SANTA CLARA, California, November 2, 2026 — Marvell Technology has agreed to acquire optical-components maker Lumentum in an all-stock transaction with an enterprise value of about $115 billion, combining custom AI chips and networking technology with lasers and optical connectivity used to move data through large computing systems. The deal is one of the largest technology takeovers announced this year and places the physical links between AI processors at the centre of Marvell’s growth strategy.

The companies said Lumentum shareholders will receive 4.5 Marvell shares for each Lumentum share, valuing the target’s equity at approximately $102.54 billion, or about $1,132 per share. The offer represents a 23.9% premium to Lumentum’s unaffected closing price on October 28, before reports of the negotiations emerged. After completion, Marvell shareholders are expected to own roughly 67% of the combined company and Lumentum shareholders about 33%.

The agreement has been approved by both boards, but it is not yet a completed acquisition. It remains subject to shareholder votes, regulatory approvals and other closing conditions. The companies expect the transaction to close in the first half of 2028. Until those steps are completed, the businesses remain separate and the announced financial benefits remain projections rather than realised results.

Modern AI data centres connect large numbers of processors that must exchange model parameters, training data and intermediate results at high speed. As those systems grow, moving information between chips can become a bottleneck alongside computing power itself. Optical components use light to carry data and can support high bandwidth over distances and configurations where traditional copper connections face practical limits.

Marvell designs custom processors, networking silicon and connectivity products used in cloud and data-centre systems. Lumentum develops lasers, optical components and related technologies that help transmit data through fibre-based networks and increasingly complex optical systems. The companies argue that bringing those capabilities together could allow them to design electrical and optical components as a more integrated system rather than treating them as separate pieces of a data-centre build.

That is the strategic logic behind the acquisition. AI clusters are becoming more interconnected, and customers want higher throughput, lower latency and better energy efficiency as they expand the number of accelerators working together. Faster processors alone cannot deliver those gains if data cannot reach them quickly enough. Optical links can help reduce the communication constraints that emerge as systems scale, although the benefits depend on system design, workload, manufacturing yields and the way customers deploy the hardware.

The transaction also deepens Marvell’s competition with Broadcom, a major supplier of networking and custom silicon for large technology companies. Marvell has been expanding beyond individual chips into a broader data-centre platform, including custom AI accelerators, switching and interconnect technologies. Lumentum adds an optical portfolio that could help the combined company sell more tightly integrated solutions to hyperscalers and other operators building large AI installations.

Marvell says the acquisition would expand its addressable market to about $550 billion. That figure is the company’s estimate of the opportunity available to the combined business; it is not a forecast of revenue the company will earn. The actual market it can serve will depend on customer design choices, competing technologies, demand for AI infrastructure and how much of the planned spending on data centres is ultimately completed.

The deal follows years of investment in optical technologies and a series of moves by chip companies to secure a larger share of AI infrastructure spending. Nvidia invested $2 billion in each of Marvell and Lumentum earlier this year, according to Reuters. The companies’ products serve different layers of the AI supply chain, and their combination could strengthen the position of a supplier that wants to sell more of the connectivity needed to link large computing clusters.

Optical technology is not a single interchangeable product category. Laser sources, transceivers, optical switching and the packaging that connects photonics with electronic components have different performance requirements and manufacturing challenges. Integrating these elements with custom silicon could simplify some designs and improve efficiency, but it also increases the importance of coordinating development cycles, testing and high-volume production. Customers will still need to qualify the resulting systems against their own reliability, power and cost requirements.

The size of the transaction is the immediate financial test. Reuters reported that Marvell’s shares fell by more than 10% on Monday and Lumentum’s dropped by more than 15% as investors assessed the price and the risks of the combination. Market reactions do not determine whether a deal will succeed, but they show that investors are weighing the strategic opportunity against the cost, dilution and time required to capture the expected benefits.

Marvell is using stock rather than announcing a cash purchase funded by a large new borrowing. That structure reduces the need for the buyer to raise the full transaction value in cash, but it shifts a substantial portion of the combined company’s ownership to Lumentum shareholders and dilutes existing Marvell investors. The final value of the consideration will also move with Marvell’s share price before closing, subject to the terms of the merger agreement.

The companies have outlined more than $1 billion in potential annual cost synergies and said the combination is expected to be accretive to adjusted earnings per share in the first year after closing. Those are forward-looking estimates, not guaranteed savings or earnings. Realising them will require the companies to integrate engineering, sales, supply chains and corporate operations without disrupting product development or customer relationships.

There is also a risk that the demand assumptions behind the transaction prove too optimistic. Technology companies have committed enormous sums to AI infrastructure, but the financial returns from that spending remain uneven. Data-centre projects can be delayed by power availability, permitting, equipment shortages or uncertainty about how quickly AI services will generate revenue. If operators slow their build-outs, demand for optical components and custom processors could grow more slowly than suppliers expect.

Marvell’s bet is that optical connectivity will become a more valuable part of AI infrastructure as clusters scale and systems become more complex. The company has said it expects the addressable market to grow sharply, but competitors are pursuing similar opportunities. Broadcom, specialist photonics firms and other semiconductor suppliers are developing products for the same customers. Large cloud companies may also design parts of their own systems or qualify several vendors to avoid dependence on a single supplier.

Regulatory review is another uncertainty. The companies have not announced a final clearance timetable, and a transaction involving major suppliers of data-centre technology may receive scrutiny in several jurisdictions. The merger agreement includes termination-fee provisions for specified circumstances, according to Reuters, but those provisions do not guarantee that approval will be granted or that the transaction will close on schedule.

The deal’s long timeline creates both room for preparation and a period of uncertainty. The expected first-half 2028 closing date leaves the companies with more than a year to secure approvals, complete shareholder processes and prepare integration plans. At the same time, the AI hardware market can change quickly as chip architectures, networking standards and data-centre designs evolve. The products that customers prioritise by the time the acquisition closes may differ from today’s mix.

For customers, the central question is whether the combined business can offer better performance, reliability and total cost than purchasing components separately. A wider portfolio could make it easier to coordinate processors, switching and optical links, but customers will want evidence that the products work well together and remain available at scale. Integration can create efficiencies, yet it can also make customers more dependent on one supplier for several critical components.

The transaction is part of a broader shift in how the AI hardware industry is organised. Competition is no longer limited to the processor that performs a calculation. Memory, networking, power delivery, cooling and optical interconnects all influence the performance and operating cost of a large AI system. Companies that can help customers coordinate those layers may gain a larger role in infrastructure design, but the market remains exposed to the spending plans of a relatively small number of hyperscale buyers.

The next milestones are regulatory filings, shareholder votes and further details about integration, expected savings and product plans. Investors will also watch Marvell’s financial performance and the pace of AI-related orders as evidence of whether demand supports the valuation. Until the deal closes, neither company should be treated as already operating as a single business.

Marvell is making a large wager that faster optical connectivity will become indispensable as AI systems grow. Lumentum brings technologies that could strengthen that position, but the size of the deal means execution will matter as much as the underlying market opportunity. The acquisition’s eventual value will depend on whether the companies can turn a broad technology fit into products customers buy, margins they can sustain and returns that justify the stock issued to complete the transaction.