Marvell Bets on Photonics to Connect Next-Gen AI Systems
The chipmaker projects $1 billion in annual revenue from optical interconnect technology by 2029 as data centers shift away from copper.

Marvell pivots to optical interconnects for AI infrastructure
Marvell Technology Inc. reported second-quarter revenue of $2.74 billion, up 37% year-over-year, with data center sales climbing 46% to $2.17 billion—now representing roughly 80% of total company revenue. But the semiconductor company's most significant bet may be on the technology that connects AI chips rather than the chips themselves.
The company is positioning itself to capitalize on a fundamental infrastructure shift in artificial intelligence systems. As AI models scale, traditional copper connections face limitations in distance, power efficiency, and bandwidth. Data centers are responding by adopting optical connections that use light to transmit information between components.
Celestial AI acquisition targets $1 billion opportunity
Marvell has made substantial investments to capture this emerging market, most notably through its acquisition of Celestial AI. That deal brought Photonic Fabric technology into Marvell's portfolio, designed to link large-scale AI systems more efficiently than conventional methods.
Management projects the photonics business will reach a $500 million annualized revenue run rate by the end of fiscal 2028, doubling to $1 billion the following year. The company forecasts third-quarter revenue of approximately $3.15 billion, driven by custom AI chips, optical products, and networking equipment.
Why it matters
Marvell's photonics strategy addresses a critical bottleneck in AI infrastructure that becomes more acute as models grow larger and more complex. The company already serves major AI customers for semiconductors, giving it established relationships to cross-sell optical interconnect solutions. If the technology transition unfolds as anticipated, Marvell could capture significant revenue from the physical layer connecting AI accelerators—a market adjacent to but distinct from the chip wars dominating current AI infrastructure discussions. The projected timeline also suggests meaningful revenue contribution within the next three to four years, not a distant research project.
Execution risk remains
While Marvell holds existing customer relationships in AI infrastructure, the photonics revenue targets are projections rather than certainties. The company will provide additional details at its investor day scheduled for October 6.
The strategic logic is straightforward: Marvell supplies the chips powering AI workloads and now aims to provide the optical fabric connecting those chips at scale. Whether the technology and market timing align with management's revenue projections will determine if this bet pays off as handsomely as the company's existing AI chip business.
These details were first reported by AI Watch.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call