Marvell Technology is in advanced talks to buy Lumentum Holdings in a deal that would unite one of the artificial intelligence boom's biggest chip winners with the optical components maker whose lasers move data inside the world's largest data centres. The Marvell Lumentum deal, first reported by Bloomberg News on Wednesday, could be announced as soon as next week — but no final agreement has been reached, and the talks could still fall apart.
Investors reacted instantly. Lumentum shares rose as much as 9% in after-hours trading, while Marvell shares fell more than 8%, a split that captures the central tension of the proposed transaction: a strategically compelling combination that would also be enormous, expensive and almost certainly paid for largely in stock.
Bloomberg News reported on Wednesday that Marvell Technology, the chipmaker whose custom AI chips and networking hardware connect data-centre servers and artificial intelligence systems, is in advanced discussions to buy Lumentum Holdings, a maker of photonic products and optical components for data centres and communications networks. Reuters, citing Bloomberg's reporting, said the discussions were still private and that both companies did not immediately respond to requests for comment.
The timing matters. The report landed in the middle of a bruising week for AI stocks. Alphabet shares plunged after a disappointing earnings report, Microsoft shares rose after results topped Wall Street estimates, and Amazon was due to report after the close on Thursday. On Thursday morning, Nasdaq futures were down 1.1% and futures linked to the S&P 500 and the Dow Jones Industrial Average were down 0.3%, as the AI trade stumbled. Against that backdrop, a megadeal report was never going to be received quietly — and it was not.
Lumentum's after-hours jump of as much as 9% — some market coverage put it at 10% — reflected the prospect of a takeover premium on an $83 billion company. Marvell's drop of more than 8% reflected the other side of the same arithmetic: paying for an $83 billion target is hard without issuing a large amount of new stock, and shareholders immediately began pricing in dilution. One analysis noted that Marvell held only about $3.92 billion in cash and cash equivalents in its latest quarter, about 4.7% of Lumentum's market value, and that an all-stock deal with no premium at all would hand Lumentum shareholders roughly 27% of the combined company. Any premium would push that share higher.
Some coverage framed the possible transaction value at up to $100 billion once a premium is included. That figure is not a confirmed price. No price, premium, cash-and-stock mix or closing timeline has been reported, and neither company has confirmed the talks. Until terms are announced, every valuation figure is a market-value snapshot, not a deal price.
To understand why Marvell would want Lumentum, it helps to understand what Lumentum actually sells. Modern AI data centres are, in essence, vast networks of chips that must talk to each other at extraordinary speed. Copper connections struggle over longer distances and at higher bandwidths. Optical components — lasers, transceivers and related photonic parts — convert electrical signals into light and carry enormous volumes of data between servers and facilities at ultra-high speeds. They are, as one report put it, core AI infrastructure: without them, the fastest chips in the world cannot work together.
Lumentum makes those parts. Marvell already buys laser products from Lumentum, and the two companies are both based in Northern California. A combination would, in the language of the reporting, simplify Marvell's supply chain by bringing a critical supplier in-house — securing access to components that are currently in short supply, at a moment when every major AI builder is racing to expand capacity.
That shortage is not theoretical. Lumentum chief executive Michael Hurlston said this month that the company's optical parts were sold out through early 2029, and that Lumentum could not meet about 70% of demand for some products next year. Some customers have already begun placing orders for 2030. To meet more of that demand, Lumentum plans to invest at least $350 million in Japan to expand production of indium phosphide and other optical parts — but adding factory capacity takes time, and a deal with Marvell could give Lumentum more resources to expand faster.
Marvell, meanwhile, has emerged as one of the biggest beneficiaries of the AI boom, driven by strong demand for its custom AI chips and networking hardware. Its chairman and chief executive, Matt Murphy, addressed investors at Marvell Technology Investor Day in New York City on October 6, 2026. The company has already been moving deeper into optics: it announced plans in late 2025 to acquire optical technology company Celestial AI for $3.25 billion, another step in its push into data-centre optics. Buying Lumentum would be that strategy at a vastly larger scale.
The wider AI infrastructure land grab provides the context. Nvidia invested $2 billion in Lumentum earlier this year. Meta has raised tens of billions of dollars for AI spending, and chip and optics suppliers across the sector have seen their valuations soar as hyperscalers pour money into data centres. A Marvell-Lumentum combination worth roughly $304 billion on current market values would rank among the largest mergers and acquisitions in the AI infrastructure industry.
The focus keyword of this story is not just a corporate transaction — it is a bet on where the AI bottleneck now sits. For two years, the binding constraint in AI was widely seen as access to the most advanced processors. That constraint has not disappeared, but a second one has joined it: moving data between those processors fast enough, over long enough distances, without melting the power budget of a data centre.
Optics solve that problem, which is why optical component makers have gone from obscure suppliers to strategic assets. Lumentum's sold-out order book through early 2029 is the clearest possible signal of that shift. When a supplier cannot meet 70% of demand for some products next year, customers do not simply wait politely — they sign long-term orders, they fund capacity, and, in Marvell's case, they apparently consider buying the supplier outright.
For Marvell, owning Lumentum would mean control over a critical input to its own networking products, and a direct claim on the fastest-growing part of the data-centre bill of materials. For Lumentum, it would mean a far larger balance sheet behind its capacity expansion, and a parent company whose customers are the same hyperscalers already clamouring for its parts. The strategic logic is unusually clean. The financial logic — price, premium, dilution and regulatory scrutiny — is where the deal will be won or lost.
Neither Marvell nor Lumentum immediately responded to Reuters' requests for comment, and no executive from either company has publicly confirmed the talks. That silence is normal at this stage of a private negotiation, but it leaves the market to trade on the Bloomberg report alone — which is one reason the share moves were so sharp.
Market analysts, as tracked by TipRanks, rate both stocks a Strong Buy, with average price targets implying about 40.89% upside for Marvell and about 24.94% upside for Lumentum. Those ratings predate any announced terms and are not a verdict on the deal itself; they measure how analysts viewed each company standalone. Once terms appear, both sets of targets will be reworked around the premium, the exchange ratio and the combined company's earnings power.
The sceptical reaction in Marvell shares is the more informative signal. Investors were not rejecting the strategy — they were pricing the cost. An $83 billion target is roughly 38% of Marvell's own market value. Absorbing a company that large, integrating its factories and product lines, and satisfying regulators will take years, and the history of very large semiconductor deals is mixed. The optimistic reaction in Lumentum shares is simpler: takeover targets usually get paid a premium, and shareholders were pricing in the chance of one.
Regulatory scrutiny is a further open question. Large AI infrastructure deals are now examined closely on competition grounds, and the context this week underlined that: the US Department of Justice has been examining Nvidia's planned acquisition of AI platform Hugging Face, a separate transaction that has already weighed on Nvidia shares. A Marvell-Lumentum deal would face its own review, and no timeline for approval can be assumed.
If completed, this deal would reshape the AI supply chain in three ways. First, it would vertically integrate one of the most important links in data-centre networking, giving Marvell a captive source of lasers and optical components at a time of severe shortage. Second, it would create a combined company worth around $300 billion, with a far larger claim on hyperscaler capital spending than either company has alone. Third, it would signal that the AI infrastructure race has entered a consolidation phase, in which the winners buy the suppliers they cannot afford to lose.
For consumers and businesses, the effects would be indirect but real. Data-centre capacity determines the cost and availability of AI services, cloud computing and streaming. If vertical integration speeds up optical capacity expansion, it could ease a genuine bottleneck. If it instead concentrates pricing power in fewer hands, it could do the opposite. Regulators, customers and rivals will all be watching which outcome emerges.
For investors, the immediate lesson is narrower: in an AI market where earnings misses are punished swiftly — Alphabet's plunge this week is the latest example — deal reports move stocks violently even before terms exist. Lumentum's 9% after-hours gain and Marvell's 8% fall happened on a report, not an announcement. The announcement, if it comes, will move them again.
The next milestone is the reported window itself: an announcement as soon as next week, if the talks hold. Watch for four things in any announcement: the price and premium, the cash-and-stock mix, the treatment of Lumentum's existing capacity plans including the Japan investment, and the expected closing timeline and regulatory path. If no announcement comes next week, that does not necessarily mean the talks are dead — late-stage negotiations can stretch — but the longer the silence, the more the market will doubt them.
Also watch Amazon's earnings after the close on Thursday and the broader AI stock sell-off for context: a falling market makes stock-funded deals more expensive and can change the negotiating dynamics overnight. And watch for any comment from either company, from Nvidia as a major Lumentum investor, or from large customers. In a deal of this size, the first official word will matter more than any further leak.
If the talks collapse, Lumentum will still be sold out through early 2029 and Marvell will still need optical supply — so the strategic pressure behind this report will not disappear. It may simply reappear in a different form: a partnership, a capacity investment, or a rival bidder.
Not yet. Bloomberg reported that Marvell is in advanced talks to buy Lumentum, but no final agreement has been reached and neither company has confirmed a deal. The talks could still fall through.
No price has been reported. Lumentum's market capitalisation is about $83 billion and Marvell's is about $221 billion. Some coverage has framed a possible value at up to $100 billion including a premium, but that is not a confirmed price.
Lumentum rose as much as 9% after hours on the prospect of a takeover premium. Marvell fell more than 8% because buying an $83 billion company would likely require issuing a large amount of new stock, diluting existing shareholders.
Marvell already buys laser products from Lumentum. Optical components move data between AI servers at ultra-high speeds, Lumentum's parts are sold out through early 2029, and owning the supplier would secure a critical input for Marvell's networking products.
Sources cited by Bloomberg said a deal could be announced as soon as next week. That is a reported window, not a commitment, and late-stage talks can still break down or be extended.
Related coverage on US News Pulse: Alphabet's Q3 earnings miss and the AI spending verdict, Meta's $95 billion AI raise leads a $198 billion day of AI spending, and the DOJ probe into Nvidia's Hugging Face deal.
Leave a Comment
| Today's Date:
