HSBC initiates coverage on Marvell with a Hold rating: AI optics supports earnings, while ASIC visibility remains the weak point
AI summary card
HSBC initiates coverage on Marvell with a Hold rating: AI optics supports earnings, while ASIC visibility remains the weak point
The report argues that Marvell faces risks of Amazon share loss and delays in new projects in its ASIC business, but AI optical revenue driven by PAM4 DSP comes in above market expectations, leaving FY27e data center and AI revenue broadly in line with consensus.
- The target price is USD85.00, based on 26x FY27e P/E and FY27e non-GAAP EPS of USD3.34; versus the USD81.32 share price, this implies about 4.5% upside.
- HSBC expects Marvell FY27e ASIC revenue of USD2.0bn, up 12% year over year, below the Visible Alpha consensus of USD2.3bn and the expected 26% growth rate.
- AI optics is the main offset: HSBC expects FY27e optical revenue of USD2.5bn, up 38% year over year, 14% above the consensus of USD2.2bn.
- FY27e AI revenue is expected to be USD4.5bn, in line with consensus; FY27e data center revenue is expected to be USD7.1bn, up 18% year over year and also in line with market expectations.
- Relative to peers, Marvell shares are down 26% year to date, while Broadcom is up 53% and the S&P 500 is up 13%, reflecting the market's preference for companies with a clearer ASIC roadmap.
Report interpretation
Overview
This is HSBC Global Investment Research's initiation report on Marvell Technology. The core view is that Marvell is becoming an important AI semiconductor player, with the business driven by both ASIC and AI optics, but the quality of the two tracks differs. The total addressable market for ASIC is expanding rapidly, but there is uncertainty around Marvell's share of Amazon's Trainium project, the timing of Microsoft Maia production, and CoWoS supply capacity. By contrast, the AI optics business is benefiting from demand for 800G and 1.6T PAM4 DSP, and its revenue resilience is stronger than market expectations. Therefore, HSBC assigns a Hold rating rather than a more aggressive one.
Core views
The report's core views are as follows: first, upward revisions to hyperscaler capex are a proxy for ASIC demand growth, and ASIC's share of that capex is expected to rise from 2% in 2023 to 13% in 2027e. Second, Marvell's own ASIC roadmap lacks visibility; most of Amazon Trainium 2 and Trainium 2.5 is handled by Amazon's internal Annapurna Labs, while Trainium 3 is expected to be led mainly by Alchip. Third, Microsoft Maia-related projects may slip to at least 2027e because HSBC believes Marvell will not have sufficient CoWoS capacity in 2026e. Fourth, although the AI optics business faces competition from Broadcom in 800G and 1.6T DSP, TAM expansion is sufficient to support revenue above consensus. Fifth, Marvell's weakness in ASIC and strength in AI optics largely offset each other, leaving FY27e AI revenue and data center revenue close to market expectations.
Analysis framework
The report uses segment revenue decomposition, peer comparison, and valuation multiple analysis. On the revenue side, it assesses ASIC, AI optics, and data center businesses separately and compares them with Visible Alpha consensus; on the competitive side, it focuses on differences among Marvell, Broadcom, and Alchip in ASIC customers, project visibility, and revenue growth; on the valuation side, it derives the target price by multiplying FY27e non-GAAP EPS by the target P/E multiple.
Methodology notes
Use hyperscaler capex as a proxy for ASIC demand growth
The report argues that continuous upward revisions to cloud providers' capex reflect the intensity of AI infrastructure investment and can serve as a leading indicator for ASIC demand growth.
Estimate the net impact of ASIC downside and AI optics upside on total AI revenue separately
HSBC splits Marvell's AI revenue into ASIC and AI optics, and judges whether the shortfall from ASIC being below consensus can be offset by optical revenue driven by PAM4 DSP coming in above consensus.
Calculate the target price by multiplying the target P/E multiple by FY27e EPS
The USD85.00 target price comes from applying a 26x FY27e target P/E to FY27e non-GAAP EPS of USD3.34; 26x is roughly the three-year historical average P/E.
Assess Marvell's relative position using ASIC revenue growth, customer visibility, and share performance
The report believes Broadcom has a clearer ASIC roadmap, while Alchip has secured the main share of Amazon Trainium 3, giving both stronger growth prospects than Marvell.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- MARVELL TECHNOLOGY INC (MRVL.US)coverage target, company equity
- Strengths
- AI optics has market leadership, 800G PAM4 DSP demand is strong, and 1.6T PAM4 solutions are beginning to ship to multiple customers; data center and AI revenue are expected to be in line with consensus.
- Weaknesses
- Low visibility on the ASIC roadmap, possible loss of Amazon Trainium share, potential delays in new projects such as Microsoft Maia, and FY27e ASIC revenue forecast below consensus.
- Comparison
- Compared with Broadcom and Alchip, Marvell is weaker in ASIC revenue growth, customer wins, and share performance; its 24x FY27e P/E is below Broadcom's and Alchip's, which the report sees as justified.
- Risks
- If DSP market share is lost rapidly or ASIC projects are delayed further, earnings could come under pressure; if new ASIC customers ramp earlier than expected or PAM4 DSP adoption is faster than expected, there is upside risk.
- Broadcommain competitor and valuation reference
- Strengths
- Has better ASIC roadmap visibility, is expected to benefit more from AI ASIC TAM expansion, and is attracting multiple new customers.
- Weaknesses
- The report does not focus on Broadcom's own downside factors.
- Comparison
- Broadcom is up 53% year to date, significantly outperforming Marvell's -26%; its FY26e P/E is around 30x, higher than Marvell's.
- Risks
- For Marvell, Broadcom's product launches in 800G and 1.6T DSP will intensify competition in AI optics.
- AlchipASIC competitor and peer reference
- Strengths
- The report believes Alchip has won the main share of Amazon Trainium 3, with faster growth than Marvell.
- Weaknesses
- The report does not focus on Alchip's own operating weaknesses.
- Comparison
- Alchip is expected to deliver materially higher revenue and EPS growth in 2026e than Marvell and has the upper hand in Amazon's next-generation ASIC project.
- Risks
- For Marvell, Alchip's expanding Trainium share will dilute revenue growth from its largest ASIC customer.
Key data
- RatingHoldHSBC initiates coverage and assigns a Hold rating.
- Target priceUSD85.00Based on 26x FY27e P/E and FY27e non-GAAP EPS of USD3.34.
- Current share priceUSD81.32As of the close on 2025-11-19.
- Implied upside+4.5%Limited upside versus the current share price.
- FY27e ASIC revenue forecastUSD2.0bnUp 12% year over year, 10% below the Visible Alpha consensus of USD2.3bn.
- FY27e AI optics revenue forecastUSD2.5bnUp 38% year over year, 14% above consensus of USD2.2bn.
- FY27e AI revenue forecastUSD4.5bnIn line with Visible Alpha consensus.
- FY27e data center revenue forecastUSD7.1bnUp 18% year over year, in line with market expectations and recent management commentary.
- ASIC share of hyperscaler capexfrom 2% in 2023 to 13% in 2027eThe report sees 2026e as an important inflection point for ASIC penetration.
- Share performance comparisonMarvell -26% y-t-d; Broadcom +53%; S&P500 +13%Reflects Marvell's relative disadvantage in ASIC roadmap visibility.
Impact & implications
For investors, Marvell is not a pure laggard in the AI semiconductor theme, because its AI optics business still provides strong revenue support. However, the lack of visibility into its ASIC customer share, supply capacity, and timing of new projects limits the scope for valuation re-rating. The report is better suited to support a neutral positioning view: if the AI optics TAM continues to expand, downside risk to earnings is limited; if the market continues to price primarily off the ASIC roadmap, Marvell may continue to lag Broadcom and Alchip.
Risks
- Upside risk includes new ASIC customers ramping earlier than expected.
- Upside risk includes adoption of the latest-generation PAM4 DSP being faster than expected.
- Downside risk includes Marvell losing DSP market share rapidly.
- Downside risk includes delays in ASIC projects, especially postponement of Microsoft Maia-related projects.
- Downside risk includes further diversion of Amazon Trainium business share to Annapurna Labs or Alchip.
- Broadcom catching up in 800G and 1.6T DSP could weaken Marvell's leadership in AI optics.
What to watch
- The actual division of labor and share changes among Marvell, Alchip, and Annapurna Labs in Amazon Trainium 3.
- Whether the Microsoft Maia AI accelerator-related ASIC project contributes revenue on a 2027e timeline.
- Whether Marvell can secure sufficient CoWoS capacity to support new ASIC customer ramps.
- The sustainability of 800G PAM4 DSP demand and the pace of 1.6T PAM4 DSP adoption.
- The impact of Broadcom's 800G and 1.6T DSP products on Marvell's market share.
- Whether hyperscaler capex expectations continue to be revised upward and whether ASIC's share of capex rises as expected.
- Whether FY27e AI revenue of USD4.5bn and data center revenue of USD7.1bn remain in line with consensus.