microsoft into the fold with the Maia2 chip. It’s a shifting landscape, right? You have your AI infrastructure, which is usually either Nvidia-based or custom-built, but now the two worlds are colliding.
Nvidia is pumping $2 billion into Marvell for a reason. They’re looking to integrate, enabling these hybrid architectures that mix XPUs with Nvidia’s own NV-Link or Vera CPUs. There’s also talk of silicon photonics—basically replacing copper networking with light. Nvidia’s current tech is copper-heavy, so they’re clearly betting on Marvell’s expertise to lead that transition into optical networking. It’s a smart pivot.
It’s interesting how this played out. Over the last year, Nvidia has been shaking hands with companies that you’d think were rivals, pulling them into the ecosystem instead of fighting them. It’s a massive vote of confidence for Marvell.
After a rough 2025, Marvell looks like a clear winner in the AI build-out. With agentic AI inference requiring constant communication between models, their networking strength is going to be tested—or maybe fully realized. The stock isn’t exactly a bargain anymore at 27 times earnings, but for those banking on networking growth in the age of generative AI? It’s still on the radar.