Marvell stock price hits a historic high then pulls back: Analysis of the three major growth engines of the AI connectivity layer leader

As of early June 2026, Marvell Technology (MRVL) stock reached an intraday high of $324.20, setting a new record, with a market capitalization briefly surpassing $250 billion. The direct catalyst driving this surge was NVIDIA CEO Jensen Huang publicly stating at Computex Taipei on June 2, 2026, that Marvell is poised to become the next tech giant with a market value exceeding one trillion dollars. This statement quickly ignited market sentiment, causing the stock price to soar sharply in the short term.

Prior to this, Marvell released a record FY2026 full-year financial report in March 2026: annual revenue of $8.2B, up 42% year-over-year; GAAP net profit of $2.67 billion. In late May 2026, Q1 FY2027 revenue further reached $2.42B, up 28% year-over-year, again hitting a new high. As of the close on June 10, 2026, MRVL stock was at $252.59, showing a significant pullback from its historical peak.

However, after the market sentiment soared, a more critical question is: Is Marvell’s growth driven by short-term event-driven pulses, such as NVIDIA’s endorsement, or is it based on the sustained industry logic of the AI infrastructure supercycle?

From Financial Data to Growth Quality: AI Demand Is No Longer Just a “Story”

Marvell’s overall financial performance for FY2026 (ending February 1, 2026) provides the most direct anchor for understanding its current market position. Full-year revenue of $8.2B, up 42%, slightly above FactSet analysts’ consensus estimate of $8.18B; net profit of $2.67 billion; adjusted EPS of $2.84, up 81% year-over-year. The fourth quarter alone saw data center revenue of $1.22 billion, up 22% year-over-year.

Marvell Technology FY2026 Core Financials and Key AI Indicators

| Indicator | Data | | --- | --- | | FY2026 Full-Year Revenue | $8.2B (new high) | | YoY Growth | 0.42 | | Full-Year Net Profit | $2.67 billion | | Gross Margin | 51.0% (up 9.7 percentage points YoY) | | Q4 Data Center Revenue | $1.65 billion (75% of total) | | FY2027 Revenue Guidance | $11.0 billion (up 34% YoY) |

Data center business constitutes the main driver of growth. Q4 data center revenue reached $1.65 billion, up 9% sequentially and 21% year-over-year. Chairman and CEO Matt Murphy pointed out in the earnings report that “robust AI demand” is the core driver of performance growth.

Looking at the business structure, management disclosed in a public speech in September 2025 that approximately 50% of Marvell’s data center revenue comes from optical interconnect products, about 25% from custom silicon chips, and the remaining 25% from data center storage, switching, and security emerging product lines. This structure naturally aligns Marvell’s growth logic along two main lines: first, mass production and volume ramp-up of AI custom chips; second, ongoing upgrades in bandwidth within AI data centers.

Custom AI Chips: Riding the Next Wave of Computing Power Beyond GPUs

Marvell’s position in the AI ASIC market is shifting from “potential beneficiary” to “actual beneficiary.” According to publicly available data, Marvell holds about 20% to 25% of the global custom AI ASIC design services market, forming a de facto duopoly with Broadcom, together accounting for roughly 95% of the market. Marvell’s clients include Amazon’s Trainium and Inferentia chips, and it also supports Microsoft, Meta, and Google’s Arm-based CPU businesses.

This market segment is expanding rapidly. According to market research firm TrendForce, ASIC’s share in AI server markets will jump from 20.9% in 2025 to 27.8% in 2026, while GPU’s share will decline from 75.9% to 69.7%. The increase in ASIC’s share does not imply a shrinking demand for GPUs; rather, as the overall AI acceleration chip market expands quickly, customized solutions are gaining higher priority among cloud service providers due to their efficiency and cost advantages under specific workloads.

Evolution of AI Server Acceleration Chip Market Structure (ASIC vs GPU)

From this logic, the current situation where data center custom silicon accounts for about a quarter of Marvell’s FY2026 revenue is essentially the early stage of growth. The company’s target revenue range for AI custom chips is $9 billion to $11 billion. The growth’s verifiability rests on two levels: first, existing large client projects (Amazon Trainium/Inferentia, etc.); second, the trend that cloud providers’ capital expenditure on ASICs is expected to rise from 2% in 2023 to 13% in 2027.

Optical Interconnects: The “Main Artery” of Data Center Business

Optical interconnect products account for about 50% of Marvell’s data center revenue, making it the company’s core foundation. The growth logic of this business is based on the technological evolution of AI data center architecture: as AI training clusters expand from thousands of cards to hundreds of thousands, the communication bandwidth and data transfer efficiency between chips become critical bottlenecks.

Marvell offers a complete product matrix in optical interconnects, including DSPs, retimers, and silicon photonics engines. The silicon photonics engines support transmission speeds up to 6.4T, capable of integrating multiple components in compact modules, and are compatible with pluggable optical modules and co-packaged optical solutions. Management expects optical products to achieve double-digit sequential growth in Q3 FY2026.

In December 2025, Marvell announced the acquisition of optical interconnect startup Celestial AI for approximately $3.25 billion, with a deal comprising $1 billion in cash and about 27.2 million Marvell common shares. The transaction is expected to close in Q1 2026. Celestial AI specializes in photonic transmission technology, with core capabilities in using optical signals instead of electrical signals to achieve high-speed interconnects between AI chips and memory. Murphy stated that this acquisition is a transformative step in Marvell’s development, expanding its leadership in AI connectivity, as “scaling” has become the next frontier in AI infrastructure.

Logically, this acquisition makes sense because bottlenecks in AI infrastructure are evolving through a “sequential emergence and breakthrough” path. Before reaching the limits of computing power, the bottleneck is in processing capacity; once decoupled, data throughput becomes the bottleneck. By acquiring Celestial AI, Marvell effectively secures a strategic first-mover position in optical interconnect technology.

NVIDIA CEO Jensen Huang also explicitly expressed this view at the same event: he called his collaboration with Marvell in optical technology a “match made in heaven,” predicting that Marvell will become the next trillion-dollar tech company. While this is a personal judgment from an industry leader rather than investment advice, it sends a clear industry signal: the status of optical interconnects in the AI infrastructure value chain is being revalued.

Competitive Landscape and Valuation Logic: Two Sides Requiring Caution

Any analysis of Marvell cannot ignore its comparison with Broadcom. Broadcom holds about 70% of the AI ASIC market share, with a market cap of approximately $2.3 trillion, significantly ahead in scale and customer coverage. Broadcom expects its AI ASIC market to grow from about $12.2 billion in 2024 to $60–90 billion by 2027, and secured key orders from OpenAI in 2025.

In contrast, some institutions believe Marvell’s growth trajectory is weaker than management suggests, mainly because Broadcom’s ASIC roadmap offers stronger predictability. Additionally, over 70% of revenue is concentrated in the Asian market, posing risks, and the gross margin of custom chip business may face competitive pressures—these are factors requiring ongoing attention. These issues do not have easy solutions and need to be dynamically tracked with each earnings report.

Marvell vs Broadcom — Dual Oligopoly in Custom AI ASICs

| Dimension | Broadcom | Marvell | | --- | --- | --- | | Combined market share in AI ASIC design | Approximately 95% (monopoly/duopoly) | Approximately 95% (duopoly) | | Market share percentage | Leading, roughly 55–70% | Second, roughly 13–25% | | Key clients | Google (TPU v7), OpenAI (deep cooperation on 10GW) | Amazon (Trainium), Microsoft (Maia) | | Main differentiated competitive fields | AI network switching chips | Optical interconnects, data center DSPs |

Institutional ratings for Marvell are generally being upgraded, but target prices vary widely. As of late May 2026, 38 analysts’ average 12-month target price was about $208.64, with 32 buy ratings and 6 hold ratings. Following the sharp stock price increase, many institutions raised their targets: Stifel raised from $230 to $321; Benchmark to $275; KeyBanc to $260; Raymond James from $105 to $235. Citigroup’s adjustment was the largest, from $118 to $215. The broad range of target prices reflects differing market views on Marvell’s valuation rather than a consensus.

The stock experienced significant volatility in the days following Computex, dropping over 16% from a record $316 to around $263. This fluctuation was not due to fundamental changes but a natural correction after rapid gains—rising from $219.43 on June 2 to a peak of $324.15 on June 4, nearly a 50% increase in three days. Such volatility warns that nonlinear movements driven by extreme sentiment are difficult to sustain.

Leveraging Gate Platform for Global Tech Asset Allocation

For investors interested in Marvell and other US tech stocks, Gate’s recent launch of US stock trading offers a valuable participation channel. Users can directly trade over 10,000 stocks and ETFs listed on NYSE and NASDAQ using USDT within their Gate accounts, without switching platforms or transferring funds, enabling one-stop cross-market asset allocation within crypto accounts.

Gate’s US stock trading is based on compliant brokerage channels offering real stock spot trading, distinct from the platform’s original CFD products. Funds are managed separately, with zero holding costs, no rollover or overnight fees, and dividends paid directly in USDT to the account. For users familiar with crypto trading, Gate’s interface for stocks mirrors that of crypto trading, lowering the learning curve for cross-market investments.

Taking Marvell as an example, investors can directly trade MRVL shares on Gate to participate in its AI growth dividends; additionally, they can flexibly adjust positions during market phases, achieving efficient cross-market asset management. Against the backdrop of increasing global capital market linkage and restructured tech stock valuation logic, this trading approach offers long-term value beyond short-term arbitrage.

Conclusion

Marvell Technology, with a record $8.195 billion in revenue and 42% YoY growth in FY2026, demonstrated the quantifiable impact of AI demand on its performance. Its three main business pillars—custom AI chips, optical interconnects, and data center networking—correspond to different nodes in the AI computing value chain, with logical synergy rather than substitution.

However, high gains come with high volatility, especially for tech companies that have expanded their market cap multiple times in a short period. Marvell faces competitive pressures (notably from Broadcom), customer concentration risks, and long-term trends in gross margins for custom chips—none of which can be answered easily. These variables require continuous tracking and reassessment with each earnings cycle, rather than relying on event-driven hype or industry leader comments.