Marvell shares jump as chipmaker bolsters AI ambitions with Celestial deal

Marvell agreed to acquire photonics startup Celestial AI for $3.25 billion, sending its shares up about 9% premarket as the deal adds photonic fabric for next‑generation data centers and includes a warrant to Amazon tied to product purchases. Management expects Celestial to contribute roughly $500 million in annualized revenue by late fiscal 2028 and $1 billion a year later, with Marvell forecasting roughly $10 billion in revenue next fiscal year driven by a 25% jump in data center sales and a 20% rise in custom chip revenue, positioning the company to better compete with Broadcom and Nvidia.

Analysis

Market structure: Marvell (MRVL) becomes a direct beneficiary of photonics commercialization; Celestial AI adds a differentiated asset potentially accelerating data-center wins vs Broadcom (AVGO) and Nvidia (NVDA). Expect a modest share-shift in networking/XPUs over 2–5 years if Marvell secures hyperscaler design wins — the firm forecasts $500M run-rate by late-FY2028 and $1B by ~FY2029 versus a $10B revenue base next year, implying photonics is material but not transformational immediately. Valuation arbitrage exists (MRVL 12-month fwd P/E 27.3 vs AVGO 38.4) which markets can reprice quickly on positive/negative signals.

Risk assessment: Key tail risks are failed integration/scale of photonics, hyperscaler standards fragmentation, and potential antitrust scrutiny tied to Amazon’s warrant or exclusivity; each could truncate the $1B revenue path. Time horizons split: expect volatile price moves in days-weeks around milestones (warrant filings, PoC wins) and binary realization of revenue by 2027–2029; liquidity/IV may spike, so use cost-limited structures. Hidden dependencies include supply-chain optical components and co-packaged optics ecosystem adoption; a competitor price war could compress margins by >200–300bp.

Trade implications: Implement a size-controlled long MRVL exposure (2–3% portfolio) via 9–18 month call spreads to capture integration proof points while capping premium; consider long-dated LEAPS (18–30 months) if conviction on 2028 revenue. Pair: long MRVL / short AVGO (weight ~1.5:1) to play valuation re-rating subject to rebalancing if MRVL misses milestones; avoid naked directional NVDA shorts. Reduce pure semiconductor cyclical beta by 3–5% and rotate into AI infrastructure names with clear hyperscaler commitments.

Contrarian angles: Consensus equates Celestial with guaranteed share gains — adoption of CPO is far from certain and ecosystem winners may be those who control standards and supply (potentially AVGO/NVDA). The market may underprice execution risk; if Marvell misses early PoC orders within 6–9 months, expect >25% downside from current levels. Conversely, a confirmed Amazon-led deployment would force rapid re-rating, creating asymmetric upside for disciplined, cost-limited longs.

Key Decisions for Investors

• Establish a 2–3% portfolio long in MRVL via a 9–18 month call spread (buy ATM calls, sell 20–30% OTM calls) to limit premium; add another 1% if Marvell announces a hyperscaler PoC order within 6 months, take profits if MRVL rallies >40% from entry.
• Implement a pair trade: go long MRVL (1.5–2% weight) and short AVGO (1–1.3% weight) to exploit P/E gap; unwind or rebalance if MRVL underperforms AVGO by >20% over a rolling 60-day window or if AVGO issues explicit share-gain guidance.
• Buy MRVL 18–30 month LEAPS (or equivalent long-dated call spreads) sized 1–2% as a conviction play on 2028 revenue ramp; cap cost with vertical spreads and avoid buying into IV spikes within 7 trading days of major company disclosures.
• Monitor Amazon warrant terms and any SEC/antitrust filings over the next 30–60 days: if warrant structure implies >5% dilution or restrictive exclusivity, reduce MRVL exposure by 50% immediately; if Amazon publicly confirms material purchase commitments within 90 days, increase MRVL position by +1%.