Marvell raises fiscal 2028 revenue target to $20 billion, guides to $70 billion to $90 billion by fiscal 2031

Marvell Technology used its Investor Day in New York on October 6, 2026 to reset how Wall Street sizes its custom artificial intelligence data-center business. The chipmaker raised its fiscal 2028 revenue target to about $20 billion and, for the first time, guided to fiscal 2031 revenue of $70 billion to $90 billion. Its shares rose about 7% as investors worked through the numbers.

Custom silicon is not a new story for the company. Marvell outlined a strategy centered on custom and cloud-optimized silicon, chips designed specifically for use in data centers, at its 2021 investor day, and it has been one of the biggest beneficiaries of the AI infrastructure boom since then. That business has become a major growth engine as large technology companies develop in-house AI processors to reduce their reliance on Nvidia's general-purpose chips.

The anchor customer is Alphabet's Google. Marvell disclosed an agreement with Google in August that could generate up to $120 billion in sales through fiscal 2033 if performance milestones are achieved, and management has built its new long-range plan around that program and around similar work with Microsoft and Amazon.

Expectations ahead of the event were already elevated. The stock has more than tripled in value so far this year and the company carried a market capitalization above $200 billion into the New York presentation, which made the Investor Day as much a test of credibility as a celebration.

Key Facts

Reuters reported on October 6 that Marvell raised its fiscal 2028 revenue forecast to about $20 billion, above Wall Street estimates, as demand for its custom data-center chips grows along with a surge in AI spending. Analysts had been expecting revenue of $18.2 billion for 2028, according to data compiled by LSEG. The new target also sits roughly $2 billion above the level Marvell set on its August 27 earnings call, when it raised its revenue outlook to about $18 billion from $16.5 billion.

AlphaPilot reported on October 6 that Marvell sees a total addressable market of approximately $400 billion for its business by 2030, and that the fiscal 2031 range of $70 billion to $90 billion was disclosed for the first time at the event. At the $80 billion midpoint, that guidance exceeds Wall Street estimates of $46.85 billion, according to four analysts polled by Visible Alpha.

The Google arrangement carries the largest single number in the plan. Marvell handed Google warrants to buy as many as 58.97 million shares at $206.58 apiece, worth about $12.2 billion if Google exercises all of them, with vesting tied to cumulative qualifying revenue of $120 billion. Each additional $500 million of chip orders unlocks another tranche, and if Google hits its purchasing targets through fiscal 2033 it would end up holding close to 7% of Marvell, making it the company's fifth-largest shareholder.

Today's custom business is far smaller. Startup Fortune reported on October 6 that custom silicon generates roughly $1.5 billion a year, close to 18% of Marvell's sales in fiscal 2026, and that management expects that business to grow more than 20% in fiscal 2027 and more than double again in fiscal 2028. Marvell had previously set a target of more than $10 billion in custom revenue by fiscal 2029.

Reported results are already moving. Revenue for the second quarter of fiscal 2027, reported in late August, rose 37% to $2.739 billion, with record data-center revenue of $2.172 billion, up 46% year over year and 18% sequentially, which represented 79% of the total. Management guided data-center revenue to grow about 60% in fiscal 2027 and more than 60% again in fiscal 2028, and forecast fiscal third-quarter revenue of $3.15 billion, up more than 50% year over year at the midpoint.

Analysis

What this really means is that Marvell is asking the market to price it as a platform for hyperscaler-owned compute rather than as a supplier of commodity connectivity parts. Chairman and CEO Matt Murphy argued on stage that hyperscale cloud operators increasingly prefer application-specific chips tuned to their own workloads and owned outright over renting compute built for everyone. The $400 billion addressable market and the $70 billion to $90 billion fiscal 2031 range are the quantitative version of that argument.

The comparison that matters is Broadcom, which runs the same custom silicon strategy at far larger scale and pairs it with a substantial infrastructure software business. Startup Fortune reported on October 6 that Broadcom raised its own fiscal 2026 AI revenue guidance to around $58 billion and told investors it has line of sight to more than $100 billion in AI chip revenue in fiscal 2027, with a custom silicon backlog of $73 billion over an 18-month window. Investors treated Marvell's pitch as good news for the whole category: 24/7 Wall St. reported on October 6 that Marvell stock was at $287.59, up 6% in morning trading, while Broadcom advanced 4% to $376.09 and the iShares Semiconductor ETF rose 0.8%.

The bigger picture here is that Marvell is not trying to out-design Nvidia's merchant GPUs. It supplies the pieces that turn a hyperscaler's architecture into working silicon: ASIC design, high-speed SerDes links, HBM memory interfaces, advanced packaging, chiplet integration and access to leading-edge manufacturing. It already works with all four of the largest hyperscalers and has disclosed three XPU design wins and nine XPU-attach programs. Nvidia CEO Jensen Huang lent the strategy some external cover at COMPUTEX 2026, calling Marvell the next trillion-dollar company.

The risk is concentration and timing. Marvell designs custom chips to each client's own specification, so revenue depends on winning individual programs that then ramp over long development cycles, a model that leaves the stock more sensitive to the timing of any one program. There can be no repeat of the slip after the August earnings report, when shares fell about 10% despite record revenue because investors wanted firmer timing on the Google ramp.

Why It Matters

The Investor Day formalizes a shift in how the largest buyers of AI hardware plan to spend. Alphabet, Microsoft and Amazon are all working with Marvell on silicon tuned to their own workloads, which reduces their dependence on Nvidia and changes the bargaining dynamics of the entire accelerator market. A credible second source of custom-designed AI compute gives hyperscalers leverage on price, supply and roadmap control.

For Marvell specifically, the numbers now on the table are an order of magnitude beyond what the company reported as recently as fiscal 2026, when custom silicon accounted for roughly $1.5 billion of sales. Hitting even the low end of the fiscal 2031 range would require the Google program and its peers to ramp on schedule, which is why management described the existing $10 billion fiscal 2029 custom revenue target as carrying meaningful upside bias and, in Murphy's own words, conservative.

For the wider semiconductor sector, the event reinforced that the AI buildout is broadening beyond general-purpose GPUs into memory interfaces, packaging, high-speed links and chiplet integration. Broadcom's 4% move and the modest rise in the iShares Semiconductor ETF show that investors read Marvell's targets as a statement about the whole custom silicon category rather than about one company.

Next Up

The near-term checkpoint is Marvell's fiscal third-quarter report, guided to $3.15 billion in revenue, up more than 50% year over year at the midpoint. Beyond that, investors will watch for concrete evidence that the Google programs are becoming significant in fiscal 2029 and beyond, the point at which the $120 billion agreement and the warrant vesting schedule start to bite.

Management has also left itself room to raise the fiscal 2029 custom revenue target again. If Google, Microsoft and Amazon ramp as described, the debate will shift from whether Marvell can reach $10 billion in custom revenue to how quickly it approaches the $70 billion to $90 billion total revenue range it now says is possible in fiscal 2031.