Loading article…
Nvidia's $2 billion stake in Marvell brings custom XPUs and silicon‑photonic networking into its NVLink Fusion platform, boosting rack‑scale AI compute.
Nvidia poured $2 billion into Marvell Technology, sealing a strategic partnership that links Marvell’s custom XPUs and high‑speed networking to Nvidia’s NVLink Fusion AI ecosystem【1】. The deal, announced on March 31, 2026, makes Marvell the latest silicon designer to join Nvidia’s proprietary rack‑scale interconnect, enabling third‑party accelerators to talk directly to Nvidia GPUs over a low‑latency, high‑bandwidth fabric【2】.
NVLink Fusion, launched in May 2025, is designed to let customers build heterogeneous AI infrastructure that remains fully compatible with Nvidia’s GPUs, CPUs, DPUs and networking stack. Under the partnership, Marvell will supply custom XPUs and NVLink‑compatible scale‑up networking, while Nvidia will provide its Vera CPUs, ConnectX NICs, BlueField DPUs, NVLink interconnects and Spectrum‑X switches【2】. By marrying Marvell’s expertise in high‑performance analog, optical DSP, silicon photonics and custom silicon with Nvidia’s AI stack, the two firms aim to give data‑center operators more flexibility in assembling AI‑compute clusters without abandoning Nvidia’s ecosystem【1】.
The collaboration also extends to telecom, with both companies pledging to turn global 5G/6G networks into AI‑ready infrastructure using Nvidia’s Aerial AI‑RAN and Marvell’s photonic‑fabric technology acquired from Celestial AI last year【1】. Nvidia’s CEO Jensen Huang framed the move as a response to surging token‑generation demand, calling the moment an “inference inflection” that requires faster, more scalable AI factories【1】. Marvell’s CEO Matt Murphy echoed the sentiment, highlighting the importance of high‑speed connectivity and silicon‑photonic solutions for scaling AI workloads【2】.
While Nvidia’s rivals AMD, Intel and Broadcom have stayed out of the NVLink Fusion program, favoring the open UALink standard, Marvell’s entry underscores a growing divide in the industry over proprietary versus open interconnects【1】. Notably, Marvell already helps Amazon develop its Trainium AI accelerators, a relationship that predates the Nvidia deal and could create tension if the two cloud giants’ strategies clash【1】. The announcement did not clarify whether the partnership will affect Amazon’s Trainium roadmap, leaving that question open.
If Nvidia’s NVLink Fusion gains traction, Marvell’s custom silicon could become a key revenue source for Nvidia even as it powers non‑Nvidia accelerators, reinforcing Nvidia’s control over the rack‑scale AI market. The next test will be whether customers adopt the mixed‑vendor approach at scale, and how Nvidia’s competitors respond to an ecosystem that now includes a major custom‑chip player.
Coverage is mostly measured — 48 of 61 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 2 outlets · Jun 14, 2026 · How we report
Nvidia designs its chips but relies on Taiwan Semiconductor Manufacturing (TSMC) for the actual manufacturing process.
NVLink Fusion allows third-party accelerators to communicate with Nvidia GPUs and infrastructure over a high-bandwidth, low-latency interconnect.
CEO Jensen Huang has forecast that Nvidia could reach $1 trillion in AI-related revenue by the 2027 calendar year.
No, Nvidia partners with other firms like Marvell Technology to integrate custom silicon and networking components into its broader AI infrastructure ecosystem.