Marvell's $115 Billion Lumentum Deal: AI's Biggest Bet Is on Light, Not Chips

On 2 November 2026, Marvell Technology agreed to buy Lumentum Holdings in an all-stock deal that values the optical-components maker at an implied enterprise value of about $115 billion, the largest technology deal of the year. Lumentum shareholders would receive 4.5 Marvell shares for each share they own, worth about $1,132 per share, a premium of roughly 23.9% over the unaffected closing prices on 28 October, the last trading day before speculation about a deal surfaced. After completion, existing Marvell shareholders would own about 67% of the combined company and Lumentum shareholders about 33%. Both boards approved the deal unanimously, Marvell chairman and chief executive Matt Murphy would stay in charge, and three Lumentum directors, including chief executive Michael Hurlston, would join Marvell's board. The companies expect the deal to close in the first half of 2028, subject to shareholder votes and regulatory approvals.

The logic of the deal is not about chips in the usual sense. It is about what happens between them. As AI systems grow to thousands of processors working on a single problem, moving data between those processors has become as hard as the computing itself. Electrical connections run into limits on bandwidth, energy use and distance, and the industry's answer is increasingly light: lasers, photonic components and optical interconnects that carry information between servers with less power and over longer reaches. That is Lumentum's business, from high-performance lasers and optical transceivers to optical circuit switching, and it is the missing half of Marvell's portfolio of custom silicon, networking and interconnect technology. Marvell puts the combined company's total addressable market at about $550 billion, a company projection rather than a measured fact, but the direction of travel is visible in Lumentum's own numbers: its optical components were reported sold out through early 2029, it disclosed a second major co-packaged-optics win backed by more than $5 billion in purchase commitments on the same day as the deal, and it forecast second-quarter revenue of $1.58 billion to $1.63 billion, above the average analyst estimate of $1.48 billion.

The market did not applaud. Marvell shares fell more than 5% and Lumentum shares more than 10% in early trading on the announcement day, and analysts questioned the price and the timing. One sell-side note put mid-single-digit earnings dilution in 2028, before any of the promised synergies arrive; Marvell projects more than $1 billion in annual cost savings, but those are projections, not results from two businesses operating together. There is also a scale question that no synergy slide answers. Marvell's annual revenue grew from about $3 billion in fiscal 2021 to $8.2 billion in fiscal 2026, and its previous large deals, the roughly $10 billion purchase of Inphi in 2021 and the $3.5 billion purchase of Celestial AI in February 2026, were an order of magnitude smaller than this one. Nvidia, which invested $2 billion in each company earlier this year, clearly believes connectivity is the strategic layer; the falling share prices say that belief is not yet priced as a bargain.

The longest shadow over the deal is regulatory. The merger agreement names an outside date of 1 November 2027, extendable in steps to 1 August 2028 if regulatory approvals are still pending, and it prices failure precisely: Marvell would owe Lumentum $5.8 billion if it walks away for a better offer or drops board support, Lumentum would owe $2.9 billion in comparable circumstances, and Marvell would owe $3.2 billion if the deal collapses on regulatory grounds while the other closing conditions are ready. Clearance is needed in the United States and other jurisdictions, and reporting has flagged China in particular, since both companies sell extensively into Chinese cloud and data-centre customers. A $3.2 billion regulatory termination fee is the agreement's way of admitting that this is a real risk, not a formality.

I read this deal as a test of a specific claim: that the next bottleneck in AI infrastructure is photonics, and that it is worth paying a peak-cycle price to own it. If the claim is right, Marvell stops being mainly a chip supplier and becomes the company that moves light around the data centre, and the $550 billion market figure starts to look like a description rather than a pitch. If it is wrong, or if regulators slow the deal past its window, Marvell will have spent eighteen months and considerable credibility on a combination that never closed. That is the prediction I am making with this article: that the Marvell-Lumentum deal closes on or before 30 June 2028, in line with the companies' target. The unanimous boards, the industrial fit and the demand for optical components point toward closing; the regulatory path, the size of the bet and the market's scepticism are why I put that at 60%, not higher.