Marvell (Minutes): AI Revenue Poised to Account for Half of Total Revenue​

Marvell Technology Marvell (MRVL.O) released its Q1 FY2026 earnings report (ending April 2025) after the US stock market closed on May 30, Beijing time:

Below is the transcript of Marvell's Q1 FY2026 earnings call. For earnings interpretation, please refer to《Marvell: AI Growth 'Stalls', Where's the Next 'Trump Card'?》

1. $Marvell Tech(MRVL.US) Key Financial Information Review

1. Cash Flow: Operating cash flow of $333 million; returned $52 million to shareholders through cash dividends, and repurchased $340 million in stock (compared to $200 million last quarter).

2. Q2 Guidance:

a. Revenue: Expected around $2 billion, ±5%.

b. Profit: GAAP gross margin 50%-51%, Non-GAAP gross margin 59%-60%; GAAP diluted EPS $0.16-$0.26, Non-GAAP diluted EPS $0.62-$0.72.

c. Expenses: GAAP operating expenses approximately $735 million, Non-GAAP operating expenses approximately $495 million; other income and expenses approximately $49 million, non-GAAP tax rate 10%.

d. Shares: Basic diluted weighted average shares outstanding 864 million, diluted weighted average shares outstanding 874 million.

2. Detailed Content of Marvell's Earnings Call

2.1 Key Information from Executive Statements

1. Business Dynamics: Sold automotive Ethernet business to Infineon for $2.5 billion in cash, expected to close in 2025, enhancing capital allocation flexibility.

2. Data Center Market:

a. Q1 Performance: Revenue of $1.44 billion, up 5% QoQ, up 76% YoY.

b. Q2 Outlook: QoQ growth expected in the mid-single digits (around 5%), maintaining strong YoY growth. Optimistic about the long-term potential of the data center business, benefiting from hyperscale enterprise capital expenditure, sovereign data demand, and emerging market expansion.

c. Technology Layout:

- Promoting mass production of AI chips and shipment of optoelectronic products.

- Developing custom HBM (High Bandwidth Memory) and Co-Packaged Optics (CPO) technology to optimize AI accelerator performance, driving the transition from copper interconnects to fiber optic interconnects.

- Collaborating with NVIDIA to introduce NVLink Fusion technology, launching a multi-chip package platform to support customer-customized XPU projects, which has entered the production stage.

3. Enterprise Network and Carrier Infrastructure Market:

a. Q1 Performance: Total revenue of $306 million (Enterprise Network $168 million + Carrier $138 million), up 14% QoQ.

b. Q2 Outlook: Total revenue QoQ growth expected in the mid-single digits (around 5%), continuing recovery. Carrier and enterprise network market revenue expected to grow QoQ for the fifth consecutive quarter, maintaining growth expectations amid macroeconomic uncertainty.

4. Consumer Electronics Market:

a. Q1 Performance: Revenue of $63 million, down 29% QoQ.

b. Q2 Outlook: QoQ growth of about 50%, mainly driven by seasonality and gaming demand.

5. Automotive and Industrial Market:

a. Q1 Performance: Revenue of $76 million, down 12% QoQ (automotive growth offset by industrial decline).

b. Q2 Outlook: Overall revenue flat QoQ.

6. Future Outlook:

a. AI has already accounted for the majority of new data center business, and custom business (such as XPU projects) will drive further growth in AI-related revenue in the coming years. The XPU project in collaboration with a large US hyperscale customer is progressing smoothly, with A0 Ethernet control chip mass production achieved, and the next-generation product will use 3nm process, expected to be in production by 2026.

b. A custom AI investor event will be held on June 17, focusing on custom chip market opportunities, technology platforms (such as system design, advanced packaging), and multi-generation collaboration advantages.

2.2 Q&A Session

Q: Regarding the 3nm XPU project, is the direction of the next-generation project to increase, decrease, or remain the same? Does your company have exclusivity on this project, or is it expected to be shared with competitors?

A: Content Direction: The next-generation project is progressing, with 3nm wafer and advanced packaging capacity secured, planned for production in 2026, expected to grow custom chip revenue through project iteration as AI capital expenditure increases.

Exclusivity: Customers typically work with multiple partners to build chip portfolios, and given the increase in XPU shipments, customers may adopt multiple approaches to meet demand, so exclusivity is not present.

Q: How does Marvell view its ability to support a broader customer base?

A: Capability Foundation: The company has expansion capabilities in engineering and R&D, has continuously increased R&D investment over the past few years, and has refocused talent and teams on data center and AI fields through capital allocation strategies.

Current Progress and Plans: Has expanded cooperation in supporting multiple projects, will detail related projects at the AI investor event; the customer base is not limited to leading enterprises, and emerging hyperscale customers can also benefit from the company's technology, with the ability to further drive company growth in the future.

Q: There is much discussion about Marvell's positioning in 200G SerDes technology. Can you introduce your company's layout in this technology? Also, how is the cooperation with NVIDIA progressing?

A: 200G SerDes Technology (Ethernet Transmission Speed) Layout: The technology performance and market timing are both best in class, with 200G products already in mass production and performing excellently. Demonstrated the first single-channel 400G at the OFC exhibition, with an aggressive technology roadmap, maintaining leadership in both electrical SerDes and Optical SerDes technologies.

Progress with NVIDIA: First, the company is deeply integrated into the ecosystem with key partners; second, the market has validated the complementarity of customized solutions, and NVIDIA needs to meet customer needs for self-developed XPU while leveraging its own rack-level solution R&D investment. The company is involved and helps achieve customer goals, viewing it as a key part of providing end-to-end product portfolios to customers, involving customized HBM, co-packaged optics technology, advanced packaging, and other fields.

Q: Can you break down data center revenue, and is the AI business share this quarter still similar to last quarter (about 55%)? It is speculated that custom chips grow due to slightly lower gross margins. Also, is this year's AI revenue far exceeding the $2.5 billion target, with the actual annual scale roughly in the $3.5 - $4 billion range?

A: Data Center Breakdown: Will not break down data center revenue by quarter, previously stated that AI has become the main part of data center revenue. Although the specific share and timing cannot be clearly stated, with the recovery of the mass market and core business, it is expected that in the near future, AI will not only be the main revenue source for the data center terminal market but will also become the main revenue for the holding company, even accounting for 50% of the company's total revenue.

AI Revenue: No specific response to this year's AI revenue scale.

Q: Is the slightly lower gross margin due to the rapid growth of customer custom business?

A: In the business mix changes within the quarter, the gross margin of custom business is lower than the company's average level, which does affect the quarterly gross margin. The custom business will continue to grow strongly in Q2 and the second half of the year, and its interaction with other businesses will determine the gross margin in the second half of the year, currently expected to be in a range similar to the Q2 guidance.

Q: Is custom business expected to continue growing in the second half of the year? Will enterprise-level business grow slowly and steadily? Can you supplement the relevant information on business expectations for the second half of the year?

A: Custom Business Growth: Yes.

Enterprise-Level Business: The core business (including enterprise network and carrier business) is experiencing a strong recovery, expected to continue to recover and grow for the rest of the year.

Overall Expectations for the Second Half of the Year: Focus on macro dynamics, expect overall business growth for the year, continued demand for AI and data centers, and core business recovery will lay the foundation for FY27.

Q: Investors hope that the growth rate of the data center business will be faster than the mid-single-digit and guidance levels for the next quarter. Can you explain the growth differences between AI business and non-AI business (such as on-premises deployment, etc.), and elaborate on the specific growth rate of AI business?

A: Overall Growth of Data Center Business: In past quarters, data center business grew over 20% from Q3 to Q4, with current quarter growth of 4%, expected to grow in mid-single digits subsequently, with Q2 growth slightly faster than Q1. YoY, data center revenue grew over 70%, confident in the growth trajectory.

AI vs. Non-AI Business Differences: AI business is the fastest-growing part of the data center, currently already over half in the data center, and AI revenue will account for half of the company's total revenue in the future.

AI Business Growth Rate: No specific figures given.

Q: Can you talk about the situation of on-premises deployment business or other changing businesses?

A: The on-premises deployment business is currently small in scale, with a slight drag on data center revenue, but the overall impact is limited. Slight decline in Q1, relatively stable overall, future contribution declining, may maintain status quo or slightly rebound. Data center revenue is mainly driven by AI, with other businesses remaining flat; enterprise on-premises deployment due to spending structure shifting to AI.

Q: The guidance for data center business is mid-single-digit growth, and custom chip growth is strong. Is there a timing issue with the optical module business? Can you explain the current status of the optical module business and its consideration in the guidance?

A: As for the optical business, it is expected to continue growing throughout the year. Both custom business and optical business are showing growth trends. The strong rebound in consumer business is seasonal and in line with expectations. However, it should be noted that from a full-year perspective, this business is usually slightly below average.

Q: Regarding the cable and other component products mentioned in the cooperation with Amazon, what is the current impact scale on the company? How important is this business compared to competitors' larger data?

A: The cooperation with Amazon in the network and connectivity field is progressing smoothly, involving multiple products, but there are no major updates, and no specific scale data mentioned.

No direct comparison with competitor data, emphasizing that new category products in the cooperation (such as AEC accelerated Ethernet controllers) will gradually enter mass production this year, driving growth in the switching field, with the business overall showing a positive development trend.

Q: Leading GPU suppliers are preparing to launch next-generation platforms supporting 1.6T optical connections and Ethernet expansion solutions. Typically, optical modules are shipped before GPUs. Is it expected that the 1.6T optical module solution will grow strongly from now on? The 5nm solution has been certified by customers, will the 3nm DSP solution be integrated into the next-generation network platform?

A: 1.6T Optical Module Growth: 5nm process 1.6T optical modules have begun delivery, with strong current demand but limited initial scale. As 800G optical modules still dominate demand throughout the year, large-scale growth of 1.6T is expected to start next year.

3nm Solution Integration: Strong demand for 3nm products, smooth product development progress, the ecosystem is driving the upgrade to 3nm (based on power consumption advantages), and the company is ready for the application of 1.6T optical modules in the next-generation platform.

Q: The company once set a specific gross margin level for the ASIC business. Is this level still reasonable? Looking ahead to the next-generation products, how does the company view the sustainable gross margin of the AI ASIC business? Will it remain in the original expected range?

A: Current Gross Margin Level: The overall custom business (including ASIC) gross margin can still be controlled within the original discussion range, despite the expansion of business scale, management capabilities have not been affected.

Internal Differences in Custom Business: Within the custom business (XPU), the higher the volume, the lower the gross margin, but operating profit will increase significantly; non-XPU business (such as custom network equipment using IP, accelerators, etc.) has slightly higher gross margins, and sales exceed expectations, the overall mix has not changed the original gross margin range.

Next-Generation Product Outlook: The AI ASIC business (such as XPU) gross margin will remain at the original lower level, but the growth of high-margin non-XPU business (such as custom network products) can optimize the overall structure, and the company is confident in the gross margin, with management strategies unchanged.

Operating Profit Margin (Management Supplement): All business projects are driving the operating profit margin to the model expected level, with project scale expanding but limited investment, leverage effects have been reflected in past quarters. The increase in absolute operating profit and significant EPS growth (EPS growth rate is about twice the revenue), expected to continue due to leverage effects of large projects, operating profit is considerable.

Q: Is it possible for ASIC projects to adopt a dual-supplier model? If adopted, is there a guarantee that projects designed by the company over several years will not be eliminated by the end of the year due to customers choosing other suppliers?

A: Possibility of Dual Suppliers: Given the scale of the project, the possibility of seeking multiple suppliers is not ruled out.

Project Assurance: Confident in the continuity of revenue from initial projects and the growth of revenue from new projects, production and execution are progressing as planned. Although there is external noise, the focus is on customer needs and self-execution capabilities, without directly mentioning specific assurance measures against elimination risks.

Q: NVIDIA has opened NVLink, will your company design SerDes to be compatible with NVLink and provide it as an IP module to ASIC customers? Also, does the NVLink fusion product involve NVIDIA licensing your company's IP?

A:SerDes Compatibility and IP Module: Plans to adopt a chiplet solution, connecting XPU and NVLink chips through I/O technology (input/output), did not explicitly mention whether self-developed SerDes is directly compatible with NVLink or provided as an IP module, but emphasized cooperation with customers to design XPU and connect to the NVLink ecosystem.

IP Licensing Issue: Did not directly respond to whether NVIDIA IP licensing is involved, only stated that the cooperation focuses on the open ecosystem concept (allowing customers to leverage existing infrastructure investments), has gained customer interest, specific models to be observed.

Q: During the upgrade process from 800G to 1.6T optical modules, what is the company's market share? What impacts might the technology transition bring? Based on visibility, what is the current inventory level of customers?

A: Market Share: Maintains a leading position in the 800G field, despite new competitors entering, the position is solid; 1.6T products have been launched first (5nm process), promoting 400G I/O technology to support 3.2T, market share will not undergo fundamental changes, execution speed and product ramp-up capabilities strengthen the leading advantage; new technologies (such as Linear Pluggable Optics LPO) have a gradual impact, mainly used for extended scenarios, the company will participate in the layout based on existing technology, but its market share is expected to be small.

Inventory Level: Customer distribution inventory is extremely low, strong sales last quarter, optical module business expected to continue growing throughout the year and next year (especially during the technology transition period), inventory risk is controllable.

Q: For customers exploring multiple paths, does the company expect their demand to lean towards performance-oriented product projects, or is it possible for hyperscale customers to adopt completely different ASIC projects for completely different workloads?

A: More details (such as different business models, diverse customer needs) are expected to be further discussed on AI Day.

<End Here>

Risk Disclosure and Statement of This Article:Dolphin Research Disclaimer and General Disclosure