Marvell's AI deal with Google: $120 billion gimmick, revenue concerns in 2029

Marvell's expanded cooperation with Google: The true meaning behind the $12 billion

The $12 billion figure associated with Marvell's expanded cooperation with Google seems simple and straightforward: A chip supplier has locked in an unprecedented future sales order from one of the largest spenders of artificial intelligence infrastructure. However, this is not the case with the disclosed agreement.

This figure actually refers to the cumulative revenue threshold Google needs to meet to obtain most of the equity options, rather than a clearly agreed purchase commitment or a disclosed amount of backlog orders. The distinction is crucial because Marvell has said the more meaningful contribution from the partnership is not expected to begin to emerge until fiscal year 2029.

For a company with projected fiscal 2028 revenue of approximately $18 billion, the agreement creates an unusually large gap between the size implied by the long-term title and the sales contribution included in the near-term forecast. At the same time, it also adds to the pressure on the less glamorous parts of the investment case: the product portfolio, gross margins and reliance on a few large customers.

Google's $120 billion: Attribution threshold, not guaranteed income

According to documents filed by Marvell, Google obtained a warrant to purchase up to 58,970,907 shares of Marvell stock at $206.58 per share. Among them, 57,610,000 shares were vested in 240 phases. According to the company's Form 8-K document, Google will receive one period of ownership for every US$500 million in cumulative revenue from qualified customized products (as of fiscal year 2033).

Multiply the 240 issue by US$500 million to get the figure of US$120 billion. But the structure does not state that Google is obligated to purchase $120 billion in products, nor does it state that Marvell has included this amount in the backlog of orders. It simply sets a commercial performance threshold for the ownership of warrant shares.

This mechanism gives this number a different analytical role. It demonstrates the potential size of eligible revenue envisaged during the agreement measurement period, while determining the speed and final amount of procurement based on eligible products and actual revenue generated through fiscal year 2033.

This warrant is still significant. Arrangements that link substantial equity incentives to continuous revenue thresholds align Google's potential ownership interests with the expansion of customized product relationships. However, this consistency should not be confused with contract sales. The document describes the attribution path rather than the disclosed annual committed procurement plan.

The implied annual revenue size of Google is comparable to Marvell's 2028 overall revenue forecast.

According to relevant market reports, during the measurement period, the threshold of US$120 billion averaged approximately US$18.5 billion in qualified product revenue per year. This is in line with Marvell's own forecast of total revenue of approximately $18 billion for the 2028 fiscal year.

This comparison does not predict that Google will bring in $18.5 billion in annual revenue. The threshold is measured over a multi-year period, and Marvell's fiscal year 2028 figures are a company-level annual outlook. Still, comparing the two numbers side by side illustrates why warrants have attracted attention: At full realization, the size of eligible revenue envisioned by the attribution design is huge relative to Marvell's current planning basis.

Marvell raised its fiscal 2027 revenue outlook to approximately US$12 billion, an increase of approximately 45%, and raised its fiscal 2028 outlook to approximately US$18 billion from its previous target of approximately US$16.5 billion. These goals indicate that management already expects the business to expand rapidly. However, this does not say how much of this growth comes from Google.

The company did not disclose independent Google revenue forecasts in relevant performance reports. That leaves a core valuation question unsettled: whether the market should view the agreement as an extension of the already rising data center business or as a separate profit engine, with major implications outside the current cycle of investors 'modeling.

Fiscal year 2029: A key watershed between agreement commitments and reported growth

Marvell CEO Matt Murphy said that as of fiscal year 2028, revenue from Google-related projects has been reflected in previous forecasts. According to reports, the agreement is expected to make a more significant contribution starting in fiscal 2029.

This timing means that the agreement alone cannot explain the increase in fiscal year 2027 and 2028 guidance. Google-related business remains part of the overall picture in these years, but the disclosed framework distinguishes projects that have been included from subsequent larger-scale business growth. As a result, the $120 billion threshold extends a time frame far deeper than the title itself may imply.

There is no need to attribute all of Marvell's current momentum to delayed Google opportunities. In the second quarter of fiscal 2027, Marvell reported revenue of $2.739 billion, a year-on-year increase of 36.5%. According to its quarterly filing, data center revenue increased 46%.

These results suggest that the data center business has expanded strongly ahead of the expected larger Google contribution. This also makes the issue of timing more prominent. Even if the long-term economic benefits implied by the maximum income threshold for warrants cannot be demonstrated, short-term growth can be real and substantial.

Fiscal year 2029 has become a key dividing line in the narrative. Until then, investors could evaluate Marvell mainly based on its announced full-year goals and the revenue already included in them. After that, the question turned to whether the customized product project could be scaled up enough to make the warrant structure more than just a distant upper limit.

Broad range of customized chips expands potential benefits, but also changes earnings trade-offs

The scope of cooperation between the two parties is broader than that of the core tensor processing unit. Marvell identified the areas of custom chips covered by the protocol, including AI inference accelerators, memory controllers, network interface controllers, memory interface controllers, and near-memory computing. This scope expands the potential revenue pool because partnerships can cover multiple components around the AI system rather than a single chip category.

This also makes execution more demanding. Delivery of multiple customized product lines requires converting extensive technical partnerships into qualified income quickly enough to cross multiple ownership thresholds. The breadth of the agreement is the source of its upside, but it is also the reason why the headline revenue figures cannot replace evidence of product climbing.

Marvell's gross profit margin for the second quarter of fiscal 2027 increased to 53.1% from 50.4% in the same period last year. But revenue growth is not the only variable: the company said its ASIC end-to-end business model typically has low gross margins, so even if total revenue rises, a faster shift to customized products could dilute profit margins.

This does not mean that customized chips are a negative factor for profitability. It reveals a trade-off hidden by the headline numbers: the largest source of expected eligible revenue may have different profit margin characteristics than the business mix on which the reported quarterly improvement is based. As Google-related projects become more important, revenue growth and gross margin improvement may not go hand in hand.

Customer concentration further increases risk

Marvell disclosed in quarterly filings that its top ten customers accounted for 82% of total revenue in fiscal 2026, with two customers each accounting for at least 10%. Deepening the relationship with Google may enhance its strategic relevance in AI infrastructure, but it will also make execution with key customers more critical to both growth and downside risks.

So, the test is not whether US$120 billion is a huge number, but whether revenue from qualified customized products can grow beyond fiscal 2028 on a scale sufficient to support the warrant milestone without transforming Marvell's expansion into a lower-margin, more concentrated customer revenue base.