Broadcom Shares Drop as Investors Weigh Google’s Expanded AI Chip Partnership With Marvell

Broadcom (AVGO) slid sharply in trading after a report linked to renewed competition at the center of AI infrastructure. The move, described as a drop of roughly 6.2%, reflected how quickly investors can reprice expectations for custom chip suppliers when large hyperscalers adjust their vendor partnerships.

The catalyst reported by Yahoo Finance was that Google expanded its long-term partnership with Marvell Technology to co-develop additional AI accelerators and controllers. Accelerators are specialized computing blocks optimized for AI workloads, while controllers typically manage data movement and orchestration around those accelerators. In other words, the work goes beyond a single chip-design engagement and points to deeper integration across parts of an AI system.

Broadcom is often viewed as a key supplier in custom AI and data center chip efforts, including components that sit alongside or inside servers used to train and infer machine-learning models. In the market framing of the news, traders appeared to interpret Google’s expanded Marvell work as a possible shift in where new design activity and future procurement could land, potentially affecting Broadcom’s pipeline visibility or pricing leverage.

What changed, according to the reporting premise, is not that AI compute is moving away from specialized silicon, but that the division of labor among chip designers and silicon buyers could be reshuffled. If Google devotes more design scope to a partner like Marvell for accelerators and controllers, the competitive edge of other custom-silicon participants may be called into question. That is the type of read-through that can move shares even when no direct Broadcom contract was mentioned in the initial coverage.

Broadcom, for its part, did not provide a detailed rebuttal or clarification in the publicly circulated reporting tied to the stock move. The available information centered on investor reaction to Google’s partnership direction rather than on any disclosed change in Broadcom customer orders, backlog, or guidance.

AI chip partnerships are frequently iterative. Hyperscalers co-design hardware to match specific performance, power, and cost targets, and they may stagger adoption over multiple hardware generations. As a result, even a reported expansion in one supplier’s role does not automatically translate into an immediate reduction in another vendor’s revenue. Still, because the market tends to price expectations for future platform share, the announcement can quickly affect sentiment for companies perceived as closely tied to that platform build cycle.

Investors will likely look for follow-through indicates in subsequent company communications. These include any mention of AI-custom silicon development in Broadcom’s earnings materials, changes in commentary about data center accelerator demand, and any additional disclosures from Google or Marvell about the scope, timeline, or commercialization of the expanded co-development work. Until then, the precise magnitude of any impact on Broadcom remains uncertain because the reporting emphasis was on partnership expansion rather than on specific procurement outcomes.