Marvell to acquire Lumentum for $115 billion in major AI infrastructure deal

Marvell Technology has signed a definitive agreement to acquire Lumentum in an all-stock transaction valued at $115 billion. The merger brings together two powerhouses in the data center optics market, solidifying Marvell’s position as a leader in AI connectivity.

• Shareholders of Lumentum will receive 4.5 Marvell shares for each share held, representing a 23.9% premium.
• The deal merges Marvell’s optical DSPs and SerDes with Lumentum’s lasers, transceivers, and manufacturing footprint.
• Expected to close in the first half of 2028, pending regulatory and shareholder approval.

Revolutionizing AI Connectivity

As AI infrastructure scales, copper interconnects are reaching physical limitations, driving an industry-wide transition to optical solutions. By integrating Lumentum’s photonic capabilities, Marvell can now provide end-to-end solutions, from high-speed electronic signal processing to optical transmission, creating a unified architecture for AI clusters.

Lumentum’s expertise in indium phosphide (InP) manufacturing is a strategic pillar for this acquisition. InP is essential for creating high-performance lasers and optical components that are critical for transmitting massive data volumes at scale, far exceeding the capabilities of traditional silicon-based optical designs.

A Strategic Pivot to Hybrid Manufacturing

Historically a fabless company, Marvell will transition into a hybrid model by acquiring Lumentum’s internal fabrication facilities. This gives Marvell vertical control over the manufacturing of its optical components, offering a competitive advantage against rivals like Broadcom, which already maintains an integrated approach to electronic and optical hardware.

Impact on the Technology Ecosystem

For the industry, this consolidation signals the maturation of the optical interconnect market. With Marvell already serving major hyperscalers like Google, AWS, and Meta, this acquisition ensures that the underlying plumbing of AI — the high-speed data highways connecting thousands of accelerators — will be optimized as a single, cohesive system, ultimately accelerating the deployment of next-generation AI workloads.