Original | Odaily Planet Daily (@OdailyChina)
Author｜Azuma (@azuma_eth)
After U.S. market close on October 28, Eastern Time, the global semiconductor and optical communications market received a blockbuster piece of news.
Bloomberg and The Information, citing people familiar with the matter, reported that chip design giant Marvell Technology (NASDAQ: MRVL) is in advanced talks to acquire data center optical component leader Lumentum Holdings (NASDAQ: LITE).
The people familiar with the matter said that if all goes well, a formal agreement could be announced as early as next week (early November). However, before the final signing, the deal terms and timetable could still be delayed or even fall apart. As of now, reports from various media outlets have not disclosed the specific per-share acquisition price or the payment structure (the exact ratio of cash to stock exchange is unknown), but given Lumentum's massive size, the market generally expects the transaction to include an extremely high proportion of stock exchange.
Before the acquisition news broke, thanks to the explosive demand for optical interconnection in AI data centers, Lumentum's stock price had soared more than 360% over the past year, with a market value of about $83 billion; while the buyer Marvell's market value is about $221 billion. If the usual merger and acquisition control premium is included, the total size of this potential transaction, characterized by foreign media as an "AI super merger," is highly likely to approach the $90 billion to $100 billion mark.
Once completed, this will not only be an unprecedented merger in the history of optical communications, but will also become the most expensive gamble in the global semiconductor industry this year.
Marvell's Motivation: Securing a Position in "Optics Replacing Copper" and Buying Up Scarce Capacity
The underlying driving force behind Marvell's decision to set its sights on Lumentum at all costs at this time is that AI computing infrastructure is hitting a physical wall.
As AI computing clusters evolve from tens of thousands of cards to hundreds of thousands and millions of cards, traditional copper cable connections within servers and between racks have approached their limits in terms of signal attenuation, heat dissipation power consumption, and physical transmission distance. "Optics replacing copper" is no longer just a long-term vision, but a rigid survival rule for data center networks evolving toward 1.6T, 3.2T, and even CPO (co-packaged optics).
For Marvell, swallowing Lumentum points to two major strategic demands.
First, to complete a full-stack vertical closed loop of "electrical chips + optical devices." As early as 2021, Marvell acquired Inphi for $10 billion, securing absolute dominance in PAM4 DSP (digital signal processor), the core electrical chip for optical modules. However, as the industry enters the deep waters of ultra-high-speed optical interconnection, relying solely on DSP electrical chips can no longer meet the stringent requirements of optoelectronic co-packaging (such as silicon photonics and CPO).
The proposed acquisition target, Lumentum, possesses world-leading EML (electro-absorption modulated laser) and CW (continuous wave) high-power laser technology, and Marvell itself is a major customer of Lumentum's laser components. Bringing Lumentum into the fold means Marvell can achieve tightly coupled internal R&D from underlying computing connections, DSP electrical chips to upstream laser light sources, and capture the rich gross margins that would otherwise flow to upstream suppliers.
Second, to control the industry's most bottlenecked "indium phosphide" (InP) wafer production lines and seize supply chain allocation power.
Currently, the biggest bottleneck in the global optical communications industry is not on the demand side, but in the extreme depletion of upstream indium phosphide laser wafer capacity. Lumentum CEO Michael Hurlston said in an interview with Bloomberg in early October this year that Lumentum's optical device capacity has actually been sold out and scheduled through early 2029, with some customers even beginning to lock in orders for 2030; in 2027, as much as 70% of customer demand for some of its core products cannot be met, and the supply gap in 2028 will also reach about 30%.
This extreme shortage may directly threaten Marvell's core growth logic. In high-speed optical modules, electrical chips and optical chips are strictly matched—for every set of lasers that is out of stock upstream, Marvell sells one less set of high-margin DSP or silicon photonics solutions. More critically, as the industry evolves toward CPO, the bundled delivery of external high-power light sources with computing chips has become a major trend. Since building a new indium phosphide wafer fab typically requires a 3-to-5-year ramp-up period, Marvell's move to swallow Lumentum, which owns core global InP production lines, is essentially bringing the industry's most scarce "optical ammunition depot" for the next three years under its command and grasping supply chain capacity scheduling power in its own hands.
A Tale of Two Extremes: Why Did Lumentum Soar While Marvell Plunged?
After the acquisition rumors emerged, during after-hours trading in the U.S. stock market, the stock price movements of Marvell and Lumentum were diametrically opposite—Marvell plunged more than 10% at one point after hours, with a temporary night session quote of around $230, and its after-hours decline remained above 8%; Lumentum surged more than 17% at one point, with a temporary night session quote of around $993, and an after-hours gain of 8.65%
For the acquiree Lumentum, with the expectation of a massive "control premium" layered on top of an already extremely tight capacity cycle, it is understandable that funds would rush to bid up the stock price; but the reason the same transaction triggered a sell-off in the buyer Marvell is that Wall Street still has certain concerns about the "financial ledger" behind this multibillion-dollar deal.
The first is the staggering acquisition cost and equity dilution pressure. Lumentum is by no means a cheap asset at the bottom of the cycle; catalyzed by the AI optical interconnection narrative, its stock price has multiplied over the past year, and recently some Wall Street institutions have even given an aggressive target price of $1,200. Near its all-time high, to swallow a giant with a market value exceeding one-third of its own and pay a considerable control premium, means that Marvell, whose cash flow is not exactly abundant, will inevitably need to issue massive new shares or take on huge debt. For a considerable period after the deal closes, Marvell's existing shareholders will face severe earnings per share (EPS) dilution pain.
The deeper disagreement lies in the fact that this transaction will completely rewrite the genetic makeup of Marvell's business model. For a long time, Marvell has been a typical Fabless (fabless) chip design company, outsourcing all wafer manufacturing to TSMC, thereby enjoying an asset-light, high-margin valuation model; while Lumentum is a full-fledged IDM (integrated device manufacturer) giant, with multiple self-built indium phosphide wafer fabs and packaging and testing bases around the world. Stepping from "asset-light design" into "heavy-asset manufacturing," Lumentum's massive fixed asset depreciation and production line expansion capital expenditure (CapEx) are highly likely to drag down Marvell's originally attractive overall gross margin.
On one side is the optoelectronic closed-loop blueprint that will only be realized in the long term, and on the other is imminent equity dilution and gross margin pressure—which makes it easy to understand why Wall Street institutions chose to "sell first" after hours.
Does Coherent Reap the Benefits?
When the two protagonists at the table caused an industry shock, the stock price of another global photonics and optical communications device giant, Coherent (NASDAQ: COHR), also rose noticeably after hours, briefly breaking through $250 in the night session.
The market's bullish logic on Coherent may be related to "supply chain neutrality." Before this acquisition, Lumentum was a neutral public underlying supplier for the entire industry; its customers included not only major cloud providers and optical module manufacturers, but also chip giants such as Broadcom, Nvidia, and AMD. Once Lumentum is wholly swallowed by Marvell, it will transform from a "public arms dealer" into "an exclusive asset controlled by a direct competitor."
For Broadcom and major cloud providers, handing the lifeline (lasers) of their core network equipment over to Marvell, a rival in custom chips and switches, is absolutely unacceptable from a supply chain security standpoint. To follow the "Second Source" principle and avoid the risk of being strangled, those giant customers who were originally lined up outside Lumentum's door waiting anxiously for 2027-2028 capacity are highly likely to accelerate the shift of massive incremental orders, capacity expansion prepayments, and multi-year long-term agreements to Coherent, the only other independent optical device leader remaining on the market.
In other words, if the Marvell-Lumentum deal ultimately goes through, Coherent's "independent third-party" identity and existing mature capacity may directly enjoy a higher scarcity valuation premium.
Marvell's road to a trillion dollars
In June of this year, Nvidia founder Jensen Huang shared a fireside chat on the same stage with Marvell CEO Matt Murphy. When discussing the evolution of AI infrastructure, the two jointly offered a highly forward-looking industry judgment: The first wave of AI bottlenecks was compute power, which created the first company with a $5 trillion market cap, Nvidia; the second wave of bottlenecks is storage, which has spawned three trillion-dollar storage giants; and the next ultimate bottleneck will be "Connectivity."
On that stage, Jensen Huang pointed at Marvell and shouted to the entire audience: "Ladies and gentlemen, this is the next trillion-dollar company!"
For the helmsman Matt Murphy, to truly leap from the current $220 billion market cap into the "trillion-dollar club" predicted by Jensen Huang, relying merely on selling DSP chips or doing custom ASIC foundry work for cloud providers has a clearly visible ceiling.
Only by taking Lumentum into its fold and forging the industry's most scarce light source capacity for the next three years together with its own silicon photonics, DSP, switch, and custom XPU technologies into an inseparable closed loop can Marvell truly transform from "a core supporting player selling shovels" into "the infrastructure rule-maker" of the AI optical interconnect era.
Early next week, the cards of this nearly hundred-billion-dollar deal may be officially revealed, and the second-half reshuffling of the global AI compute supply chain has only just begun.
