Marvell Is Buying Lumentum for $115 Billion — the AI Bet Moves From Chips to Light

中文版 Chinese version

Deal terms below are from Marvell and Lumentum’s joint announcement of November 2, 2026, and Reuters’ reporting of the accompanying regulatory filing. The $115 billion is an implied enterprise value computed from share prices on October 28, 2026, not a fixed cash price.

The deal in one paragraph

Marvell Technology will acquire Lumentum in an all-stock transaction. Each Lumentum share converts into a fixed 4.500 Marvell shares, implying about $1,132 per Lumentum share — a 23.9% premium to Lumentum’s unaffected closing price on October 28, the last trading day before deal speculation surfaced. On completion, Marvell shareholders would own about 67% of the combined company and Lumentum shareholders about 33%. Both boards approved the deal unanimously; Marvell CEO Matt Murphy stays as chairman and CEO, and three Lumentum directors, including CEO Michael Hurlston, join Marvell’s board. Closing is targeted for the first half of 2028, subject to both companies’ shareholder votes and regulatory clearance.

The same day, Lumentum reported fiscal first-quarter revenue of $1.26 billion, slightly above the roughly $1.25 billion consensus cited by market coverage. Marvell, meanwhile, reported preliminary fiscal third-quarter revenue in line with its prior guidance midpoint. The companies claim line of sight to more than $1 billion in annual run-rate cost synergies and a combined addressable market of $550 billion — the latter is a company estimate, not a measured market, and should be read as positioning.

Why light, why now

The strategic logic is about a bottleneck that has moved. For years the constraint on AI training was compute: more GPUs, faster accelerators. As clusters scale to tens of thousands of chips, the binding constraint becomes moving data between them — and copper interconnects hit physical limits on bandwidth, reach, and power. Marvell’s own framing is blunt: the industry is shifting “from discrete, modular components toward highly integrated optical solutions” that require electrical and optical engineering to be designed together, not bought separately.

That is what Lumentum brings: high-performance lasers, optical transceivers, optical circuit switching, and compound-semiconductor manufacturing (including indium phosphide) that Marvell’s electrical interconnect, switching, and custom-silicon business does not have. The combined portfolio is meant to span everything from connections between distributed data centers down to connections inside an advanced semiconductor package — the scale-up and scale-in domains the industry now argues about most.

The competitive target is named, if only indirectly: Broadcom, which has long been Google’s main custom-chip partner and already sells a broader AI data-center hardware menu under one roof. Reuters notes Marvell signed a deal in August to supply chips alongside Google’s AI processors worth up to $120 billion in sales through fiscal 2033. Buying Lumentum is Marvell’s attempt to match that breadth — and to keep NVIDIA, which invested $2 billion each in Marvell and Lumentum earlier this year according to Reuters, as a customer rather than a competitor in optics.

The number to be careful with

The headline — $115 billion, one of the biggest deals of 2026 — deserves a footnote that matters more than the headline. Because the exchange ratio is fixed, the dollar value of the deal floats with Marvell’s share price until closing. A $10 move in Marvell’s stock changes the consideration by $45 per Lumentum share. The $115 billion and $1,132 figures are snapshots from October 28 reference prices, not money anyone will necessarily receive.

The regulatory filing adds real break costs, as reported by Reuters: Marvell would owe Lumentum $5.8 billion if its board drops support or it walks away for another takeover, Lumentum would owe Marvell $2.9 billion in the mirror case, and Marvell would pay $3.2 billion if regulators block the deal. Those numbers tell you how seriously both sides take the antitrust risk — and how expensive a failed bet would be.

The case against

There is a genuine counter-argument, and it arrived on the same day as the announcement. Reuters’ own framing notes the deal lands as investors question whether AI spending is sustainable, with chip stocks selling off on tightening margins and slowing cash flow. Lumentum’s valuation — about 22.8 times its latest quarterly revenue annualized, per one independent calculation — prices in years of uninterrupted AI-infrastructure growth. If hyperscaler capex pauses, or if co-packaged optics ramps slower than the pluggable optics it replaces, Marvell will have paid a peak-cycle price for a cyclical supplier.

The other risk is execution. Marvell’s last comparable integration, the $10 billion Inphi purchase in 2021, was far smaller. This deal asks a chip company to absorb a photonics manufacturer with different fabs, different yields, and different customers — and to do it while the technology itself (co-packaged optics versus pluggables) is still being decided by the market.

The judgment

Strip away the superlatives and this is a coherent, expensive, well-timed-in-one-direction bet: if AI clusters keep scaling, the company that co-designs the electrical and optical path wins a structural advantage, and there are very few independent optical assets left large enough to buy. Lumentum was one of the last. But the price is a peak-cycle price paid in floating stock, the close is more than a year away, and the break fees show both sides know regulators could still say no. The right question is not whether $115 billion is big — it is — but whether Marvell’s stock holds up long enough for the snapshot to become the reality.

Sources: Marvell–Lumentum joint announcement, November 2, 2026; Reuters, “Marvell to buy Lumentum in $115 billion deal in AI data-center bet,” November 2, 2026.