Marvell has agreed to acquire Lumentum in an all-stock deal with an implied enterprise value of $115 billion. The real story is what happens between AI chips: moving massive amounts of data quickly enough to keep them working.
Marvell designs custom AI silicon, networking chips and high-speed interconnects. Lumentum makes lasers, optical transceivers and optical circuit switches that send data using light. Putting those technologies together could help AI infrastructure builders connect far larger systems without letting the network become the bottleneck.
Why Light Matters More Now
AI systems are expanding from servers into clusters of racks and even multiple facilities. That creates an enormous traffic problem: processors constantly exchange data, and moving it can consume both time and energy.
Copper cables remain useful, but their reach and power demands become harder to manage at higher speeds. Optical connections can carry large volumes of data over longer distances, making them increasingly important as AI clusters grow.
The companies want to go further than pairing existing products. They plan to design the electrical and optical sides together, from the silicon that processes signals to the lasers and components that transmit them.
They Have Already Worked Together
This is not their first technical collaboration. In March, Marvell and Lumentum demonstrated their products working together for optical circuit switching in AI data centers.
The demonstration linked Marvell optical signal-processing technology with Lumentum’s R300 optical circuit switch. The idea is to create direct paths for data between endpoints, reducing intermediate processing and helping cut latency and power consumption.
That earlier project offers a glimpse of what the combined companies want to build at a much broader scale: systems designed as a whole rather than collections of separately optimized parts.
What Comes Next
Marvell says the combination would cover connections inside and between AI data centers, alongside specialized manufacturing for the lasers and photonic components those systems need. Its goal is to deliver more fully tested, integrated optical solutions to customers.
The transaction still requires shareholder and regulatory approvals and is expected to close in the first half of 2028. Until then, the proposed combination is a plan, not a completed merger.
The bigger question is whether the companies can make connectivity a defining part of AI system design, rather than a collection of components chosen after the compute hardware.
Get stories like this in your inbox every week. Subscribe to The Week in Perspective.