NEW YORK, Oct 6: Marvell Technology has raised its fiscal 2028 revenue forecast to approximately $20 billion, signalling growing confidence in the semiconductor industry’s ability to benefit from the rapid expansion of artificial intelligence infrastructure.
The revised outlook, announced during the company’s investor day in New York on Tuesday, is above the roughly $18.2 billion revenue expected by analysts. Marvell’s shares gained about 6% following the announcement, while shares of rival chipmaker Broadcom also moved higher.
The company’s stronger outlook is closely linked to rising demand for custom chips used in data centres. As technology companies invest heavily in artificial intelligence, they are increasingly looking beyond general-purpose processors and developing specialised silicon designed for specific workloads.
Marvell has been building its strategy around custom and cloud-optimised chips since 2021. The company has benefited from the expansion of AI infrastructure as cloud providers and other large technology businesses seek greater control over computing hardware.
Artificial intelligence systems require enormous amounts of computing power. Training and operating advanced models can involve large clusters of processors working together, creating demand not only for AI accelerators but also for networking, storage, optical connectivity and other components required to link data-centre systems.
This broader infrastructure requirement has created opportunities for semiconductor companies that supply technologies surrounding AI processors.
Marvell’s custom chip business has consequently become an important part of its long-term growth strategy. The company said it now expects revenue from custom chips to reach approximately $12 billion in fiscal 2029, raising its previous target of $10 billion.
The development reflects a major change taking place within the technology industry. Large cloud and internet companies are increasingly designing or commissioning processors tailored to their own requirements.
Such chips can be optimised for specific applications and may help technology companies manage computing costs, performance and energy consumption. The strategy also allows major customers to diversify their hardware supply chains as AI workloads continue to expand.
Marvell’s relationship with major technology companies has become increasingly important to its growth plans. In August, the company disclosed an agreement involving Alphabet’s Google that could generate as much as $120 billion in sales through fiscal 2033 if specified performance milestones are achieved.
The company’s latest projections indicate that management believes the AI infrastructure market could remain a significant growth engine for years.
Marvell is forecasting fiscal 2031 revenue of between $70 billion and $90 billion. At the midpoint of $80 billion, the projection is substantially higher than the current Wall Street estimate of about $46.85 billion.
The scale of the forecast illustrates how dramatically expectations for the semiconductor sector have changed since the beginning of the AI boom.
Companies that supply processors, networking equipment and data-centre components have benefited from technology firms increasing capital expenditure to support AI services. The construction and expansion of data centres require large quantities of specialised hardware, while the increasing complexity of AI workloads is creating demand for faster connections between computing systems.
Marvell’s strategy is not limited to processing chips. Its portfolio also includes technologies that allow data to move efficiently between different parts of a data centre.
High-speed connectivity is becoming particularly important as AI systems use large clusters of processors. If computing units cannot exchange data quickly enough, overall system performance can be restricted even when the processors themselves are highly capable.
This has made networking and interconnect technologies an increasingly important part of the AI hardware market.
The company has also been expanding beyond dependence on individual customers. Its custom silicon business serves major technology companies, while its broader product portfolio gives it exposure to several parts of the data-centre infrastructure market.
Marvell’s latest projections suggest that management expects this diversification to support sustained growth rather than a temporary increase linked to one AI product cycle.
The financial market response highlighted investor enthusiasm surrounding AI-related semiconductor businesses. Marvell’s shares have risen sharply during 2026, reflecting expectations that demand for data-centre infrastructure will remain strong.
The company had already raised its full-year revenue outlook in August, increasing it to approximately $18 billion from $16.5 billion. The latest long-term forecast represents another significant increase in expectations.
However, the semiconductor industry remains highly competitive. Marvell operates alongside companies involved in different parts of the AI chip ecosystem, including Nvidia and Broadcom. Technology firms are also investing heavily in their own hardware programmes, creating both opportunities and challenges for chip suppliers.
The growing preference for custom silicon could gradually alter the balance within the AI hardware market. Instead of relying entirely on standard processors, major technology companies may increasingly use combinations of commercial products and chips designed specifically for their platforms.
For semiconductor suppliers, this creates opportunities to provide highly specialised solutions while maintaining long-term relationships with large customers.
For cloud providers and technology companies, custom chips can provide greater control over infrastructure design. They can be developed around particular AI models, workloads or software environments, potentially improving efficiency and reducing dependence on external hardware.
The development is also closely connected to the rising cost of operating AI systems. Advanced models require substantial computing resources, and companies are under pressure to make those systems more efficient as usage increases.
Energy consumption is another major consideration. Data centres already require significant amounts of electricity, and AI workloads can increase those requirements substantially. Hardware that improves computing efficiency can therefore become strategically valuable.
Marvell’s outlook reflects the broader transformation of data centres from traditional computing facilities into specialised AI infrastructure hubs.
These facilities increasingly combine high-performance processors, advanced networking, optical systems, storage technologies and specialised cooling and power infrastructure.
The growth of AI applications in areas such as search, cloud computing, enterprise software and digital services is expected to continue driving investment in this infrastructure.
Marvell’s management therefore sees the current AI cycle as a long-term technology shift rather than a short-lived demand surge.
The company’s projected revenue range for fiscal 2031 remains an ambitious target, and actual results will depend on customer demand, competition, technology development and broader economic conditions.
Still, the decision to raise the 2028 forecast demonstrates the company’s confidence in the underlying market.
As AI moves from experimental applications into increasingly widespread commercial use, demand for the hardware required to run these systems is becoming one of the most important forces shaping the semiconductor industry.
Marvell’s latest projections provide another indication that the AI boom is not only creating demand for model developers and software companies but is also reshaping the economics of the global chip and data-centre markets.