Marvell Technology (MRVL) is trading higher as investors react to renewed optimism that Google is expanding its custom silicon supply chain to include Marvell for AI inference-focused chips. Recent reporting described discussions around two projects: a memory processing unit intended to work alongside Google’s Tensor Processing Units and a next-generation TPU tailored for inference workloads—exactly the area where hyperscalers are trying to cut cost per token and reduce latency as AI “agent” usage grows.

A Google program would be strategically meaningful because it would reinforce Marvell’s positioning as a large-scale custom silicon and connectivity supplier to hyperscalers, potentially adding another major customer alongside existing cloud relationships. Even before any contract details are known, traders typically re-rate custom silicon stories when a top-tier hyperscaler design win looks plausible because it can extend multi-year revenue visibility and pull-through demand for high-speed interconnect and optical components.

Marvell also announced on April 22, 2026 that it acquired Polariton Technologies, adding plasmonics-based modulation technology and silicon photonics expertise to its optical roadmap. The company framed the deal as a way to support scaling toward 3.2T and beyond in coherent and optical interconnect platforms, a theme tightly linked to AI cluster buildouts and the rising bandwidth needs of scale-up and scale-across architectures.

Key swing factors now are whether Google provides any confirmation around suppliers or silicon partners and whether Marvell offers incremental color on program timing, scope, and expected contribution to data-center revenue. Investors will also watch follow-on details around integration of Polariton’s technology into Marvell’s optical DSP and silicon photonics portfolio, since optical attach rates and bandwidth transitions can directly influence growth expectations.