TL;DR · 30-second read
The Short Version
Marvell, a large American chipmaker, agreed to buy Lumentum for about $115 billion. Lumentum makes tiny lasers that turn data into light, so it can travel through the glass cables that link computers inside artificial intelligence data centers.
Investors are uneasy. Marvell’s stock is down more than half from its peak, and Lumentum’s has fallen 42%.
One detail stands out: almost half of Lumentum’s sales come from just two customers. If either one slows its spending on artificial intelligence, the deal pays off less. The purchase is not expected to be completed until 2028.
Marvell Technology’s planned acquisition of laser maker Lumentum Holdings, valued at $115 billion, is facing a skeptical market. GuruFocus reported that Marvell shares have fallen 56% from their record close, slipping below the stock’s 200-day moving average (the average closing price over roughly the past 40 weeks, a common trend gauge) amid concerns about slowing AI infrastructure spending and dilution from the purchase. The companies expect the deal to close in the first half of 2028, once shareholders and regulators approve it.
The pressure extends to the target. BigGo Finance reported that Lumentum’s shares have sunk 42% as its chief executive defends the transaction, and 24/7 Wall St. reported the CEO acknowledging that “shareholders are not happy.” Lumentum’s two largest customers accounted for 48% of its September-quarter revenue.
Executive Summary
Marvell is buying Lumentum to add laser technology to its optical networking business. Optical networking is the fiber-optic links that carry data between processors, servers and switches inside AI data centers. Management forecasts more than $1 billion in annual cost reductions and higher adjusted earnings per share in the first full year after the deal closes.
The market is weighing that forecast against three risks that are easier to quantify. Marvell may issue new shares to fund the purchase, which would dilute existing holders. Lumentum’s revenue is concentrated, with two customers producing nearly half of it. And the deal will not close for more than a year, during which AI spending could slow.
The concentration figure matters most for the deal’s revenue case. Marvell controls the $1 billion savings target. It does not control the purchasing decisions of the two buyers that drive 48% of the business it is acquiring.
Savings Are Marvell’s to Deliver. Revenue Belongs to Two Buyers.
Merger cases usually have two parts. The first is cost: duplicated functions get removed, purchasing gets consolidated, overhead shrinks. The second is revenue: the acquired business keeps growing and fills the acquirer’s product line. Marvell’s headline forecast of more than $1 billion in annual cost reductions sits in the first part. Management has direct levers over that number, including headcount, facilities and procurement.
The second part is where Lumentum’s customer mix comes in. Two customers accounted for 48% of Lumentum’s September-quarter revenue. For every dollar of revenue Marvell is paying to acquire, nearly 50 cents depends on the capital spending of two organizations. Their identities, contract lengths and committed volumes have not been disclosed alongside the deal. If AI infrastructure spending slows, which is one of the concerns already weighing on Marvell’s stock, a pullback by either buyer would show up quickly in the acquired revenue line. Management’s savings would not offset it.
Concentration of this kind is common among suppliers to large-scale AI buildouts, because only a small number of companies buy optical components in hyperscale volumes. That makes it a structural feature, not a red flag on its own. It does mean the deal’s value case depends heavily on a few customers’ buildout plans. It also raises a question for those customers: their laser supplier is about to become part of a chip company. Whether that changes how they allocate purchases is something only they and Marvell can answer.
A Long Runway Before Integration Starts
The companies expect completion in the first half of 2028, subject to shareholder and regulatory approval. That leaves more than a year in which both companies operate separately while investors price the combined entity. During that window, the AI spending cycle that justified the deal is exposed to the same concerns now cited for Marvell’s 56% drawdown.
The analyst framework also starts late. Consensus estimates project 62% annualized growth in earnings per share from 2028 to 2030, a period that begins only once the deal closes. Those projections assume the demand environment holds through the closing period and into integration. The stock’s slide below its 200-day moving average suggests a meaningful share of investors are less willing than analysts to make that assumption today.
Marvell’s own operating results give the bullish case real support. Data center sales rose 46% year over year in the company’s fiscal second quarter, driven by custom AI processors and high-speed connectivity chips. Adding lasers extends that connectivity franchise from the chips that drive optical links into the light sources inside them. The strategic logic is coherent. The open question is timing and price.
What the Valuation Numbers Do and Do Not Say
Marvell traded at 13.34 times trailing sales on November 18, compared with 13.55 times five years earlier. A trailing sales multiple is the share price divided by the last twelve months of revenue. Even after a 56% decline from the record close, the stock is not cheap by this measure. It has returned to roughly where it stood five years ago, before the AI buildout accelerated. Readers can view that as a reset to a historical norm or as evidence of how stretched the peak had become. The multiple alone does not settle the question.
Wall Street’s consensus is firmly positive: 24 Buy ratings, 5 Holds, no Sells, and an average 12-month price target of $323.63, about 134% above the recent price. A gap that wide between targets and trading price usually means one of two things. Either analysts expect a sharp rebound, or targets have not yet caught up with a changed outlook. Investors should treat the consensus as one input, not a verdict. Dilution is the variable most likely to move per-share math. Management says adjusted earnings per share will rise in the first year, but the number of new shares has not been disclosed, so outside investors cannot yet test that claim.
Lumentum’s 42% decline fits the same framework. When an acquirer may pay partly in stock, the target’s share price generally moves with the acquirer’s and with the perceived odds that the deal closes. A falling Lumentum share price does not, on its own, indicate that the market doubts the value of Lumentum’s lasers. It may mainly reflect what investors think of Marvell’s currency and timeline.
A Bigger Board Heading Into a Bigger Company
Separately, Marvell expanded its board from seven to ten directors effective November 12, according to an 8-K filed with the SEC that day. The new directors are Walmart Chief People Officer Donna Morris, Lam Research CFO Douglas Bettinger and Intuit Chairman and CEO Sasan Goodarzi. Bettinger joins the Audit Committee and was designated an audit committee financial expert. He previously served as CFO of Avago Technologies, now Broadcom, from 2008 to 2013. The filing does not link the appointments to the Lumentum transaction, but the audit committee will oversee financial reporting during an integration of considerable scale.
The announcement also carried a clear statement of management’s outlook. “Marvell is entering an extraordinary period of growth as demand for AI infrastructure continues to accelerate,” Chairman and CEO Matt Murphy said in Marvell’s press release. That view is the opposite of the slowing-spend concern cited for the share decline. The gap between the two is the valuation test the deal now faces. Over the coming quarters, the evidence that resolves it will be customer orders, not commentary from either side.
Background
Marvell Technology, based in Santa Clara, California, has designed data infrastructure semiconductors for more than 30 years. Its customers include cloud, enterprise and telecom carriers. The company’s recent growth has come mainly from data centers, where it sells custom AI processors and the high-speed connectivity chips that move data between them. Lumentum, spun out of JDS Uniphase in 2015, makes lasers and photonic components used in optical networks, including the fiber links that connect equipment inside large data centers.
Optical connectivity has become a central constraint in AI buildouts. Large training clusters link thousands of processors that must exchange data continuously, and fiber-optic links are the main way to do that at scale. Owning both the chips that drive those links and the lasers that generate the light would make Marvell one of the few suppliers with both pieces in-house. Source: Marvell’s $115 Billion Lumentum Deal Creates a New Valuation Test (GuruFocus via TradingView): Marvell’s share decline, deal timeline, projected savings and Lumentum’s customer concentration. Also referenced: Lumentum CEO Defends $115 Billion Marvell Deal as Shares Sink 42% (BigGo Finance) and Lumentum’s CEO Admits “Shareholders Are Not Happy” About the Marvell Deal (24/7 Wall St.). Primary sources: Marvell Technology Form 8-K filed November 12, 2026 (board expansion and director appointments); Marvell press release, Exhibit 99.1: “Marvell Appoints Three New Independent Directors to its Board”.Sources

