As competition in artificial intelligence (AI) data centers intensifies, optical communication technology is emerging as a core element of next-generation infrastructure. The analysis suggests that beyond competition centered on GPUs and memory, optical networks that efficiently connect large-scale AI clusters have become the decisive technology determining performance.

Lumentum recently highlighted AI infrastructure trends, emphasizing that AI systems must encompass not only computing but also data delivery, power efficiency, and stable connection structures, much like the human body. The company explained that optical interconnect technology serves as the "circulatory system" of AI infrastructure, forming the core foundation that supports massive data movement.

◆ Hyperscaler Investment Expansion Drives Optical Communication Demand

The expansion of AI data center investment is also accelerating the growth of the optical communication market. Major U.S. hyperscale cloud operators are reportedly planning to increase their capital expenditures to up to $725 billion by 2026. This represents a significant increase from the previous year's record investment levels, and demand for optical communication equipment is expected to rise rapidly alongside data center expansion.

Market research firm LightCounting predicted that the Ethernet optical transceiver market will maintain strong growth through 2026. In particular, demand for indium phosphide (InP)-based EML lasers and laser chips continues to outstrip supply capacity, with some product categories experiencing demand that significantly exceeds supply.

◆ Lumentum Projects Approximately 85% Annual Growth for InP-Based Optical Products

At the OFC 2026 investor briefing, Lumentum projected that demand for InP-based optical products, including EML, CW, and UHP lasers for AI data centers, will grow at an annual rate of approximately 85%. The company explained that this represents not merely an increase in optical transceiver sales, but a "structural growth" that requires expanded laser chip and wafer production capacity.

As AI models grow larger, data center architectures are also evolving rapidly. Connecting thousands or more GPUs and accelerators as a single system requires ultra-low latency and high-capacity data transmission. The role of optical technology is expanding across all domains, from scale-up and scale-out to inter-campus connectivity.

◆ "Interconnect, Memory, and Architecture Are as Important as Computing"

At the OFC 2026 event, Richard Ho of OpenAI stated that in the AI era, interconnects, memory, and system architecture have become as important as computing performance, and that optical technology is an essential connectivity solution in terms of bandwidth and power efficiency.

Lumentum CEO Michael Hurlston said, "Lumentum's inclusion in the NASDAQ-100 is an example that demonstrates the growing importance of optical technology in AI-based infrastructure," adding, "As speed and bandwidth within data centers continue to increase, demand for optical solutions is expanding in tandem."

The industry expects that future AI data centers will evolve toward combining multiple optical technologies tailored to diverse environments — such as scale-up, scale-out, and long-haul connectivity — rather than being built on a single optical technology. Accordingly, the importance of optical components across the board, including optical transceivers, lasers, optical semiconductors, and silicon photonics, is expected to grow even further.

[This article was written with the assistance of AI. This article is not intended to solicit investment, and the final decision and responsibility for investments lie with the investor.]