Marvell Raises AI Chip Targets

Original editorial illustration representing the expansion of custom AI chips, data-center connectivity and global capital investment.

Marvell Raises AI Chip Revenue Targets Through 2031

Marvell Technology has dramatically raised its long-term revenue ambitions as demand for custom artificial-intelligence chips and data-center connectivity accelerates. At its investor day in New York on October 6, 2026, the semiconductor company said it expects fiscal 2028 revenue of approximately $20 billion, above the roughly $18.2 billion analysts had been expecting. Marvell also introduced a much larger fiscal 2031 revenue range of $70 billion to $90 billion. The announcement sent Marvell shares about 6% higher and lifted rival Broadcom as investors reassessed the potential size of the custom-silicon market. The company says its strategy, developed around custom and cloud-optimized chips since 2021, is benefiting from hyperscalers building their own AI processors rather than relying exclusively on Nvidia.

Marvell Bets On A Much Larger AI Chip Market

Marvell's investor-day announcement represents one of the most aggressive growth forecasts from a major semiconductor company this year.

The company raised its fiscal 2028 revenue target to approximately $20 billion, compared with its previous outlook of about $18 billion. The new target is also above the approximately $18.2 billion consensus estimate compiled by LSEG.

More strikingly, Marvell expects fiscal 2031 revenue to reach between $70 billion and $90 billion.

At the midpoint of $80 billion, that would put Marvell far above the roughly $46.85 billion fiscal 2031 revenue estimate cited by Reuters from Visible Alpha analysts.

The forecast is not a guarantee of future performance. It is management's current outlook based on expected demand, existing programs and future design opportunities.

Custom AI Chips Are Driving The Forecast

Marvell's strategy is different from the business model of companies focused primarily on general-purpose AI accelerators.

Its custom-silicon business helps large technology companies develop processors tailored to their own workloads and infrastructure.

That market is becoming increasingly important as hyperscalers seek greater control over AI computing costs and performance.

Companies operating enormous data centers have an economic incentive to develop specialized processors when the volume of their computing workloads is sufficiently large.

A custom chip can potentially be optimized for particular workloads, memory configurations, networking systems or power requirements.

Marvell has positioned itself as one of the major semiconductor companies providing the engineering and manufacturing ecosystem required for those projects.

Marvell Raised Its Custom Chip Target Again

Marvell said it expects its custom-chip business to generate approximately $12 billion in fiscal 2029 revenue.

That is up from the company's previous target of approximately $10 billion.

The increase is important because custom silicon is becoming one of the company's principal growth engines.

The market is expanding as large technology companies attempt to reduce their reliance on a single accelerator supplier and create chips designed specifically for their own computing architectures.

That does not necessarily mean Nvidia is losing its leadership.

Instead, it suggests the overall AI-chip market may be becoming more diversified, with Nvidia's general-purpose accelerators competing alongside custom processors designed by individual hyperscalers.

Google Could Become A Huge Marvell Customer

One of the most important developments supporting Marvell's long-term outlook is its relationship with Alphabet's Google.

Marvell disclosed in August that a deal with Google could generate up to approximately $120 billion in sales through fiscal 2033 if specified performance milestones are achieved. 

The potential size of that arrangement helps explain why investors are taking Marvell's custom-chip strategy seriously.

Google has invested heavily in its own AI accelerators, known as TPUs, as it seeks to optimize computing for its AI and cloud businesses.

Marvell's role in such programs demonstrates how semiconductor companies can participate in AI growth without selling a conventional GPU product directly to every customer.

Hyperscalers Want More Control Over AI Computing

The economic logic behind custom chips is straightforward.

Large cloud companies operate enormous fleets of servers. At that scale, even relatively small improvements in performance, power consumption or utilization can translate into significant financial savings.

Developing a custom processor requires substantial engineering investment, but the economics can become attractive when the chip is deployed across millions of computing workloads.

AI has increased that incentive because processors represent one of the largest costs in advanced computing infrastructure.

Companies also want greater control over their technology roadmaps.

Instead of waiting for a general-purpose processor vendor to define the next generation of hardware, a hyperscaler can design a chip around its own model architectures and software stack.

Nvidia Still Sets The Market Benchmark

Marvell's growth does not mean the AI-chip market is moving away from Nvidia.

Nvidia remains one of the industry's most important suppliers of AI accelerators and networking technology.

Its software ecosystem and integrated hardware platforms make its products attractive to organizations that want a broadly supported AI-computing environment.

Custom chips address a different part of the market.

Large technology companies with enough scale can justify the cost of designing specialized processors, while smaller organizations may prefer standardized accelerators.

The result could be a more segmented AI semiconductor industry rather than a simple winner-takes-all market.

Marvell's Stock Reaction Shows Investor Confidence

Marvell shares rose approximately 6% after the investor-day announcement, while Broadcom shares gained roughly 4%, according to Reuters.

The market reaction indicates that investors interpreted Marvell's forecasts as evidence of stronger-than-expected demand for AI infrastructure.

Marvell's shares have already more than tripled in 2026, according to Reuters.

That performance means expectations are already high.

When a stock has risen sharply, investors typically require continued evidence that earnings and revenue can catch up with the valuation.

Marvell therefore has a significant execution challenge ahead.

The 2031 Forecast Is Extremely Ambitious

The $70 billion to $90 billion fiscal 2031 revenue range represents a major expansion from Marvell's existing business.

The midpoint of the forecast is approximately $80 billion.

For comparison, Marvell generated approximately $8.2 billion in fiscal 2026, according to market reporting around the investor day.

That means the company is effectively projecting a transformation from a semiconductor business measured in single-digit billions of annual revenue to one approaching the scale of the world's largest chip companies.

Such a trajectory would require sustained demand across several business lines and successful execution on numerous customer programs.

It also assumes that the AI infrastructure boom continues creating opportunities for custom silicon and high-speed connectivity.

Connectivity Is Another Major Growth Engine

Marvell's opportunity is not limited to custom AI processors.

Modern AI data centers require extremely fast networking and interconnect technologies because thousands or even hundreds of thousands of processors may need to communicate efficiently.

AI workloads can generate enormous quantities of data that must move between processors, memory systems and storage.

As computing clusters become larger, networking performance can become a critical factor in overall system efficiency.

This creates opportunities for semiconductor companies that provide optical connectivity, networking chips and related infrastructure.

Marvell has invested heavily in these areas, giving the company multiple ways to benefit from data-center expansion.

The AI Data Center Is Becoming A System, Not A Chip

One of the most important changes in the semiconductor industry is that AI infrastructure is increasingly being designed as an integrated system.

Processors are only one component.

Modern AI facilities also require:

• High-bandwidth networking.
• Optical interconnects.
• Advanced memory systems.
• High-capacity storage.
• Power-management components.
• Specialized cooling infrastructure.
• Software optimized for distributed computing.

Marvell's strategy places it across several of these infrastructure layers.

That diversification could reduce its dependence on the success of a single chip category.

AI Is Changing The Economics Of Semiconductor Design

Historically, developing custom silicon was largely limited to companies with enormous resources.

The AI boom is changing the economics because the potential workload volume has become so large.

A hyperscaler that operates massive AI infrastructure can spread development costs over a huge installed base.

That makes custom chips more financially attractive.

Marvell's business model benefits from this trend because it can provide the technology expertise required to create these processors without every customer having to build a semiconductor design organization from scratch.

Marvell's Strategy Began Before The Current AI Boom

The company has emphasized that its custom and cloud-optimized silicon strategy dates back to 2021.

That timing is significant because it means Marvell began building the capabilities needed for today's AI infrastructure market before the latest wave of generative AI investment reached its current scale.

Reuters reported that company executives attribute the current position partly to years of work developing the technology and customer relationships required for custom silicon.

This illustrates an important feature of the semiconductor industry: major technology opportunities often require years of preparation before they become visible in revenue.

Google Is Not The Only Potential Customer

Google's potential $120 billion arrangement is particularly significant, but the broader market includes other hyperscalers and technology companies developing proprietary AI infrastructure.

Cloud providers are competing to offer AI services at lower cost and higher performance.

Custom silicon can become part of that competition.

If several major cloud providers adopt specialized processors, semiconductor companies such as Marvell could gain multiple large revenue streams.

The diversification would also reduce the risk associated with dependence on one customer.

Why Custom Chips Could Matter For AI Costs

AI computing is expensive because advanced processors consume significant amounts of electricity and require sophisticated supporting infrastructure.

Improving performance per watt can therefore have a direct economic benefit.

A custom processor can be designed around the exact workload it is expected to perform.

That may allow a company to reduce unnecessary functionality and optimize the chip for specific operations.

The financial benefit becomes larger as the processor is deployed at massive scale.

For hyperscalers operating global data centers, even a small efficiency improvement can potentially translate into substantial savings over time.

Marvell's Forecast Depends On Execution

The largest uncertainty surrounding the new targets is execution.

Marvell must turn design wins into production volumes.

It must meet performance requirements set by customers.

It must navigate complex semiconductor manufacturing supply chains.

And it must continue investing in new generations of technology as AI architectures evolve.

A design win does not always translate into the maximum possible revenue originally expected.

Customer requirements can change, competing technologies can emerge and projects can be delayed.

Marvell's management therefore has to convert a large pipeline of potential business into actual shipments and recurring revenue.

The Semiconductor Cycle Still Matters

AI has created a powerful source of demand, but Marvell remains exposed to the broader semiconductor cycle.

Chip companies can experience significant swings in inventory, pricing and capital expenditure.

AI infrastructure may reduce some of that cyclicality if hyperscaler spending remains strong, but it does not eliminate the risk.

Large customers can also adjust capital expenditure rapidly if economic conditions change.

That makes Marvell's long-term forecast dependent not only on AI adoption but also on the willingness of customers to continue investing billions of dollars in data centers.

AI Infrastructure Spending Is Becoming More Concentrated

Another issue for investors is customer concentration.

A relatively small number of hyperscalers account for a substantial share of global data-center investment.

This can create enormous opportunities for semiconductor suppliers, but it can also make them dependent on the capital-spending plans of a few companies.

If a major cloud provider delays a new data-center program, semiconductor suppliers can feel the effect through reduced orders or delayed shipments.

Marvell's ability to serve multiple hyperscalers is therefore strategically important.

Marvell's Full-Year Outlook Had Already Improved

The latest investor-day announcement follows another increase in Marvell's outlook in August.

At that time, the company raised its full-year revenue forecast to approximately $18 billion from $16.5 billion.

The repeated upward revisions suggest that management sees stronger demand than it previously expected.

However, investors will eventually judge those forecasts through actual quarterly results.

The ability to consistently meet or exceed guidance will be critical if Marvell wants to maintain the market's confidence in its long-term targets.

What The Marvell Forecast Means For Broadcom

Broadcom's share price also rose following Marvell's announcement.

That reaction may appear counterintuitive because Marvell is expanding its custom-chip business in an area where Broadcom is also a major competitor.

But investors may interpret Marvell's forecast as evidence that the overall custom-silicon market is expanding rapidly enough to support several major suppliers.

In other words, the market may be large enough for multiple companies to grow simultaneously.

Broadcom also has a much broader semiconductor and infrastructure portfolio, so its exposure to AI extends beyond custom processors.

What Investors Should Watch Next

• Custom-chip revenue: The path toward Marvell's $12 billion fiscal 2029 target will be an important test of its strategy.
• Google production milestones: The potential $120 billion relationship depends on achieving specified performance milestones.
• Data-center spending: Continued capital expenditure by hyperscalers will determine the size of the addressable market.
• Gross margins: Revenue growth must eventually translate into profitable growth.
• Customer diversification: Additional major design wins could reduce dependence on individual hyperscalers.
• AI networking demand: Rapid growth in cluster size could increase demand for Marvell's interconnect technologies.
• Competition: Broadcom, Nvidia and other semiconductor suppliers will continue competing for the same infrastructure spending.

Marvell's New Targets At A Glance

Metric	Latest Marvell Outlook	Previous/Market Reference
Fiscal 2028 revenue	Approximately $20 billion	Analyst consensus about $18.2 billion
Fiscal 2029 custom-chip revenue	Approximately $12 billion	Previous target about $10 billion
Fiscal 2031 revenue	$70 billion–$90 billion	Analyst estimate about $46.85 billion
Fiscal 2031 midpoint	$80 billion	Significantly above current analyst expectations
Potential Google sales	Up to $120 billion through fiscal 2033	Subject to performance milestones

The Bigger AI Semiconductor Shift

Marvell's forecast points to a broader change in the AI semiconductor industry.

The first phase of the generative-AI boom was dominated by demand for general-purpose accelerators, particularly Nvidia GPUs.

The next phase is becoming more diversified.

Hyperscalers are designing custom processors.

Cloud companies are developing proprietary architectures.

Networking requirements are expanding alongside processor demand.

And semiconductor suppliers are increasingly selling complete infrastructure components rather than individual chips.

That shift could create a much larger addressable market for companies such as Marvell.

The Key Question Is Whether AI Spending Stays Elevated

Marvell's targets ultimately depend on one major assumption: that AI infrastructure investment remains strong for years.

So far, hyperscalers have continued announcing large data-center and AI infrastructure programs.

But the spending cycle is still relatively young compared with traditional infrastructure markets.

Investors will eventually want evidence that the computing capacity being built today is generating sufficient revenue and productivity improvements to justify further investment.

If that happens, Marvell's custom-chip market could expand rapidly.

If AI infrastructure spending slows before the company's major programs reach full production, its long-term revenue targets would become harder to achieve.

Why The Forecast Matters For The Global AI Economy

Marvell's announcement is significant beyond its own stock.

It provides another indication of how much money technology companies are prepared to invest in AI computing.

Custom processors require years of engineering and billions of dollars of infrastructure investment.

Companies would not pursue them at this scale unless they expected AI workloads to remain strategically important for a long period.

The forecast also shows that the AI boom is spreading deeper into the semiconductor supply chain.

The opportunity is no longer limited to the companies designing the most visible AI models or accelerators.

Chip designers, networking suppliers, optical-component companies and data-center infrastructure providers are all becoming important beneficiaries of the same investment cycle.

A High-Expectation Moment For Marvell

Marvell has now given investors an unusually ambitious view of where it believes the company can be by fiscal 2031.

The opportunity is enormous, but so is the execution requirement.

To reach even the bottom of the $70 billion to $90 billion range, Marvell would need sustained growth across custom silicon, networking and connectivity businesses.

Its Google relationship provides a potentially significant foundation, while the wider adoption of proprietary AI processors creates additional opportunities.

But the company's forecasts remain forward-looking expectations rather than guaranteed outcomes.

The next stage will be measured through customer deployments, production volumes, revenue growth and profitability.

For the global technology market, however, the message is already clear: the AI infrastructure boom is creating an increasingly competitive and valuable market for custom semiconductors, and Marvell believes it is positioned to capture a much larger share of it.

Frequently Asked Questions

What did Marvell announce on October 6, 2026?

At its investor day, Marvell raised its fiscal 2028 revenue forecast to approximately $20 billion and introduced a fiscal 2031 revenue target of $70 billion to $90 billion.

Why is Marvell benefiting from the AI boom?

Marvell supplies custom and cloud-optimized semiconductor technology used in data-center infrastructure. Hyperscalers are increasingly developing specialized AI processors, creating demand for companies that can help design and deliver those chips.

How much custom-chip revenue does Marvell expect?

Marvell expects its custom-chip business to generate approximately $12 billion in fiscal 2029 revenue, up from its previous target of about $10 billion.

What is the significance of Google's relationship with Marvell?

Marvell disclosed in August that its relationship with Alphabet's Google could generate up to approximately $120 billion in sales through fiscal 2033 if specified performance milestones are achieved.

Does Marvell compete with Nvidia?

Marvell and Nvidia participate in different but overlapping parts of the AI semiconductor market. Nvidia is a major supplier of general-purpose AI accelerators, while Marvell focuses heavily on custom silicon and data-center connectivity. Some hyperscalers use both approaches.

Is Marvell's $70 billion–$90 billion 2031 forecast guaranteed?

No. It is a management forecast. Achieving it will depend on customer demand, production execution, semiconductor supply chains, AI infrastructure spending, competition and the successful conversion of design opportunities into revenue.

Why are custom AI chips becoming more important?

Large cloud companies operate computing infrastructure at enormous scale. Custom processors can potentially be optimized for specific workloads, performance requirements and power consumption, making them economically attractive when deployed across sufficiently large data-center fleets.

Sources

• Reuters, October 6, 2026 — Reporting on Marvell's fiscal 2028 and fiscal 2031 revenue forecasts, custom-chip strategy and Google relationship.
• Reuters reporting from Marvell's October 6 investor day — Market reaction and details of the company's custom-silicon and AI infrastructure strategy.
• Reuters, October 6, 2026 — Reporting on the broader market reaction to Marvell's new long-term targets.