Second only to NVIDIA, Marvell Tech becomes the second-highest chip company in AI share

"Optical interconnects, Ethernet switch chips, active cables (AEC), and custom ASICs have given rise to the golden growth acceleration period of Marvell Tech."

Driven by the AI supercycle, Marvell has emerged as the second-highest chip company globally in terms of AI revenue share, only behind Nvidia.

Under the impetus of the AI supercycle, Marvell has rapidly risen to become the second-highest chip company globally in terms of AI revenue share, thanks to its technological advantages in optical interconnects, Ethernet switch chips, active cables (AEC), and custom ASICs, second only to Nvidia.

According to the latest financial report, Marvell's revenue in the AI sector grew by 98% year-on-year and 25% quarter-on-quarter, demonstrating strong growth momentum. Particularly driven by AI custom silicon and high-speed interconnect solutions, Marvell not only meets the high-performance demands of hyperscale data centers for AI training and inference but has also solidified its market position through long-term cooperation agreements with giants like AWS.

The company continues to push for technological innovation, from launching the world's first 3nm 1.6Tbps PAM4 optical DSP to showcasing PCIe Gen 7 connectivity technology, further enhancing its competitiveness in the AI infrastructure market. The arrival of the AI supercycle has brought unprecedented opportunities for Marvell, and its diversified product portfolio and rapidly expanding AI market share make it another leading enterprise in the AI chip field after Nvidia, with significant growth potential ahead.

Marvell Tech's AI Revenue Share Rapidly Soars

1. Key Points and Highlights from the Financial Report

Revenue and Profit: Third-quarter revenue was $1.516 billion, exceeding the guidance midpoint by $66 million, a year-on-year increase of 7% and a quarter-on-quarter increase of 19%. Non-GAAP earnings per share (EPS) were $0.43, a quarter-on-quarter increase of 43%, exceeding the guidance midpoint of $0.03. Gross margin: Non-GAAP gross margin was 60.5%, slightly below guidance, mainly due to the increased revenue share of custom silicon products. Operating cash flow: $536 million, a quarter-on-quarter increase of $230 million.

Key Business Performance:

• Data center business revenue reached a record high of $1.1 billion, a year-on-year increase of 98% and a quarter-on-quarter increase of 25%.
• Strong growth in AI-related revenue drove the data center business to become the main contributor to the company's revenue, accounting for 73%.
• Demand for electro-optics products and custom silicon products exceeded expectations, driving overall performance beyond forecasts.

Customer Cooperation and Industry-Leading Technology: Signed a five-year multi-generational cooperation agreement with Amazon AWS, covering various data center semiconductor solutions, including custom AI products, optical DSPs, and Ethernet switch chips. Launched the world's first 3nm 1.6T DSP chip, further consolidating its technological leadership in the optical market II. Summary of Key Business Segments and Performance

Data Center Business: Revenue of $1.1 billion, a year-on-year increase of 98% and a quarter-on-quarter increase of 25%. Main drivers: rapid growth of custom AI silicon products, strong demand for optical products, and growth in Ethernet switch chips and interconnect products. Revenue from the data center business is expected to grow by 20%-25% quarter-on-quarter in the fourth quarter.

Enterprise Networking and Carrier Business:

• Enterprise networking revenue: $151 million; carrier revenue: $85 million.
• Quarter-on-quarter growth of 4%, with expected double-digit percentage growth in the fourth quarter.
• Significant signs of recovery, but still below end-market consumption levels.

Consumer Market: Revenue of $97 million, a quarter-on-quarter increase of 9%. A quarter-on-quarter decline in the mid-double-digit percentage is expected in the fourth quarter, mainly affected by seasonal weakness in gaming demand.

Automotive and Industrial Market: Revenue of $83 million, a quarter-on-quarter increase of 9%. A quarter-on-quarter growth in the low to mid-single-digit percentage is expected in the fourth quarter.

III. Business Outlook for Next Quarter and Full Year

Fourth Quarter Guidance:

• Revenue: Expected to be $1.8 billion, with a fluctuation of 5%, a year-on-year increase of 26%.
• Non-GAAP gross margin: Expected to be 60%.
• Non-GAAP earnings per share: Expected to be $0.54-$0.64.

Full Year Outlook:

• AI Revenue: Expected to significantly exceed the previous full-year target of $1.5 billion.
• Full-year revenue growth: Expected quarter-on-quarter growth of 19% in the fourth quarter, with a significant increase in full-year revenue growth rate.
• Orders and Reservations: Strong order performance, with demand consistently exceeding expectations. There is a sufficient order volume for custom silicon products and optical products, and supply chain capabilities have been ensured to meet customer growth demands.

I. Highlights of Marvell's AI Revenue Growth

Rapid increase in AI revenue share: Marvell's AI-related revenue achieved a quarter-on-quarter growth of 25% in FY25 Q3, nearly doubling year-on-year. According to management disclosures, the share of AI revenue in the company's overall revenue has significantly increased and is expected to exceed 50% in the coming years.

Explosion of data center business: Data center business revenue grew by 98% year-on-year and 25% quarter-on-quarter, with AI-related products being the core driver. Strong demand for AI custom silicon and optical interconnect products has driven rapid revenue growth.

Market size and potential: Marvell expects the AI-related market TAM (Total Addressable Market) to reach $75 billion, with the custom silicon market size at $40 billion. The company's technological leadership in AI custom silicon and optical interconnect fields enables it to rapidly expand its market share II. Marvell's Core Competitiveness in the AI Field

Custom Silicon: Custom silicon is the core hardware for AI training and inference. Marvell has become a leader in this field through long-term collaborations with hyperscale customers such as AWS and Meta. The company optimizes the performance and power consumption of AI workloads by supporting custom chips with a 3-nanometer process, meeting customers' demands for high efficiency.

Optical Interconnects: Optical interconnect technology is key to achieving high-speed data transmission in AI data centers. Marvell's 1.6T PAM DSP chip is a market leader. Hyperscale customers are accelerating the adoption of optical interconnect technology, further driving Marvell's revenue growth.

Ethernet Switch Chips: The demand for high-speed network switching in AI training clusters has surged, and Marvell's Ethernet switch chips provide higher bandwidth and lower latency, making them highly favored by customers.

Diverse Customer Base: In addition to the GPU market dominated by NVIDIA, Marvell has successfully established a strong customer base in the AI custom silicon field through collaborations with customers like AWS and Meta. The addition of the third-largest customer (Customer C) further enhances the company's market coverage.

III. AI Revenue Share Second Only to NVIDIA

NVIDIA's Dominance in AI: NVIDIA's GPUs nearly monopolize the AI training and inference market, with its AI-related revenue share exceeding 80%, making it the absolute leader in the industry.

Marvell's Rapid Rise: Although Marvell's AI revenue share is not as high as NVIDIA's, it has surpassed competitors such as AMD, Intel, and Broadcom. The strong growth of AI custom silicon and optical interconnects is the core reason for the increase in its AI revenue share.

Comparison with Other Chip Companies:

• AMD: Despite launching AI accelerators (such as MI300), its AI revenue share is far lower than Marvell's.
• Intel: Its layout in the AI market is relatively scattered and has not yet formed a strong revenue source.
• Broadcom: Primarily focused on network chips and ASICs, but its AI revenue share is still lower than Marvell's.

IV. Future Growth Outlook

Continued Growth in AI Revenue: Marvell expects significant growth in AI-related revenue in FY25 Q4, becoming the core driver of overall revenue growth. Management stated that the growth rate of AI revenue will significantly exceed that of the company's other business segments.

Technological Innovation and Market Expansion: Through technological leadership (such as 3-nanometer process, 1.6T optical DSP) and customer expansion, Marvell will further increase its AI market share in the coming years. The rapid expansion of the Total Addressable Market (TAM) in emerging areas (such as active cables and Ethernet switch chips) also creates additional growth potential for the company Expected Revenue Proportion (Next 3-5 Years)

The change in this revenue proportion will make Marvell more focused on high-growth, high-profit core markets while maintaining a certain level of diversified revenue sources.

With its technological breakthroughs in AI custom silicon and optical interconnects, as well as long-term cooperation agreements with hyperscale customers, Marvell has become the second-highest chip company in terms of AI revenue proportion, only behind NVIDIA. As the AI supercycle continues to advance, Marvell's position in the AI market will be further solidified, and it is expected to achieve a higher revenue proportion and market share in the coming years. This marks the rapid rise of Marvell's importance in the global AI chip market.

Marvell Tech's Golden Growth Accelerator Portfolio

The AI supercycle is driving rapid growth in the data center semiconductor market, with the company expecting the AI-related market's Total Addressable Market (TAM) to reach $75 billion, of which the custom silicon market size is $40 billion.

The rapid expansion of TAM in emerging fields such as optical interconnects, Ethernet switch chips, and Active Electrical Cables (AEC) creates significant growth potential for the company.

In-depth analysis of the AI supercycle and the TAM expansion in emerging fields:

1. The AI supercycle drives rapid growth in the data center semiconductor market

Market size and growth potential, with the AI-related market's TAM reaching as high as $75 billion and the custom silicon market size exceeding $40 billion.

Custom silicon (ASIC) is the key technology to meet the high-performance demands of AI workloads, with the customization needs of hyperscale customers (such as AWS, Meta, etc.) becoming the core driving force for market growth.

The data center semiconductor market is experiencing an AI-driven supercycle, with the rapid growth of AI training and inference demands driving an exponential increase in computing power requirements.

Growth Drivers:

• Surge in AI training and inference demands: The complexity of AI models is continuously increasing (e.g., the parameter scale of the GPT-4 model), requiring more efficient hardware support. Data centers are beginning to shift towards customized silicon solutions to optimize the performance and power consumption of AI workloads.
• Long-term cooperation with hyperscale customers: The company's multi-generational cooperation agreements with customers like AWS ensure sustained demand for custom silicon products.
• Technological leadership: Marvell's 3nm process and 1.6T optical DSP technology meet the stringent requirements for high performance and low latency in AI data centers Future Outlook: AI-related revenue is expected to account for over 70% of the company's total revenue in the next 2-3 years, becoming the core driver of long-term growth. Custom silicon and optical interconnect products will continue to dominate the technological upgrades and revenue growth in the data center market.

II. TAM Expansion in Optical Interconnects, Ethernet Switch Chips, and Active Electrical Cables (AEC)

Optical Interconnects: TAM expansion to $20 billion

• AI training and inference require processing massive amounts of data, significantly driving the demand for optical DSP and high-speed interconnect solutions.
• Hyperscale data center customers are accelerating the adoption of optical interconnect technology to replace traditional electrical interconnects.
• The explosive growth of traffic in data centers raises higher demands for high-bandwidth, low-latency optical interconnect solutions.
• Marvell's optical DSP technology (such as 1.6T PAM DSP) can support higher data transmission rates, adapting to future data center needs.

Ethernet Switch Chips, TAM expected to reach $15 billion within five years

• The distributed computing characteristics of AI workloads require faster network switching capabilities.
• Marvell's Ethernet switch chips support higher throughput and lower latency, catering to AI and cloud computing scenarios.
• The data center network architecture is shifting from traditional three-tier architecture to a flattened architecture (such as CLOS architecture), leading to a surge in demand for high-performance Ethernet switch chips.

Active Electrical Cables (AEC), TAM expansion: expected to reach $3 billion within five years

• AEC technology is widely used in AI training clusters and storage networks, becoming a complementary solution to optical interconnects.
• Marvell has a significant competitive advantage in the AEC market, leveraging its electro-optical interconnect technology accumulation.
• AEC replaces traditional copper cables in short-distance high-speed interconnects, providing higher bandwidth and lower power consumption.

III. Growth Potential Analysis in Emerging Fields

Strategic Significance of Optical Interconnects:

• Technological Leadership: Marvell launched the world's first 3-nanometer 1.6T DSP chip, significantly reducing power consumption and enhancing performance.
• Market Demand: Optical interconnect technology is key to supporting the scaling of AI models and distributed computing, with market demand expected to continue growing.
• Customer Base: The rapid adoption by hyperscale customers brings stable revenue growth to the company.

Market Potential of Ethernet Switch Chips:

• Technological Upgrade: Marvell's Ethernet switch chips support throughput of up to 800Gbps, meeting the high-speed data exchange needs within AI clusters.

• Market Expansion: As AI workloads become more prevalent, the demand for high-performance network switch chips in data centers will continue to rise Application scenarios of active cables:

• Short-distance interconnection: AEC has significant advantages in short-distance, high-speed data transmission scenarios and is an effective complement to optical interconnection.

• Cost and performance balance: AEC offers lower costs and higher energy efficiency, making it suitable for large-scale deployment.

IV. Marvell's Strategic Advantages and Competitiveness

• Technological leadership: Marvell's technological accumulation in optical DSP, Ethernet switch chips, and custom silicon ensures its leading position in the market. The world's first 1.6T PAM DSP and the application of 3-nanometer technology further consolidate the company's market competitiveness.
• Diversified product portfolio: The company's layout in custom silicon, optical interconnection, Ethernet switch chips, and AEC forms a complete solution that can meet the diverse needs of data center customers.
• Customer base and partnerships: Long-term cooperation agreements with hyperscale customers (such as multi-generational cooperation with AWS) create stable revenue sources for the company. The diversification of the customer base enhances revenue stability and risk resistance.

V. Future Growth Outlook

• Continued expansion of AI-related business: The AI supercycle will drive long-term growth in the data center semiconductor market, with the company's AI-related revenue expected to significantly exceed the FY25 target.
• Rapid growth of optical interconnection and Ethernet switch chips: Optical interconnection and Ethernet switch chips will become core technologies for data center network upgrades, with market demand expected to maintain high growth.
• Popularization of active cables: AEC, as a complementary solution to optical interconnection, will achieve broader applications in short-distance high-speed interconnection scenarios.

VI. Potential Challenges and Risks

• Market competition: The market for optical interconnection and Ethernet switch chips is highly competitive, with competitors (such as Broadcom and NVIDIA) potentially accelerating technological iterations. The custom silicon market faces technological and price competition from other manufacturers.
• Macroeconomic uncertainty: A global economic slowdown may affect customers' capital expenditures, negatively impacting market demand.
• Speed of technological breakthroughs: The market's demand for technological upgrades is accelerating, and the company needs to continuously invest in R&D to maintain technological leadership.

The expansion of the Total Addressable Market (TAM) for the AI supercycle and emerging fields (optical interconnection, Ethernet switch chips, active cables) creates significant growth potential for Marvell. With technological leadership, a diversified product portfolio, and long-term partnerships with hyperscale customers, the company is well-positioned in these areas. However, market competition and macroeconomic uncertainty remain key risk points that the company needs to monitor. Through continuous technological innovation and customer expansion, Marvell is expected to maintain strong growth momentum in the coming years.

Author of this article: BayesCrest, Source: The Beauty of Bayesian, Original title: "Second only to NVIDIA, Marvell becomes the chip company with the second highest AI proportion"

Risk warning and disclaimer

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not consider individual users' specific investment goals, financial conditions, or needs. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances Invest based on this information at your own risk