Marvell to buy chip startup Celestial AI for $3.25 billion, bullish on growth next year

Marvell agreed to acquire photonics-focused startup Celestial AI for $3.25 billion ($1.0B cash + 27.2M shares worth $2.25B) and issued a related warrant to Amazon allowing up to ~$90M of stock purchases tied to photonic product buys through 2030. Management guided to roughly $10 billion in revenue for next fiscal year with a 25% increase in data center revenue and 20% growth in custom chip revenue, gave Q4 revenue guidance of ~$2.2 billion (+/-5%) and reported Q3 revenue of $2.07 billion (+36.8% YoY). Marvell expects Celestial-related meaningful revenue in H2 FY2028 with a $500M annualized run rate by Q4 FY2028 and $1B by Q4 FY2029; the deal is expected to close in Q1 2026, positioning Marvell to compete with Broadcom and Nvidia in photonics-enabled AI infrastructure.

Analysis

Market structure: Marvell (MRVL) materially improves its vertical stack by adding Celestial's silicon photonics capability, boosting potential data-center TAM and pricing power versus peers (Broadcom AVGO, Nvidia NVDA) for custom interconnects. Short-term share reallocation favors MRVL and specialized photonics suppliers; legacy copper/ASIC incumbents face margin pressure in cloud custom silicon. Cross-asset: expect elevated MRVL equity vols and tighter credit spreads for Marvell if revenue guidance sustains; modest positive spill to semiconductor-equipment names (ASML, LRCX) and specialty materials, limited commodity impact.

Risks: Tail risks include failed integration, photonics under-adoption, or a competitive strike from AVGO/NVDA via M&A—each could wipe >30% of projected Celestial-driven value. Time horizons are elongated: market reaction immediate (days), meaningful revenue only in H2 FY2028 and FY2029 (500M and $1B run-rates pledged), so execution risk dominates quarters-to-years. Hidden dependencies: customer POCs (Amazon/Microsoft) and the Amazon warrant tie economics to purchase volumes through 2030.

Trade implications: Tactical: size exposure to MRVL with multi-stage entries — capitalize on post-deal volatility but anchor to milestones (deal close Q1 2026, design-win announcements 2026–27, run-rate targets 2H FY2028). Use pair trades (long MRVL, short AVGO 0.5–1% notional) to hedge macro semiconductor beta; buy 12–18 month MRVL LEAP calls funded by short 3–6 month call overlays to monetize high near-term IV. Rotate 2–4% portfolio overweight into photonics/equipment suppliers and trim broad networking names vulnerable to custom silicon disruption.

Contrarian: Consensus underprices execution and capital intensity — the market is likely over-excited on immediate revenue; short-term pop (~13% after-hours) is probably overdone. Historical parallels (Broadcom-style rollups) show long gestation with diluted EPS initially; watch dilution from the 27.2M share issuance and the $1B cash outlay. Unintended consequences: customer lock-in could provoke competitive bundling or regulatory scrutiny; set hard stop-losses tied to missed 2026 design-win cadence.

Key Decisions for Investors

• Establish a 1.5–2.5% long position in MRVL on a 24–36 month horizon, scaling in between now and Q1 2026; take 25% profits if share price rises >35% pre-deal-close and take another 40% at announced $500M run-rate achievement (expected Q4 FY2028); initial stop-loss at -20%.
• Initiate a 0.5–1.0% pair trade: long MRVL / short AVGO (equal dollar) to isolate photonics/custom silicon alpha versus broad semiconductor infrastructure risk; rebalance if AVGO outperforms by >10% or MRVL misses a 2026 design-win milestone.
• Allocate 0.5–1.0% to MRVL options: buy Jan 2027 LEAP calls ~ATM or ~10% OTM and sell 1–3 month calls to fund 30–50% of premium; target >2x payoff if Celestial revenue trajectory materializes in FY2028–29.
• Overweight (2–4% portfolio) semiconductor equipment and photonics suppliers (e.g., LRCX, ASML, smaller optical component specialists) for 12–36 months; concurrently trim 1–2% positions in legacy switching/networking vendors vulnerable to custom silicon displacement.
• Monitor three binary catalysts: (A) deal close Q1 2026 (if delayed >90 days reduce MRVL exposure by 50%), (B) proof-of-concept/design-win announcements from Amazon/MSFT by end-2026 (no announcements → reduce MRVL by 50%), (C) Celestial-related revenue proofs in H2 FY2028 (miss → exit options bets).