Marvell's 2024Q3 Earnings Call: Contradictions in Custom Compute Revenue, Carrier Market Projections, and AI Growth Expectations from 2024Q3 to 2026Q2

Generated by AI AgentEarnings DecryptReviewed byAInvest News Editorial Team
Tuesday, Nov 25, 2025 3:11 am ET4min read
Aime RobotAime Summary

-

reported Q3 revenue of $1.42B, driven by 21% data center growth from AI/cloud demand, exceeding guidance midpoints.

- Enterprise networking declined 28% YoY due to inventory corrections, while carrier infrastructure grew 17% YoY but faces Q4 softness.

- Q4 guidance projects >30% sequential data center growth with AI exit rates >$200M, but carrier revenue expected to drop mid-40% sequentially.

- Custom silicon revenue is concentrated among hyperscalers, with $400M AI optics revenue in 2024 and $800M long-term targets, though margin pressures persist.

- Management emphasized AI/cloud momentum despite near-term enterprise/carrier weakness, with Q4 non-GAAP gross margin guidance at 63.5–64.5%.

Date of Call: None provided

Financials Results

  • Revenue: $1.49B (Willem) / $1.42B (Matt); declined 8% YOY, grew 6% sequentially
  • EPS: Non-GAAP $0.41 per diluted share (Willem/Matt), $0.01 above midpoint; GAAP loss $0.19 per diluted share
  • Gross Margin: GAAP 38.9%; Non-GAAP 60.6%, up 30 bps sequentially
  • Operating Margin: GAAP -10.3%; Non-GAAP 29.8%

Guidance:

  • Q4 revenue expected $1.42B ±5%.
  • GAAP gross margin 48.2–50.7%; non‑GAAP gross margin 63.5–64.5% (midpoint back toward low end of long‑term target).
  • GAAP OpEx ≈ $680M; non‑GAAP OpEx ≈ $430M.
  • Q4 non‑GAAP EPS $0.41–$0.51; GAAP EPS range -$0.08 to $0.02.
  • Other expense ≈ $50M; non‑GAAP tax rate 6% (Q4), 7% in FY25.
  • Data center revenue expected +mid‑30% sequential in Q4 and to be >50% of total revenue.
  • Carrier Q4 expected down ~mid‑40% seq; enterprise down mid‑single digits; auto/industrial down ~20%; consumer down mid‑teens.

Business Commentary:

* Revenue and Data Center Growth: - Marvell reported revenue of $1.42 billion for Q3, growing 6% sequentially above the midpoint of guidance. - Data center revenue was the largest contributor, reaching $556 million, well above guidance, driven by strong AI and cloud demand.

  • Enterprise and Carrier Market Challenges:
  • Revenue in the enterprise networking segment was $271 million, declining 28% year-over-year and 17% sequentially.
  • The decline was attributed to weak demand and inventory corrections, despite Marvell's efforts to manage inventory and adjust production.

  • Carrier Infrastructure Performance:
  • Carrier infrastructure revenue was $317 million, above guidance, with a 15% sequential increase and 17% year-over-year growth.
  • The growth was due to Marvell's specific product cycles, but the company expects a decline in the fourth quarter due to demand softness and reduced carrier CapEx.

  • Automotive and Industrial Segment Fluctuations:
  • Marvell's automotive and industrial segment revenue was $107 million, growing 26% year-over-year but declining 3% sequentially.
  • The sequential decline was due to lumpy order patterns in aerospace and defense, while year-over-year growth was supported by demand in automotive.

  • Guidance and Future Expectations:

  • Marvell's guidance for Q4 expects data center revenue to grow by mid-30% sequentially, with strong growth from AI and standard cloud infrastructure.
  • The company anticipates revenue from the enterprise on-premise portion of the data center to decline sequentially, reflecting ongoing soft demand.

Sentiment Analysis:

Overall Tone: Positive

  • Management reported revenue and non‑GAAP EPS above midpoints, data center revenue up 21% Q3 and guided data center +mid‑30% QoQ for Q4 with AI exit rate 'well north of $200M.' They guided a >300 bps sequential non‑GAAP gross‑margin improvement in Q4 and repeatedly emphasized strong AI/cloud momentum despite near‑term softness in enterprise/carrier.

Q&A:

  • Question from Toshiya Hari (Goldman Sachs): Can you size the AI business for the October quarter (Q4) and provide visibility into optical and compute businesses into calendar 2024?
    Response: AI is driving a strong data center rebound: data center grew >20% in Q3, Q4 data center guide +mid‑30% QoQ with AI exit 'well north of $200M'; optics demand strong and custom compute (ASIC) ramps are progressing for production next year but sizing remains early.

  • Question from Timothy Arcuri (UBS): Why is enterprise networking down (inventory increasing at customers) — is this Marvell product or other suppliers — and can total revenue grow in fiscal Q1 given data center strength?
    Response: Enterprise decline is an industry‑wide OEM inventory correction and softer demand; it will take several quarters to work through, and Q1 growth is uncertain because declines across roughly half the business may outweigh data center strength.

  • Question from Vivek Arya (Bank of America): How should we think about your ASIC/custom silicon business size, customer concentration, visibility and revenue potential next 1–2 years?
    Response: Custom silicon is highly concentrated among a few hyperscalers; optics drove roughly $400M AI revenue this year and the overall AI/custom opportunity is now larger than previously estimated, but management cannot yet precisely size custom silicon for next year.

  • Question from Harsh Kumar (Piper Sandler): With carrier guide down mid‑40% QoQ, which subsegment (wired vs wireless) takes the bigger hit and which will recover first?
    Response: Both wired and wireless will be soft; wireless outperformed this year but both are lumpy and cyclical, timing of recovery is uncertain though share gains and new 5nm base station wins should drive eventual recovery.

  • Question from Tore Svanberg (Stifel): What does the NVIDIA–Marvell collaboration mean — is it multiyear and any additional color?
    Response: It's a long‑standing, multi‑year complementary partnership (not a new product announcement) focused on optical interconnects and joint work to enable AI deployments.

  • Question from Tore Svanberg (Stifel): Is the less‑bad sequential decline in enterprise networking a sign of stability or is visibility still limited?
    Response: Visibility remains limited; enterprise demand is still declining and management is monitoring customer balance sheets before declaring stability.

  • Question from Ross Seymore (Deutsche Bank): How much of the weakness outside AI/cloud is cyclical versus de‑prioritized structurally?
    Response: Most weakness (enterprise, carrier, storage) is cyclical and expected to recover; consumer is the structural decline the company has de‑emphasized.

  • Question from Matthew Ramsay (TD Cowen): Are customers leaning toward merchant, ASIC houses, or in‑house silicon, and are they investing software at scale to support custom programs?
    Response: Customers are pursuing multiple models (merchant, custom, in‑house); design activity is very high and customers are well‑capitalized and investing in software/services, supporting sizable custom program ramps.

  • Question from Ambrish Srivastava (BMO): How should we think about gross‑margin profile next year given AI/custom and potentially lower‑margin customers?
    Response: If custom grows materially it could pressure gross margin, but management expects custom to be accretive to operating margin and EPS; team is controlling costs to protect profitability.

  • Question from Ambrish Srivastava (BMO): You mentioned a >1:1 attach rate for optics; what changed versus earlier?
    Response: Clarification: attach rate is >1:1 because optics attach not only accelerators to switches but also between switch layers, increasing overall optics per accelerator.

  • Question from Harlan Sur (JPMorgan): Are you seeing a sustainable pickup in data‑center storage (HDD/storage customers) heading into next year?
    Response: There are encouraging signs (inventory normalizing, cautious optimism from customers) but management remains cautious and is not yet calling a recovery until backlogs and unit demand clearly improve.

  • Question from Christopher Rolland (Susquehanna): Are you tracking above or below the $200M custom expectation this year, and is the $800M target still on track for FY25–26?
    Response: Management is tracking close to the ~$200M target this year; the $800M long‑term target (FY25–26 timeframe) remains plausible and chip programs are taped‑out and looking ready to ramp.

  • Question from Christopher Rolland (Susquehanna): Is Q4 the bottom for carrier or could there be further declines; will additional content wins meaningfully lift the segment?
    Response: Softness is expected to continue into Q1; carrier will recover over time and content increases (more sockets, new DSPs) will provide a meaningful tailwind when demand normalizes, but timing is uncertain.

  • Question from Srinivas Pajjuri (Raymond James): Given supply chain constraints (HBM, packaging), when will you start building wafer/packaging inventory to support ramp next year?
    Response: Capacity planning is underway, some wafers are being started and capacity/reservations are in place; management expects materially better visibility by around March as ramps and lead times clarify.

  • Question from Srinivas Pajjuri (Raymond James): Given strong recent AI momentum, why not guide for Q1 growth now?
    Response: Management prefers to focus on executing Q4 and will provide Q1 guidance next quarter; while year‑over‑year trends look favorable, near‑term dynamics remain too fluid to guide Q1 now.

  • Question from Chris Caso (Wolf): What drove the cloud rebound — cyclical inventory or product‑cycle demand?
    Response: Cloud strength is driven by both product cycles (400G, 12.8T switches ramping) and increased networking demand to support AI deployments, not just transient inventory effects.

Contradiction Point 1

Custom Compute and ASIC Programs Revenue Expectations

It involves differing expectations for revenue contributions from custom compute and ASIC programs, which are strategic growth areas for Marvell.

Are you above or below the $200 million expectation for the custom silicon opportunity, and what is your long-term goal? - Chris Rolland (Susquehanna)

2024Q3: Marvell is tracking close to the $200 million expectation for fiscal 2024 and maintains the long-term goal of exceeding $800 million between FY '25 and '26. - Matthew Murphy(CEO)

What is the momentum of design wins for custom products, and what proportion of H2 revenue is expected from new programs versus existing design wins? - Jeremy Lobyen Kwan (Stifel)

2026Q2: We are now pursuing several billion-dollar projects, which further enhances our confidence in achieving a 20% market share target by 2028. - Christopher Koopmans(COO)

Contradiction Point 2

Carrier Market Demand and Demand Projections

It highlights differing perspectives on the demand projections and weakness in the carrier market, which directly impacts revenue expectations.

Which segment—wired or wireless—has been more affected by the carrier market's challenges, and when do you expect a recovery? - Harsh Kumar (Piper Sandler)

2024Q3: Both segments are weak, with carrier expected to decline mid-40% sequentially. Wired has been weakening, while the wireless segment saw a strong 5G rollout completion. Uncertainty in global carrier CapEx is a concern. A multi-year digest is expected before a return to growth. - Matthew Murphy(CEO)

What are your expectations for the second half of the year? - Chris Caso (Wolfe Research)

2026Q1: We expect the carrier market to grow around 5% to 10% in the second half of this year, and then we expect the recovery to start in earnest in fiscal '27. - Matt Murphy(CEO)

Contradiction Point 3

AI and Custom Compute Revenue Projections

It involves differing projections for AI and custom compute revenue, which are critical growth areas for the company.

What is the size of Marvell's AI business this quarter, and what visibility do you have for its optics and compute businesses in 2024? - Toshiya Hari (Goldman Sachs)

2024Q3: The data center business grew over 20% in Q3. Q4 guidance shows AI revenue exceeding $200 million. AI and cloud infra will be growth drivers into next year. - Matthew Murphy(CEO)

Can you detail data center revenue and discuss the year's AI revenue trajectory? - Timothy Arcuri (UBS)

2026Q1: AI continues to be the fastest-growing portion of our data center business. We expect AI to become the majority of HoldCo. The custom business runs at a lower gross margin, but we've managed this overall with strong recovery in other businesses. - Matt Murphy(CEO)

Contradiction Point 4

Optics Revenue and Growth Opportunities

It involves the company's revenue projections and growth potential in the optics segment, which is a significant part of their business strategy.

What is the size of Marvell's AI business this quarter, and what visibility do you have for its optics and compute businesses in 2024? - Toshiya Hari (Goldman Sachs)

2024Q3: Optics revenue is around $400 million, with AI being a significant portion. - Matthew Murphy(CEO)

How do you plan to maintain your lead with XPU customers given a competitor's market share gains? - Ross Seymore (Deutsche Bank)

2025Q4: Optics revenue was $467 million, a record for us. - Matthew Murphy(CEO)

Comments



Add a public comment...
No comments

No comments yet