Broadcom shares came under pressure after Marvell Technology announced an expanded partnership with Google covering custom silicon for artificial intelligence infrastructure. The market reaction reflects growing investor sensitivity to changes in the competitive landscape for custom AI accelerators, as hyperscalers increasingly develop specialized chips to meet rising computing demand and manage infrastructure costs.

Marvell’s Expanded Google Partnership Changes the Competitive Picture

Marvell’s agreement with Google extends beyond traditional custom-chip development, encompassing technologies associated with Google’s Tensor Processing Units, including AI accelerators, networking, storage and memory. The broader scope signals that Google is seeking greater flexibility across multiple layers of its AI infrastructure rather than relying on a single supplier for every component.

As part of the agreement, Google received a warrant to purchase up to approximately 59 million Marvell shares at $206.58 per share, representing potential consideration of about $12.2 billion. The structure links the potential share purchases to performance and revenue milestones extending through 2033, giving Marvell a significant long-term commercial opportunity while strengthening the strategic relationship between the two companies.

Why Broadcom Shares Came Under Pressure

Broadcom has been a major supplier in Google’s custom TPU ecosystem, making the Marvell announcement particularly important for investors assessing the future allocation of Google’s AI infrastructure spending. Broadcom shares fell nearly 5% following the news as investors considered whether an expanded Marvell relationship could eventually reduce Broadcom’s share of Google’s custom silicon business.

However, the development does not necessarily mean that Broadcom is being displaced. Google’s AI infrastructure requirements are expanding rapidly, and the company has continued to invest heavily in custom accelerators and supporting data-center technologies. Broadcom also has relationships with other major technology companies, including an expanded custom AI silicon partnership with Meta, highlighting the broader demand for specialized chips among hyperscalers.

AI Infrastructure Demand Remains the Bigger Variable

The Marvell-Google agreement illustrates a broader shift in the semiconductor industry: hyperscalers are increasingly developing custom AI processors designed for specific workloads rather than depending entirely on general-purpose GPUs. Custom silicon can potentially improve performance, power efficiency and total infrastructure costs, particularly as AI workloads move from training toward large-scale inference.

Marvell has also been expanding its capabilities in AI memory, storage and high-speed connectivity, areas that are becoming increasingly important as AI models require larger memory capacity and faster movement of data. This gives the company an opportunity to participate in AI infrastructure beyond the accelerator itself.

The next focus for investors will be whether Marvell converts the expanded Google relationship into the revenue growth implied by the long-term agreement and whether Broadcom maintains its position as a leading supplier of Google’s custom AI infrastructure. The competitive landscape will likely remain fluid as Google and other hyperscalers diversify their chip supply chains. For Broadcom, the key question is whether the expanding overall AI market can offset potential pressure from greater supplier diversification; for Marvell, execution against demanding performance targets will determine how much of the opportunity ultimately translates into financial results.