GF Securities Highlights AI Infrastructure Bottlenecks Shifting from GPUs to Optical Interconnects, Including CPO

June 20, 2026

Futunn News (GF Securities Report) China

Overview

GF Securities has released an analysis report indicating that performance bottlenecks in AI infrastructure are gradually shifting from conventional GPU computing capabilities to optical interconnects, including Co-Packaged Optics (CPO). This change signifies that as AI workloads scale, data transfer between GPUs and communication speeds within data centers are becoming primary factors limiting overall system performance. GF Securities forecasts that this trend will expand growth opportunities for optical communication companies, emphasizing the importance of investing in advanced technologies like CPO.

In Depth

Key Findings

Analysts at GF Securities, in their latest market analysis, have pointed out an evolving nature of performance bottlenecks within artificial intelligence (AI) infrastructure. While GPU computing power traditionally served as the primary constraint, there is now an increasing emphasis on faster and more efficient optical interconnects, particularly Co-Packaged Optics (CPO), with bottlenecks shifting from GPUs to interconnects.

Technical / Clinical Details

Bottleneck Shift: As AI models grow in scale and complexity, mere improvements in standalone GPU performance are no longer sufficient to fully leverage overall system processing capabilities. The ability to efficiently move massive amounts of data between GPUs, or between GPUs and memory, has emerged as a new bottleneck.
Role of CPO: Co-Packaged Optics (CPO) dramatically shortens electrical signal paths by integrating optical modules within the same package as the host ASIC (e.g., GPU). This significantly reduces signal loss, power consumption, and latency, supporting terabit-scale data rates.
Shift to Optical Technology: GF Securities predicts a surge in demand for 800G, 1.6T optical transceivers, and CPO solutions in data centers and HPC (High-Performance Computing) over the next few years. This serves as a direct solution to the limitations of traditional electrical interconnects.
This trend presents significant growth opportunities for companies providing optical fibers, optical transceivers, photonic integrated circuits (PICs), and related packaging technologies.

Background & Context

The explosive growth of AI is fundamentally changing the design philosophy of data centers. AI training GPU clusters require interconnect bandwidth on the petabit per second (Pb/s) scale, which is unachievable with conventional data center networks. GF Securities’ analysis highlights the strategic value of optical technology within the AI ecosystem and its impact on capital markets.

Strategic Significance & Outlook

GF Securities predicts that with the evolution of AI infrastructure, the optical interconnect market will expand significantly, with innovative technologies like CPO playing a central role. This shift will further accelerate investment in the optical communication sector and establish new standards for data center performance and power efficiency. The symbiotic relationship between AI and optical technology is expected to be a key theme driving growth in the high-tech industry for years to come.

Source: https://news.futunn.com/en/post/74434409/gf-securities-bottlenecks-are-gradually-shifting-toward-interconnects-with-cpo

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.