Marvell Technology has emerged as the semiconductor industry’s breakout story of 2026. The custom AI chip maker’s stock has surged 50% year-to-date, with a 30% gain in April alone, as Wall Street wakes up to a company that quietly became the backbone of hyperscale AI infrastructure. With record fiscal 2026 revenue of $8.195 billion, a Bank of America upgrade to Buy, and a deepening partnership with Nvidia in silicon photonics, Marvell is no longer flying under the radar. Here is why analysts are calling it the most undervalued name in AI semiconductors, and what the rally means for the broader chip market.

Marvell’s Record Fiscal 2026 Earnings: $8.195 Billion in Revenue

Marvell Technology reported its fiscal year 2026 results on March 5, 2026, delivering numbers that exceeded Wall Street expectations across every key metric. Full-year net revenue hit $8.195 billion, a 42% year-over-year increase and a new company record. The fourth quarter alone generated $2.219 billion in revenue, surpassing guidance by $19 million and representing a 22% year-over-year gain.

The earnings quality was equally impressive. GAAP gross margin came in at 51.7%, while non-GAAP gross margin reached 59.0%. GAAP net income for the full year was $2.670 billion, translating to $3.07 per diluted share. Non-GAAP earnings per share hit $2.84, an 81% increase from the prior year. Fourth-quarter non-GAAP EPS landed at $0.80, a penny above the $0.79 consensus estimate from 12 analysts.

“Marvell delivered record fiscal 2026 revenue of $8.195 billion, driven by exceptional data center growth,” CEO Matt Murphy said during the earnings call. “Design wins hit an all-time record, with bookings accelerating at a record pace into fiscal 2027.”

The data center segment was the star performer, with Q4 data center revenue reaching a quarterly record of $1.65 billion. For the full fiscal year, data center revenue exceeded $6 billion, making it by far the largest contributor to the company’s top line. Management guided fiscal Q1 2027 revenue to $2.4 billion, an 8% sequential increase, and projected full fiscal year 2027 revenue at approximately $11 billion, a 30% jump from the record just set.

The 50% Stock Rally: What Is Driving MRVL in April 2026

Marvell Technology stock has climbed 50% since January 2026, making it one of the best-performing semiconductor names of the year. In April alone, MRVL rose 30% and closed in the green for six out of seven active trading sessions, according to Stocktwits data. The stock hit prices near $95 by late March, up from the $48.09 low recorded over the past 52 weeks.

👁 The 50% Stock Rally: What Is Driving MRVL in April 2026

Several catalysts converged to drive the rally. First, the blowout Q4 2026 earnings on March 5 sent shares spiking roughly 20% in a single session, from $75.68 to approximately $90. Second, Bank of America upgraded Marvell from Neutral to Buy on March 7, raising its price target from $90 to $110. Analyst Vivek Arya cited accelerating custom silicon momentum and the company’s expanding addressable market in AI infrastructure.

“Marvell is in the early stages of a strong multiyear growth cycle,” Arya wrote in his upgrade note. “The custom silicon opportunity alone could drive revenue well beyond current Street estimates.”

Barclays followed with an Overweight rating in early April, adding further institutional endorsement. The broader semiconductor sector also provided a tailwind, with the Philadelphia Semiconductor Index hitting fresh 2026 highs as both Broadcom and Marvell led the sector rally. Zacks Research highlighted both names as breakout leaders, noting double-digit percentage gains in early April driven by AI chip deals and networking growth.

The 52-week range of $48.09 to $133.20 shows just how volatile the stock has been, but the current trajectory suggests the market is repricing Marvell as a core AI infrastructure play rather than a cyclical semiconductor name.

Custom AI Chips: How Marvell Became the Hyperscaler’s Secret Weapon

At the heart of Marvell’s transformation is its custom silicon business. Unlike Nvidia, which sells standardized GPU accelerators, Marvell designs bespoke AI chips tailored to the specific workloads of individual hyperscale cloud providers. These custom XPUs (accelerator processors) allow companies like Amazon, Google, and Microsoft to optimize their AI infrastructure for cost, power efficiency, and performance.

Marvell’s custom silicon revenue has been growing at an extraordinary pace. The company reported record design wins in fiscal 2026, with CEO Murphy noting that bookings are “accelerating at a record pace” heading into fiscal 2027. Analysts project Marvell’s custom silicon business could reach approximately $1 billion in annual revenue within the next two years, with CXL memory interconnect attach rates expected to double annually.

The hyperscaler strategy is straightforward: as Amazon builds out its Trainium and Graviton chips, as Google scales its TPU architecture, and as Microsoft develops its Maia AI accelerators, each of these programs requires specialized chip design expertise. Marvell provides the silicon design, packaging, and interconnect technology that makes these custom programs possible.

“What makes Marvell unique is their ability to serve as the design partner for multiple hyperscalers simultaneously,” said Stacy Rasgon, senior semiconductor analyst at Bernstein Research. “They are building a portfolio of custom programs that collectively represent a massive and growing revenue opportunity.”

Management has guided fiscal 2028 revenue to approximately $15 billion, up from an earlier $13 billion outlook, with custom silicon as the fastest-growing driver. That implies the company expects to nearly double its revenue in just two years.

Silicon Photonics and the Nvidia Partnership

One of the most significant catalysts for Marvell’s April rally was a reported $2 billion investment from Nvidia focused on silicon photonics collaboration. Silicon photonics technology uses light instead of electrical signals to transfer data between chips and across data center racks, enabling dramatically higher bandwidth at lower power consumption.

Marvell’s electro-optics business has been positioned as a critical enabler for the next generation of AI data centers. As GPU clusters scale to tens of thousands of interconnected chips, the networking fabric that connects them becomes a major bottleneck. Traditional copper-based interconnects simply cannot keep up with the bandwidth demands of large-scale AI training workloads.

Marvell’s PAM4 and coherent DSP (Digital Signal Processor) technology provides the ultra-high bandwidth, low-latency connections that modern AI fabrics require. The company’s co-packaged optics (CPO) solutions integrate optical transceivers directly into switch packages, reducing power consumption and latency while increasing port density.

“The optical interconnect market is experiencing explosive growth,” said Rick Schafer, semiconductor analyst at Oppenheimer. “Marvell’s photonics business connects thousands of GPU chips in data centers, and as cluster sizes grow, this becomes an increasingly critical piece of the infrastructure puzzle.”

Marvell has also made strategic acquisitions to strengthen its photonics capabilities. The purchases of Celestial AI and XCON Technologies have enhanced its AI interconnect portfolio, providing both the silicon and the system-level expertise needed to compete in this rapidly growing market segment.

Marvell vs. Broadcom: The Custom AI Chip Market Landscape

Marvell’s primary competitor in the custom AI chip space is Broadcom, which has built a dominant position designing custom accelerators for Google’s TPU program and other hyperscalers. Broadcom’s ASIC deals generally command higher margins and larger contract values, but Marvell is closing the gap with a broader portfolio of interconnect and networking solutions.

👁 Marvell vs. Broadcom: The Custom AI Chip Market Landscape

Metric	Marvell Technology (MRVL)	Broadcom (AVGO)
Fiscal 2026 Revenue	$8.195 billion	$56.1 billion (projected FY2025 ending Oct)
YoY Revenue Growth	42%	44% (AI segment)
Data Center Revenue	$6+ billion	$8.4 billion (AI revenue, Q4 FY2025)
Non-GAAP Gross Margin	59.0%	77.5%
Stock YTD Gain (2026)	~50%	~35%
Custom AI Chip Focus	XPU design, interconnect, photonics	ASIC design, networking switches
Key Hyperscaler Clients	Amazon, Microsoft, Meta	Google, Meta, ByteDance
Analyst Consensus Rating	Buy (BofA $110 PT)	Buy

The key differentiator is scale and margin. Broadcom’s AI revenue surged 106% year-over-year to $8.4 billion in its most recent quarter, dwarfing Marvell’s data center segment. However, Marvell’s 42% revenue growth rate and expanding design win pipeline suggest the gap could narrow significantly over the next two fiscal years.

Both companies benefit from a secular shift: hyperscalers are increasingly designing their own custom AI chips rather than relying solely on Nvidia’s off-the-shelf GPUs. This “custom silicon” trend is creating a massive addressable market for chip design partners like Marvell and Broadcom, with industry analysts estimating the custom AI chip TAM could reach $30 to $50 billion by 2028.

Wall Street Upgrades and Analyst Price Targets

The analyst community has turned decisively bullish on Marvell Technology in 2026. Bank of America’s upgrade from Neutral to Buy on March 7, with a price target increase from $90 to $110, was the highest-profile call. Analyst Vivek Arya highlighted the company’s accelerating custom silicon pipeline and improving margin trajectory as key reasons for the upgrade.

Barclays assigned an Overweight rating in early April, adding to the bullish consensus. The stock’s 50% YTD gain has already surpassed several prior price targets, forcing analysts to revise upward. Consensus estimates for fiscal 2027 project revenue growth in the 32% to 33% range, which would bring annual revenue to approximately $11 billion.

“The risk-reward profile for Marvell has shifted dramatically,” said Chris Caso, semiconductor analyst at Wolfe Research. “With design wins accelerating and the custom silicon TAM expanding faster than expected, this is a name that deserves a premium valuation.”

The forward guidance is what has analysts most excited. Murphy’s projection of $15 billion in fiscal 2028 revenue implies the company expects to nearly double from its current run rate. If Marvell executes on this trajectory, the stock could have significant upside from current levels, even after the 50% rally.

Not everyone is convinced, however. Bears point to Marvell’s lower margins compared to Broadcom, execution risk on multiple simultaneous custom chip programs, and the company’s dependence on hyperscaler capital expenditure cycles. A slowdown in cloud spending could hit Marvell disproportionately, given its concentrated customer base.

The $750 Billion Data Center Capex Boom Fueling the Rally

Marvell’s growth story cannot be understood without the broader context of the AI data center investment boom. According to Bloomberg New Energy Finance (BNEF), capital expenditures by the 14 largest publicly owned data center operators are projected to reach nearly $750 billion in 2026, up from under $450 billion in 2025. Alphabet, Amazon, Meta, and Microsoft alone plan to invest over $650 billion this year.

Gartner estimates $582 billion will be invested globally in AI infrastructure in 2026, a 19% increase from 2025. Over 23 gigawatts of data center IT capacity is currently under construction worldwide, with approximately 75% of that concentrated in the United States. BNEF tracked 831 sites under data center construction as of late 2025.

This unprecedented spending wave directly benefits Marvell in three ways. First, every new AI data center requires custom silicon and interconnect solutions. Second, as cluster sizes grow from thousands to tens of thousands of GPUs, the networking and optical interconnect opportunity scales nonlinearly. Third, the shift toward custom AI chips means hyperscalers need design partners like Marvell to help them build proprietary architectures.

“We are in the early innings of a multiyear infrastructure buildout,” said Matt Murphy during Marvell’s most recent earnings call. “The demand signals we are seeing across our hyperscale customer base are unprecedented in the company’s history.”

Marvell’s Data Center Revenue Breakdown and Growth Trajectory

Period	Data Center Revenue	Total Revenue	DC as % of Total	YoY Growth
FY2024	~$3.5 billion (est.)	$5.77 billion	~61%	Baseline
FY2025	~$4.5 billion (est.)	$5.77 billion	~78%	~29%
FY2026	$6+ billion	$8.195 billion	~73%	~33%
FY2026 Q4	$1.65 billion	$2.219 billion	74%	21% YoY
FY2027 (guided)	~$8 billion (est.)	~$11 billion	~73%	~33%
FY2028 (guided)	~$11 billion (est.)	~$15 billion	~73%	~36%

The data center segment now represents approximately three-quarters of Marvell’s total revenue, up from roughly 61% two years ago. This concentration reflects both the explosive growth of AI-related demand and the relative maturity of Marvell’s other business lines, including carrier infrastructure, enterprise networking, and consumer products.

👁 Marvell's Data Center Revenue Breakdown and Growth Trajectory

The Q4 2026 data center figure of $1.65 billion implies an annualized run rate of $6.6 billion, with management expecting sequential acceleration through fiscal 2027. If the company hits its $11 billion full-year target, data center revenue would need to reach approximately $8 billion, a 33% increase from fiscal 2026 levels.

The Semiconductor Sector Rally: Beyond Marvell

Marvell’s surge is part of a broader semiconductor rally that has seen the Philadelphia Semiconductor Index (SOX) hit fresh 2026 highs in April. Broadcom led the charge with double-digit percentage gains, driven by its own AI chip deals and networking growth. The semiconductor sector has outperformed the S&P 500 year-to-date, reflecting investor confidence in the durability of the AI infrastructure spending cycle.

The rally has been concentrated in names with direct AI exposure. Companies like Marvell, Broadcom, and Nvidia have significantly outperformed semiconductor firms focused on traditional markets like automotive, industrial, or consumer electronics. This divergence highlights a fundamental shift in where semiconductor value is being created.

TSMC, the world’s largest contract chipmaker, reported Q1 2026 revenue of $25.53 billion (NT$892.55 billion), a 35% year-over-year increase, further confirming the strength of AI chip demand. The company’s $56 billion capital expenditure plan for 2026 is the largest in semiconductor history, with the majority directed toward advanced AI chip manufacturing capacity.

For Marvell specifically, the sector tailwind is amplified by its unique positioning. While Nvidia dominates the GPU market and Broadcom leads in custom ASICs, Marvell occupies a differentiated niche at the intersection of custom silicon, optical interconnects, and data center networking. This multi-product approach gives the company multiple vectors for growth as AI infrastructure scales.

Key Risks and the Bear Case Against Marvell

Despite the bullish momentum, there are legitimate risks that investors should consider. Marvell’s non-GAAP gross margin of 59% trails Broadcom’s 77.5% by a wide margin, suggesting the company may be competing on price in certain custom chip programs. The Trainium3 design loss to Alchip, where Marvell’s chiplet-based proposal was reportedly rejected in favor of a monolithic die approach, demonstrates that hyperscaler design wins are far from guaranteed.

Customer concentration is another concern. With data center revenue representing 73% of total sales and a handful of hyperscalers accounting for the bulk of that, Marvell is heavily exposed to changes in cloud capex cycles. A meaningful slowdown in spending by Amazon, Google, or Microsoft could materially impact the company’s growth trajectory.

“The custom silicon business is inherently lumpy,” noted Hans Mosesmann, managing director at Rosenblatt Securities. “Design wins take years to ramp and can shift in timing. Investors need to be prepared for quarters where revenue comes in below expectations, even if the long-term trajectory remains intact.”

Geopolitical risks also loom. The ongoing US-China chip trade tensions, including 25% chip tariffs implemented in 2025, could disrupt supply chains or limit Marvell’s ability to serve certain customers. The company’s fabless model, which relies on TSMC and other foundries for manufacturing, creates dependency on a supply chain that has proven vulnerable to geopolitical disruption.

Valuation is also a concern after the 50% rally. Marvell is now trading at a significant premium to its historical multiples, and any execution misstep could trigger a sharp correction. The 52-week range of $48.09 to $133.20 illustrates just how quickly sentiment can swing in either direction.

5 Predictions for Marvell Technology and the Custom AI Chip Market

Based on the current trajectory and market dynamics, here are five predictions for Marvell and the broader custom AI chip landscape:

👁 5 Predictions for Marvell Technology and the Custom AI Chip Market

1. Marvell will hit $11 billion in fiscal 2027 revenue. Management’s guidance aligns with the accelerating design win pipeline and hyperscaler spending trajectory. With Q1 already guided at $2.4 billion, the company is on pace to deliver.

2. The custom AI chip TAM will exceed $40 billion by 2028. As more hyperscalers develop proprietary silicon, the market for design partners like Marvell and Broadcom will expand faster than current estimates suggest. Amazon, Google, Microsoft, and Meta are all scaling custom chip programs aggressively.

3. Silicon photonics will become Marvell’s fastest-growing segment by fiscal 2028. The Nvidia partnership and strategic acquisitions in optical interconnect technology position Marvell to capture a disproportionate share of the data center networking upgrade cycle.

4. At least two additional hyperscalers will announce new custom chip programs with Marvell by end of 2026. The trend toward custom silicon is accelerating, and Marvell’s track record of successful design partnerships makes it a natural choice for companies looking to reduce their dependence on Nvidia.

5. Marvell’s stock could reach the $120 to $130 range by end of 2026 if execution continues. The combination of revenue acceleration, margin expansion, and multiple analyst upgrades provides a path toward the upper end of the 52-week range. However, any macro or spending slowdown could cap the upside significantly.

Historical Context: Marvell’s Transformation From Storage to AI

Marvell’s rise as an AI infrastructure company represents one of the most dramatic corporate transformations in the semiconductor industry. The company was founded in 1995 and spent its first two decades primarily focused on storage controllers, networking chips, and wireless communications. Under previous leadership, it was considered a second-tier semiconductor firm, overshadowed by larger competitors like Broadcom, Qualcomm, and Intel.

The pivot began when Matt Murphy took over as CEO in 2016. Murphy, a former Maxim Integrated executive, refocused the company on data center infrastructure and began building the custom silicon capabilities that would later become Marvell’s crown jewel. The $10 billion acquisition of Inphi in 2021 was a transformative deal that brought critical electro-optics and interconnect technology into Marvell’s portfolio.

The results speak for themselves. In fiscal 2023, Marvell generated $5.5 billion in revenue. By fiscal 2026, that figure had grown to $8.195 billion, and management is projecting $15 billion by fiscal 2028. The company’s market capitalization has grown in tandem, reflecting the market’s recognition that Marvell has successfully repositioned itself at the center of the AI infrastructure buildout.

What This Means for the Broader AI Chip Market

Marvell’s success signals a structural shift in how AI infrastructure is being built. The era of one-size-fits-all GPU computing is giving way to a more diverse ecosystem where custom silicon, optical interconnects, and specialized networking solutions play equally critical roles. This shift benefits companies like Marvell that can offer the full stack of data center silicon, from custom XPU accelerators to high-speed photonics.

For Nvidia, the rise of custom AI chips is both a challenge and an opportunity. On one hand, every dollar spent on custom silicon is a dollar not spent on Nvidia GPUs. On the other hand, Nvidia’s reported $2 billion investment in Marvell suggests the company sees custom silicon partners as complementary rather than competitive. Large AI clusters will likely use a mix of Nvidia GPUs for general-purpose training and custom accelerators for inference and specialized workloads.

For investors, Marvell’s rally offers a template for identifying the next wave of AI beneficiaries. The first wave was dominated by Nvidia and, to a lesser extent, AMD. The second wave is about the picks-and-shovels plays: the companies that provide the interconnect, networking, and custom design services that make AI data centers function. Marvell sits squarely in this second wave, alongside Broadcom, Arista Networks, and Coherent.

Market Impact and Investment Implications

Marvell’s 50% stock rally has added billions of dollars in market capitalization and attracted significant institutional attention. The stock’s inclusion in multiple semiconductor and technology ETFs means its performance directly influences broader market indices. As of late March 2026, the stock was trading near $95, implying a market capitalization of approximately $82 billion based on its outstanding share count.

👁 Market Impact and Investment Implications

The investment case for Marvell hinges on two critical assumptions: that hyperscaler AI capex will continue growing at current rates, and that Marvell will execute on its expanding pipeline of custom silicon programs. If both hold true, the company’s revenue trajectory from $8.2 billion to $15 billion over three years represents one of the most compelling growth stories in the semiconductor sector.

Institutional ownership has increased meaningfully in Q1 2026, with multiple large fund managers adding to their positions following the earnings beat and Bank of America upgrade. The average daily trading volume has also picked up, with 15.3 million shares changing hands on March 31 alone, reflecting heightened investor interest.

For portfolio allocation purposes, Marvell offers exposure to multiple AI infrastructure tailwinds: custom silicon, optical interconnects, data center networking, and the broader hyperscaler capex cycle. This diversified exposure within the AI theme makes it an attractive complement to pure-play GPU names like Nvidia.

Related Coverage

• TSMC’s $35.71B Q1 2026 Revenue: Inside the 35% Surge and $56B Capex Reshaping the AI Chip Market
• AMD’s MI400 Series: Inside the 320B-Transistor Chip and $7.2B Bet to Break Nvidia’s AI GPU Grip
• NVIDIA Blackwell vs AMD MI350: The Top AI GPU Comparison (2026)
• Broadcom’s AI Revenue Surges 106% to $8.4 Billion: Inside the Custom Chip Empire Reshaping the Semiconductor Industry
• AI Chips 2026: The Guide to the Semiconductor Market
• Big Tech’s $700 Billion AI Infrastructure Bet: Inside the 2026 Spending Race
• NVIDIA’s $4 Billion Silicon Photonics Bet: Inside the Lumentum and Coherent Deals Reshaping AI Data Center Connectivity

Frequently Asked Questions

Why is Marvell Technology stock going up in 2026?

Marvell Technology stock has surged 50% year-to-date in 2026, driven by record fiscal 2026 revenue of $8.195 billion (up 42% YoY), a Bank of America upgrade to Buy with a $110 price target, record custom AI chip design wins, a deepening partnership with Nvidia in silicon photonics, and the broader semiconductor sector rally fueled by AI data center spending.

What does Marvell Technology do?

Marvell Technology designs and sells semiconductor solutions for data centers, including custom AI accelerator chips (XPUs), silicon photonics and optical interconnect technology, data center switches, and networking processors. The company’s primary customers are hyperscale cloud providers like Amazon, Microsoft, and Meta who use Marvell’s chips to build AI infrastructure.

How does Marvell compete with Broadcom in custom AI chips?

Both Marvell and Broadcom design custom AI accelerator chips for hyperscale cloud providers. Broadcom commands higher margins (77.5% vs. 59%) and larger contract values, particularly through its long-standing Google TPU partnership. Marvell differentiates through its broader portfolio of interconnect and networking solutions, including silicon photonics and high-speed optical technology.

What is Marvell’s revenue forecast for fiscal 2027 and 2028?

Marvell management has guided fiscal 2027 revenue to approximately $11 billion, a 30% increase from the $8.195 billion recorded in fiscal 2026. For fiscal 2028, the company projects approximately $15 billion in revenue, implying a near-doubling from fiscal 2026 levels. Data center revenue is expected to remain the primary growth driver.

Is Marvell Technology stock a good investment in 2026?

Analysts are generally bullish, with Bank of America setting a $110 price target and Barclays assigning an Overweight rating. Bulls cite accelerating custom silicon demand, the expanding AI data center market, and Marvell’s unique positioning in optical interconnects. Bears point to lower margins compared to Broadcom, customer concentration risk, and valuation concerns following the 50% rally. Investors should consider their own risk tolerance and investment horizon.

What is silicon photonics and why does it matter for Marvell?

Silicon photonics uses light instead of electrical signals to transfer data between chips and across data center racks. As AI clusters scale to tens of thousands of GPUs, optical interconnects become critical for maintaining bandwidth and reducing power consumption. Marvell’s PAM4 and coherent DSP technology, combined with its co-packaged optics solutions, positions it as a key supplier of this technology to AI data centers. Nvidia’s reported $2 billion investment in Marvell’s photonics capabilities underscores the strategic importance of this business.