Marvell buys Lumentum for $115B in all-stock AI optics deal
Marvell has agreed to buy Lumentum, the laser and optical-component maker, in an all-stock deal with a $115 billion enterprise value. It is the largest bet yet that the next bottleneck in AI data centers is not the chip but the light path between chips.
Marvell Technology will acquire Lumentum in an all-stock transaction that values the target at roughly $1,132 a share — a 23.9% premium over Lumentum's closing price on October 28, the last trading day before speculation about a deal began, according to the companies' announcement. Lumentum stockholders receive a fixed 4.500 Marvell shares per Lumentum share. When it closes, Marvell holders will own about 67% of the combined company and Lumentum holders about 33%. Marvell puts the combined addressable market at $550 billion and says it has line of sight to more than $1 billion in annual run-rate cost synergies, with the deal accretive to non-GAAP earnings per share in year one.
The strategic logic is narrower and more interesting than the headline number. Lumentum makes indium phosphide lasers, optical transceivers and optical circuit switches. Marvell makes the DSPs it inherited with Inphi, silicon photonics, custom accelerators and the switching silicon around them. Today a hyperscaler buys those pieces from three or four vendors and hopes the combination holds at 1.6T and beyond. Marvell is buying the right to sell the whole signal path and pre-validate it before a customer ever racks it.
That matters because the wire is becoming the constraint. As clusters grow to hundreds of thousands of accelerators, links that used to run a few metres over copper are being pushed to optics, and optics is moving from pluggable modules at the edge of the switch to co-packaged connections inside it. Every one of those transitions increases the number of lasers per system. Lasers are also the part of the chain that is hardest to scale: compound semiconductor fabs take years to bring up, and yield problems show up as supply shortages, not headlines.
Marvell is paying up for that scarcity. Lumentum was worth about $76 billion by market capitalization and had risen 128% this year before the deal, Bloomberg reported, making this the largest M&A transaction of 2026. Note the structure: Lumentum holders are not taking cash. They are taking Marvell stock at a fixed ratio, which means a fixed bet on a company whose shares have been volatile all autumn.
One accounting note that changes how the number reads: $115 billion is enterprise value, which counts Lumentum's debt, not the equity cheque. The shares changing hands are worth less than the headline, and the premium is calculated on the equity. Marvell is not writing a $115 billion check; it is issuing stock and inheriting a balance sheet.
Volatile is the polite word. Since talks were first reported in late October, Lumentum shares have fallen about 8% and Marvell about 22%, in a broad chip selloff that accelerated after Intel's weak forecast, CNBC reported. A 23.9% premium measured against October 28 is therefore a premium against a price that no longer exists; whether Lumentum holders are actually being paid a premium depends on where Marvell trades when the votes are counted.
Marvell is not a stranger to this playbook. It bought Inphi for $10 billion in 2021, closed a $3.5 billion purchase of Celestial AI in February, and in August signed a deal to supply silicon alongside Google's AI processors worth up to $120 billion in sales through fiscal 2033, with Google holding the right to buy a $12.2 billion stake in Marvell. Revenue has gone from $3 billion in fiscal 2021 to $8.2 billion in fiscal 2026, and the company's market value now tops $170 billion. The Broadcom comparison is the one Marvell wants: Broadcom sells custom silicon and the networking around it as one package, and has long been Google's main custom-chip partner. We looked at how that bundle actually gets financed in Broadcom lends Anthropic $42B to lease the chips it designs.
There is a quieter detail worth flagging: Nvidia invested $2 billion each in Marvell and Lumentum earlier this year, per CNBC. The largest buyer of AI optics has been writing cheques to both halves of this merger. Read that as Nvidia hedging its interconnect supply, or as Nvidia deliberately funding a second source to Broadcom's optics bundling — either way, the consolidation is happening inside a supply chain Nvidia already owns a piece of, which is a strange place for an antitrust regulator to find itself.
Integration is the part markets usually underrate, and the bull case here rests on Marvell having done it before. Analyst Patrick Moorhead argued on X that Marvell "times and integrates its M&A well" and called Inphi, Celestial AI and now Lumentum the right strategy at a big price. He is right that the pieces fit; the open question is whether a company that has never absorbed a business this large can keep two laser fabs and a custom-silicon design team on the same roadmap for three years.
The counter-argument is the calendar. Marvell expects the deal to close in the first half of calendar 2028. That is not a typo and not a formality: a fifteen-month-plus path to close tells you the companies expect a real review, and analyst Ben Bajarin singled out China as the only regulatory challenge he sees. A deal priced on 2026 AI capex, closing in 2028, is a bet that the spending curve does not bend in between. If it does, Marvell will have issued a third of its equity for lasers at the top of a cycle.
The competitive field the combined company lands in is genuinely crowded. Coherent competes directly in optics; China's Zhongji Innolight does too, and a proposed US ban on new Chinese optical transceivers would push demand toward domestic suppliers — a tailwind for Marvell-Lumentum that has nothing to do with engineering. Broadcom, meanwhile, is unlikely to sit still; expect an interconnect answer within two quarters.
What to watch: the shareholder votes, China's antitrust review, and laser capacity commitments. Anyone can promise $1 billion of synergies. Only the fabs can promise the lasers, and the whole thesis rests on them.
Would you rather own the chip or the light between the chips? Tell us where you think the margin ends up.
