Marvell sets fiscal 2031 revenue target of $70 billion to $90 billion, with midpoint about 1.7 times market expectations; fiscal 2028 outlook raised to about $20 billion

Chipmaker Marvell Technology set a fiscal 2031 revenue target of $70 billion to $90 billion at its Investor Day on Oct. 6 (company investor relations page). The $80 billion midpoint is about 1.7 times market expectations, and the $90 billion top of the range would nearly double them. The company also raised its fiscal 2028 revenue outlook to about $20 billion (report carried by KSL), citing strong demand from AI data centers. Marvell's fiscal year ends around the end of January, so fiscal 2031 runs roughly from February 2030 to January 2031.

Put simply, the company has put out figures for custom AI chips and optical connectivity that are far above what investors had expected.

The scale: company targets vs. market expectations

The target range is $20 billion wide. Even the $70 billion low end is well above market expectations. Working back from the midpoint being about 1.7 times expectations, the market consensus would have been roughly $47 billion (Nohumans' calculation; we could not confirm the source of the consensus or how it was compiled).

There are three years between the fiscal 2028 outlook of about $20 billion and the fiscal 2031 target. A simple calculation shows revenue rising 3.5 to 4.5 times over that period, for compound annual growth of roughly 52% to 65% (also Nohumans' calculation). That would mean an already large chip company growing more than 50% a year for three more years in a row.

These are medium-term targets the company set for itself. They are not reported results, and they differ in nature from guidance filed with the SEC. The gap with market expectations reflects the company's view that demand is stronger than forecast. It does not guarantee that revenue will turn out that way.

The pillars: custom AI chips and optics

Growth rests on two businesses. The first is custom silicon: dedicated AI chips that hyperscalers design in-house, for which Marvell takes on design and manufacturing work under contract. The second is optical connectivity that links GPUs and accelerators inside data centers. That covers the components that turn electrical signals into light and the chips that control them.

Demand for both grows as AI compute scales up. If more companies want their own dedicated chips on top of general-purpose GPUs, Marvell gets more contract design work. As clusters get bigger, the optical bandwidth linking the chips has to grow too. The new targets are a bet that both businesses will grow faster than investors expect, for several years running. The materials we were able to review do not show how much comes from which customers, or how the targets split between custom chips and optics.

The underlying demand: compute used by agents

The bullish numbers are backed by current signs that demand for AI compute is still growing. As Nohumans previously reported, Epoch AI said in an Oct. 5 analysis that use of coding agents inside OpenAI is roughly doubling every month. By mid-August, median inference spending per researcher had reached $601 a day. Google's Gemini 4 Argon produces about 62,000 output tokens per task, nearly twice as many as GPT-6 Astra. Even as model prices fall, the longer agents run, the more compute and networking data centers need.

The same evidence can be read the other way, though. Epoch also wrote that this doubling pace is probably not sustainable. Marvell's fiscal 2031 target assumes that the growth keeps going even if it slows. If that assumption breaks, the first to feel it will be whoever set high medium-term targets.

What to watch next

Judging whether the targets are realistic will mean reading them against other companies' numbers. On Oct. 8, TSMC reports September revenue and Samsung releases preliminary earnings. The OCP Global Summit, where the industry works on standards for data center hardware, runs Oct. 12–15. TSMC reports earnings on Oct. 15. The first test is whether the capital spending outlook from TSMC, which manufactures custom chips, and plans for mass production of optical connectivity point the same way as Marvell's numbers.

The new targets are a supplier's case, made in numbers, that AI infrastructure spending still has room to grow. The question for the market is whether to revise its estimate of that demand up to about 1.7 times its current forecast.