Skip to content
MRVL
~3 min read · 698 words ·updated 2026-04-29 · confidence 78%

Executive Description

Oracle Corporation (NYSE: ORCL) operates Oracle Cloud Infrastructure (OCI), the fastest-growing of the major cloud platforms in 2024-2026 by capex acceleration (FY2026 capex guidance $50B, +136% YoY). Despite this growth, Oracle is essentially a non-customer for Marvell on a direct silicon basis. OCI’s strategic posture is Nvidia-first for accelerated compute (Blackwell, GB200 NVL72 deployments scaling H1 2026) and Broadcom-first for networking switching silicon. Oracle has not disclosed any custom-silicon partnership with Marvell, and Marvell’s optical content into Oracle is at most secondary, flowing through merchant module OEMs that may incorporate Marvell DSPs but with no exclusive design-in. Estimated FY2026 Marvell revenue from Oracle is **$0-50M** (effectively zero), or <0.2% wallet share of Oracle’s $14.8B addressable AI silicon spend. Cross-link customer wallet share Part 2.5. The relevant context for Oracle is the Stargate program (the OpenAI / SoftBank / Oracle joint AI infrastructure initiative announced January 2025, framed as $300B+ over multiple years) — though even there, Marvell’s exposure is indirect.

Marvell Relationship

  • Direct custom silicon: None disclosed. Oracle has not publicly named Marvell as a custom-XPU partner. ✓ (absence)
  • Optical interconnect: Oracle’s data centers use merchant optical pluggables that may contain Marvell DSPs (via Coherent, Innolight, Lumentum module-OEM channel — cross-link partner supplier chain). The pass-through nature of this revenue makes attribution to Oracle imprecise. ◐
  • Networking: Broadcom dominates Oracle’s switching plane. Marvell has minimal networking presence at OCI. ◐
  • Estimated FY2026 revenue: $0-50M, or <0.2% wallet share. Cross-link customer wallet share Part 2.5. ◐
  • Strategic posture: Oracle’s ecosystem is Larry Ellison-driven, cost-flexible, and Nvidia-aligned. The probability of Oracle pivoting to a Marvell-enabled custom XPU under Ellison’s tenure is low.

Stargate $300B Context

  • Stargate is the joint AI infrastructure program announced January 21, 2025 (post-inauguration political event) involving OpenAI, SoftBank, and Oracle, with a stated commitment of “up to $500B” over four years (often summarized as $300B+ near-term). Texas (Abilene) is the anchor first build site. ✓
  • Marvell relevance: Stargate’s compute substrate is largely Nvidia GPU-centric (with selected custom-silicon optionality TBD). Marvell’s exposure to Stargate is indirect at best — through the merchant optical-module ecosystem.
  • OpenAI custom silicon: Reported OpenAI custom-silicon design effort (as of 2025-2026 industry reporting) is more associated with Broadcom than Marvell. Stargate could host this silicon at scale if/when production-ready, further benefiting Broadcom rather than Marvell.
  • Risk to Marvell thesis: A massive Stargate-like Nvidia-and-Broadcom-dominant build-out could absorb hyperscaler capex without flowing meaningful revenue to Marvell.

Recent News

  • 2026-Q1 (March 2026) — Oracle Q3 FY2026 earnings: capex guidance accelerating; OCI revenue growth 50%+; Nvidia GPU build-out continues. Marvell not referenced. ✓
  • 2026-Q1 — Stargate Texas Abilene site build advances; Nvidia GPU shipments scaling. ◐
  • 2025 — Oracle expands Nvidia partnership; GB200 NVL72 first deployments confirmed. ✓
  • 2025-01-21 — Stargate announcement (OpenAI / SoftBank / Oracle); “up to $500B” framing; Texas anchor site. ✓
  • 2024 — OCI growth accelerates; Oracle reports >100% capex YoY growth as cloud demand outpaces existing footprint.

Risks & Catalysts

Risks (For the Marvell investor)

  • Oracle remains a non-opportunity: Even with $50B+ FY2026 capex and Stargate optionality, Marvell’s direct Oracle exposure is structurally near zero. Wallet upside requires a strategic pivot that is not in the public roadmap.
  • Stargate absorbs hyperscaler capex without Marvell participation: If Stargate-class infrastructure attracts OpenAI compute that would otherwise have flowed through Microsoft Azure (Maia-related), Marvell’s Microsoft trajectory could soften.
  • Indirect optical exposure is low-margin pass-through: Module-OEM revenue is structurally lower-margin than direct hyperscaler design wins.

Catalysts (Low-probability)

  • Oracle custom-silicon pivot — would require a strategic shift under Ellison; not currently priced into any Marvell narrative. Low probability.
  • Stargate optical-content disclosure — if Stargate’s networking topology requires Marvell-grade optical DSPs in volume, indirect upside emerges.
  • OpenAI direct custom-silicon partnership with Marvell — public reporting currently associates OpenAI custom silicon with Broadcom; any pivot to Marvell would be a meaningful catalyst.

Sources

KB cross-references