---
title: "MRVL to buy LITE: Strategy looks sound, but the market votes with its feet first."
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/dolphin/post/44622246.md"
description: "Worth it or not hinges on how long the InP shortage lasts. Duration is the key driver."
datetime: "2026-10-29T08:55:37.000Z"
locales:
  - [en](https://longbridge.com/en/dolphin/post/44622246.md)
  - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44622246.md)
  - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44622246.md)
author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)"
generator: "portal-rs"
---

# MRVL to buy LITE: Strategy looks sound, but the market votes with its feet first.

On Oct 28, reports surfaced that Marvell (MRVL) is in advanced talks to acquire optical leader Lumentum (LITE), with an announcement as early as next week. Both sides declined to comment and no agreement has been signed. After-hours, LITE jumped ~11% to a market cap of approx. $91bn; MRVL fell ~8%–9%, wiping out close to $20bn in value.

Dolphin Research's take: **Strategically, MRVL is plugging its biggest CPO-era gap — lasers — so the direction is right**. But the deal would be struck near the top of the optical cycle and is likely largely stock-funded. **The market is voting with its feet over price, not strategy**.

**1) The deal: where things stand**

**Only three things are clear: they are talking, the deal is large, and MRVL lacks the cash**. MRVL had about $3.9bn in cash and equivalents last quarter, less than 5% of LITE's market cap. A transaction this size is highly unlikely to be all-cash, **and would most likely be primarily stock, plus some cash and new debt**.

Four key unknowns remain: **purchase price and premium**, **stock exchange ratio**, **NVDA's stance as a strategic investor in LITE**, and **whether there is a breakup fee and any regulatory structuring**.

**2) Rationale: you cannot remove the laser in CPO, and MRVL does not have it**

MRVL's optical interconnect strength is the DSP inside optical modules. As CPO/NPO proliferate, light engines will be co-packaged with switch chips and XPUs, compressing DSP value capture; but external light sources (CW, ultra high-power lasers) remain indispensable. In late 2025, MRVL announced a $3.25bn acquisition of Celestial AI, whose optical interconnect also relies on external light sources. In short, MRVL has the electrical chips and packaging, **but lacks lasers**.

**Celestial AI also set the stage for the LITE approach**. MRVL bought Celestial to scale up, and **Celestial's Photonic Fabric is a silicon photonics approach that does not emit light, requiring external sources; if Celestial ramps, MRVL's bottleneck shifts from engine design to InP capacity controlled by LITE and Coherent**.

Lasers are the tightest bottleneck across the optical stack: **roughly half of 800G/1.6T module EMLs globally come from Lumentum**. Management says AI optics orders are booked into early 2029 (vs. 2028 six months ago); pump lasers are essentially sold out, and EML shipments in the Dec qtr are guided to grow 50%+ YoY yet still fall short of demand. InP capacity adds are slow and yield ramps are hard, making this the most difficult asset to replicate.

In Mar 2026, NVDA committed $2bn each to LITE and Coherent to secure advanced optics output and capacity rights. AXT signed a prepay-based long-term deal for InP substrates with LITE. **With both downstream customers and upstream materials providers using capital to lock LITE capacity, MRVL's outright acquisition upgrades 'prepay to lock capacity' into 'control to lock capacity'**.

**Dolphin Research sees this as benchmarking Broadcom** — AVGO owns EML/VCSEL lines and can ship a full stack from switch, DSP to light sources. Buying LITE would round out MRVL's 'electrical + optical + system' capability.

Over the past two years MRVL has assembled optical assets: Celestial AI (Photonic Fabric) and Polariton (modulators). The missing pieces are light sources and device manufacturing. **If, as CPO rolls out, light sources and engines get locked into the AVGO or NVDA ecosystems, MRVL risks being relegated from a 'core DSP supplier' to a complementary part in others' reference designs**. **Viewed this way, the deal has both offensive and defensive motives**.

**3) Market reaction: the pushback is price, not strategy**

**First, it looks expensive**. $Lumentum(LITE.US) was at ~$83bn pre-rumor, roughly 17x LITE's annualized revenue. Add a 20%–30% control premium and **the tag could reach $100bn–$110bn, about half of MRVL's own market cap**. LITE is already up 148% YTD, so paying now effectively front-loads the entire optical-cycle optimism.

**Second, stock is the only realistic funding, implying dilution**. MRVL holds only ~$3.9bn cash and has about $5.0bn of interest-bearing debt, making an all-cash bid implausible.

On an all-stock basis with no synergies: at zero premium, MRVL's CY2028 EPS dilution is ~6%; at a 25% premium ~12%, and at 30% ~13%. That is far worse than chatter that 'on current earnings there is little dilution', **because MRVL's own earnings are projected to grow faster into 2028 (custom XPU ramp), making the shares it issues more 'valuable' on a forward basis**. The 8%–9% after-hours drop in MRVL already prices in part of this dilution.

**Third, business-model mix could reset valuation anchors**. **MRVL is a fabless, asset-light company with capex at ~4% of revenue**; **LITE runs fabs, with capex around 15%**. Combined Adj. GPM could slip from 58%–59% to roughly 55%, and the balance sheet would become more asset heavy.

**Lasers are cyclical — 'add capacity, cut price, clear inventory'**. Folding a peak-cycle, capital-heavy asset into an ASIC + networking growth story is another source of investor concern.

**The real bull-bear debate is how long InP tightness lasts**: bulls, on bandwidth math, still see major upward revisions post-2028. Bears argue that once InP capacity comes online, or part of demand shifts to VCSEL, Micro-LED, or CW schemes with lower InP intensity, light-source pricing power will ease. **MRVL is effectively prepaying for the 'optimistic case'**.

**4) Risks and uncertainties: channel conflict, NVDA, and China review**

**i) Channel conflict:** MRVL sells DSPs to module makers such as InnoLight, Eoptolink, and Coherent, while LITE's Cloud Light also ships modules and LITE sells substantial device volumes to China-based module makers. Post-deal, module customers could try to 'de-Marvellize' and pivot to AVGO or other DSP vendors, which bears watching.

**ii) NVDA's stance:** In Mar 2026, NVDA invested $2bn each in LITE and Coherent to lock capacity. Alphabet (GOOGL) is both a MRVL TPU companion-chip customer and a major buyer of LITE's OCS. If LITE folds into MRVL, whether NVDA shifts more share to Coherent becomes a key swing factor.

**iii) China antitrust:** Around 79% of LITE's revenue comes from outside the U.S., with large EML shipments to China module makers. Prior M&A by both firms took 5–9 months to clear with China's SAMR; in today's U.S.-China context, reviews could take longer, include conditions such as 'continued fair supply', or even stall.

**5) Supply chain: light-source scarcity gets 'priced'**

If the deal happens, **it effectively sets an M&A clearing price for high-end light sources (200G/400G EML, ultra high-power CW)**. Near term, peers like Coherent would likely re-rate, with read-through for potential M&A premia.

**For China module makers, impacts skew neutral-to-cautious**: the largest EML supplier would be controlled by a chip company that also sells DSPs and is pushing CPO. Over time this should accelerate validation of domestic EML and CW sources, strengthening the import-substitution case.

For AVGO, this is a competitive signal, not a threat: AVGO's opto-electronic integration started earlier and is more complete. MRVL would need to close and integrate, then win share before CPO ramps in H2 2027; execution risk is non-trivial.

**Dolphin Research believes adding lasers would clearly strengthen MRVL's bargaining power in the CPO era; the strategy is sound**. But whether the deal is worth it **hinges on InP tightness lasting beyond 2028**. **If InP capacity releases in 2027–2028 and prices normalize, MRVL would be swapping its more valuable long-term equity for the priciest light-source asset at a cycle peak**.

Given MRVL's need to shore up lasers (especially post-Celestial AI), **we think MRVL still wants to buy LITE, but completion depends on terms, regulators, and partner stances**.

**MRVL's after-hours selloff broadly discounts dilution under a neutral premium (20%–30%)**. Next, watch the terms: a sub-20% premium, a higher cash mix, or long-term agreements with NVDA and cloud buyers could support a rebound; a >30% premium and all-stock would imply further downside. If talks collapse, MRVL could recoup much of the drop near term.

**For LITE, the current price already embeds some premium expectations; further upside depends on the final offer**. Downside risks include a breakdown in talks and a slower-than-expected capacity ramp in the upcoming earnings.

**Key signposts from here: (1) deal announcement with premium and cash/stock mix; (2) LITE's next print and management's view on InP supply-demand; (3) China review outcome and reactions from NVDA, Alphabet, and other key customers**.

\<End here>

**Dolphin Research on MRVL: archive**

Hot topics:

Oct 7, 2026 Investor Day: '[Marvell: Long-term guide surges — 'confidence' or 'storytelling'?](https://longportapp.cn/en/topics/44216340)'

Aug 20, 2026 Hot Take: '[Marvell: Paying to ally with Google — can it pry share from AVGO?](https://longportapp.cn/en/topics/43507754)'

Earnings:

Oct 7, 2026 Call Trans: '[Marvell (Trans): DC capex to top $3tn; five years of capacity secured](https://longportapp.cn/en/topics/44216864)'

Aug 28, 2026 Earnings Take: '[Marvell: No NVDA, can it still meet lofty hopes — time for multiple compression?](https://longportapp.cn/en/topics/43643427)'

Aug 28, 2026 Call Trans: '[Marvell (Trans): No raise to ASIC guide; DC growth 60%+ next year](https://longportapp.cn/en/topics/43641329)'

May 28, 2026 Call Trans: '[Marvell (Trans): FY2028 revenue guide at $16.5bn, with upside potential](https://longportapp.cn/en/topics/41197136)'

May 28, 2026 Earnings Take: '[Marvell: In the AI inference era, is connectivity the scarce asset?](https://longportapp.cn/en/topics/41191713)'

Mar 6, 2026 Call Trans: '[Marvell (Trans): Raises rev. guide again; a second XPU to ramp in FY28](https://longportapp.cn/en/topics/39099470)'

Mar 6, 2026 Earnings Take: '[Marvell: ASIC card miss — winning back with interconnect](https://longportapp.cn/en/topics/39096551)'

Dec 3, 2025 Call Trans: '[Marvell (Trans): AWS becomes Celestial AI's 'lead lighthouse' customer](https://longportapp.cn/en/topics/36868576)'

Dec 3, 2025 Earnings Take: '[Marvell: Filling gaps via M&A — a faster 'NVDA alternative' path?](https://longportapp.cn/en/topics/36866810)'

Aug 29, 2025 Call Trans: '[Marvell (Trans): Still not addressing the Alchip challenge head-on](https://longportapp.cn/en/topics/33459106)'

Aug 29, 2025 Earnings Take: '[Marvell: Cloud spend is generous — why didn't ASICs benefit?](https://longportapp.cn/en/topics/33455225)'

Risk disclosure and statement: [Dolphin Research disclaimer and general disclosure](https://support.longbridge.global/topics/misc/dolphin-disclaimer)

### Related Stocks

- [LITE.US](https://longbridge.com/en/quote/LITE.US.md)
- [COHR.US](https://longbridge.com/en/quote/COHR.US.md)
- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [AVGO.US](https://longbridge.com/en/quote/AVGO.US.md)
- [MRVL.US](https://longbridge.com/en/quote/MRVL.US.md)
- [LITU.US](https://longbridge.com/en/quote/LITU.US.md)
- [LITX.US](https://longbridge.com/en/quote/LITX.US.md)
- [LITZ.US](https://longbridge.com/en/quote/LITZ.US.md)
- [COHX.US](https://longbridge.com/en/quote/COHX.US.md)
- [NVDD.US](https://longbridge.com/en/quote/NVDD.US.md)

## Comments (1)

- **Chun Hui · 2026-10-29T15:36:59.000Z**: ok


---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**