Story
Nov 6, 2026
Marvell’s $115 Billion Lumentum Bet Tests Faith in AI’s Optical Future
Marvell sees optical links as the next critical layer of AI infrastructure, while the market response shows that investors are not yet convinced the enormous spending cycle will deliver without risk.
Marvell Technology said Monday it would acquire optical-components maker Lumentum in an all-stock transaction valued at roughly $115 billion, putting one of the technology sector’s largest AI infrastructure wagers behind the increasingly crucial task of moving data between chips.1
The logic is straightforward: as companies assemble ever-larger systems to train and run large language models, conventional electrical connections face mounting limits. Marvell argues that optical technology is becoming central to linking those systems across racks, facilities and regions.
“AI infrastructure is undergoing a fundamental architectural shift toward optical connectivity,” Marvell Chairman and CEO Matt Murphy said, calling it a once-in-a-generation opportunity. He added that companies moving early could “set the standard for decades to come.”2
The deal gives Lumentum shareholders 4.5 Marvell shares for each share they own. Because there is no cash component, its ultimate value will rise or fall with Marvell’s stock; the companies expect to close in the first half of 2028.2
Marvell had already signaled its direction earlier this year by completing a multibillion-dollar acquisition of Celestial AI, another company focused on optical interconnect technology. The Lumentum purchase greatly expands that strategy, combining Marvell’s electrical technology with Lumentum’s optical components as demand for AI connectivity accelerates.2
But the market offered an immediate note of caution. Marvell shares fell 3% and Lumentum slipped about 4% before the opening bell, even as Lumentum stock had more than doubled during 2026. The contrast captures the tension behind the transaction: Marvell is buying into a future it considers inevitable, while investors are still weighing whether the AI infrastructure boom can justify such a vast price tag.2