Marvell raises long-term AI chip targets, sees revenue up to $90 billion by 2031

Marvell Technology told investors on Tuesday that it expects its annual revenue to climb as high as $90 billion by its 2031 fiscal year, a roughly tenfold increase from last year's results, as the chipmaker leans further into designing custom processors for the data centers run by Amazon, Google, Microsoft and Meta. The announcement, made at an all-day investor presentation in New York, sent Marvell shares up as much as 11 percent in early trading before the stock settled to close 5.8 percent higher.

Chairman and Chief Executive Matt Murphy used the gathering to raise the company's near-term targets as well, lifting the fiscal 2028 revenue forecast to about $20 billion, up from the $18 billion figure Marvell gave in August and above Wall Street's roughly $18.2 billion consensus estimate. The company's current fiscal year, which ends in January, is now expected to produce around $12 billion in revenue, up 45 percent from the prior year.

The numbers behind the pitch

Marvell is a fabless semiconductor designer, meaning it does not operate its own factories and instead contracts manufacturing to foundries such as Taiwan Semiconductor Manufacturing Company. Its business has split in recent years between general-purpose networking and storage chips and a faster-growing unit that designs custom silicon, known in the industry as application-specific integrated circuits, tailored to the specific workloads of individual cloud computing customers.

That custom silicon business generated roughly $1.5 billion in revenue last fiscal year, about 18 percent of the company's total sales. At the investor day, Marvell executives said they now expect custom revenue to exceed $12 billion by fiscal 2029, an upward revision from a prior target of more than $10 billion. Marvell also said it expects data center revenue, which includes both custom chips and the networking components that connect them, to grow more than 60 percent in fiscal 2028. Taken together with Marvell's 2031 scenario of $70 billion to $90 billion in total revenue, the company is betting that the market for custom AI accelerators and the high-speed links between them will keep expanding faster than the broader semiconductor industry. Marvell pointed to a combined addressable market for AI infrastructure of roughly $400 billion by 2030 as the opportunity it is chasing.

How Marvell got here

Marvell's pivot toward custom silicon accelerated this year after it struck an expanded agreement with Google covering chips that attach to Google's own Tensor Processing Unit lineup, including inference accelerators, storage controllers and memory interface chips. Under the terms disclosed in an August filing with the Securities and Exchange Commission, Google received a warrant to buy as many as 58.97 million Marvell shares at $206.58 apiece, a stake worth roughly $12.2 billion if fully exercised. The warrant vests in tranches tied to Google's custom-chip purchases, unlocking one block of shares for every $500 million in revenue Google generates for Marvell through fiscal 2033 — a structure that, if every milestone is hit, could be worth about $120 billion in cumulative sales over that period.

The broader industry shift explains why Marvell is making this bet. Nvidia still supplies the large majority of processors used to train and run large AI models, but Amazon, Google, Microsoft and Meta have each built or commissioned their own chips, including Amazon's Trainium line, Google's TPUs, Microsoft's Maia processors and Meta's MTIA accelerators, in an effort to cut costs and reduce reliance on a single supplier. Broadcom has emerged as the dominant partner for hyperscalers designing that silicon, while Marvell has positioned itself as the leading alternative, working with multiple cloud providers on both full custom processors and the surrounding chips that move data in and out of them.

"We are seeing broad-based strength across our Data Center portfolio, including strong demand in Connectivity and a significant acceleration in our Custom business beginning in the second half of fiscal 2027. We look forward to showcasing these growth drivers and sharing Marvell's long-term strategy for enabling the continued expansion of AI infrastructure at our Investor Day on October 6th, 2026," Matt Murphy, Marvell's chairman and chief executive, said in the earnings statement that previewed Tuesday's event.

Who is watching the numbers

The audience for Tuesday's presentation extended well beyond Marvell's own shareholders. Its four largest custom-silicon customers, Amazon, Google, Microsoft and Meta, are each weighing how much of their AI infrastructure spending to route toward in-house chip programs that Marvell helps design and manufacture, rather than toward Nvidia's general-purpose graphics processors or Broadcom's rival custom-chip business. Competing chipmakers, foundry partner TSMC and semiconductor equipment suppliers all have a stake in whether Marvell's growth targets hold up, since a miss would ripple through a supply chain that has priced in years of AI-driven expansion.

Marvell's roughly 7,500 employees and its investors are also directly exposed to the swings in the stock, which has risen more than 200 percent so far this year on the strength of the AI buildout, making Tuesday's targets a high-stakes marker for whether that rally can continue. The company's non-GAAP gross margin guidance, which has been drifting lower as the lower-margin custom business grows as a share of total sales, is a particular point of scrutiny for analysts trying to judge how much of the added revenue will reach the bottom line.

Reaction on Wall Street

The initial market reaction was positive but not uniformly so. Shares jumped as much as 11 percent in the first hours of trading, touching their highest level since June, before paring those gains to close up 5.8 percent at $287.01, according to trading data reported by the Motley Fool. Some of the moderation reflected caution among analysts about how far out Marvell's biggest numbers reach: the $70 billion to $90 billion fiscal 2031 target is five years away, and the company's custom-chip revenue has previously disappointed investors who expected faster near-term contributions from the Google partnership specifically.

Coverage of the event noted that Marvell's new fiscal 2028 figure of about $20 billion already exceeds what most Wall Street analysts had modeled, a sign that management is trying to get ahead of skepticism that built up after the stock fell sharply following its most recent quarterly earnings report in August, when investors judged otherwise strong results against an AI rally that had priced in even faster growth.

What happens next

Marvell's next scheduled test of these targets comes with its fiscal third-quarter earnings report, expected in early December, when the company will disclose whether bookings and shipments are tracking toward the raised fiscal 2027 and 2028 figures. Investors are likely to pay particular attention to any updates on the pace of Google's custom-chip purchases, since the pace at which Google's warrant vests is a direct, contractually disclosed proxy for how quickly that relationship is generating revenue. Analysts covering the stock are expected to publish revised price targets and models in the coming days incorporating the new long-term guidance, which will determine whether Tuesday's rally holds or proves to be another overreaction in a sector where expectations for AI infrastructure spending have repeatedly run ahead of delivered results.