September 2026 | AI Infrastructure & Optical Communications
As AI models and computing clusters continue to scale, AI infrastructure is facing a new challenge: the bottleneck is shifting from computing power to data communication.
A recent white paper, Photonics at the Speed of AI, published by Yole Group, highlights that AI model sizes are expected to grow by approximately 100× every two years, while computing capabilities have increased by around 3.3× over the same period. As large-scale AI clusters deploy tens of thousands of GPUs, data transfer latency, bandwidth and power consumption are becoming increasingly critical to overall computing efficiency.
From the “Compute Race” to the “Connectivity Race”
Over the past several years, AI infrastructure development has focused heavily on GPUs, processors and computing performance. But as AI clusters continue to expand, simply adding more computing resources is no longer enough.
When tens of thousands of GPUs need to exchange massive volumes of data continuously, overall performance increasingly depends on how quickly and efficiently these devices can communicate.
This is shifting the focus of AI infrastructure from:
Compute → Communication
From chips and servers to switches, optical fibers and optoelectronic components, connectivity technologies within data centers are becoming an increasingly important part of the AI infrastructure stack.
Yole forecasts that the global optical transceiver market could grow from approximately US$10 billion in 2021 to US$112 billion by 2031, representing a CAGR of around 35%. Meanwhile, global hyperscalers are expected to spend approximately US$670 billion in capital expenditure in 2026.
New Opportunities for Optical Modules and Photonics
As AI training and inference clusters continue to expand, traditional electrical interconnects are facing increasing challenges in terms of bandwidth, power consumption and transmission distance.
As a result, technologies including high-speed optical modules, silicon photonics, optoelectronic integration and co-packaged optics (CPO) are attracting growing attention.
Yole expects optical connectivity to move increasingly closer to GPU and CPU packages, with technologies such as LPO, XPO and CPO emerging as important solutions for future high-speed interconnect architectures. The CPO optical engine market is projected to grow from approximately US$600 million in 2026 to US$11.2 billion by 2031.
This evolution means that the growth of AI is creating broader manufacturing demand — not only for more powerful chips and servers, but also for advanced optical modules, PCBs, packaging, testing and electronics manufacturing technologies that can support high-speed data transmission.
AI Infrastructure Is Reshaping the Electronics Manufacturing Ecosystem
From high-speed optical modules and advanced packaging to PCB assembly and automated production, the development of AI infrastructure is accelerating the transformation of the electronics manufacturing ecosystem.
For manufacturers, improving the production efficiency, precision and testing capabilities of high-speed electronic products will be increasingly important in meeting the requirements of next-generation AI infrastructure.
At the same time, as product complexity and production volumes continue to increase, technologies such as SMT, PCB Assembly, automation, machine vision, smart factories, and advanced inspection and testing will play an increasingly important role in this evolving industry landscape.
NEPCON ASIA 2026: Connecting the Electronics Manufacturing Ecosystem for the AI Era
Against this backdrop, NEPCON ASIA 2026 will take place from 27–29 October 2026 at Shenzhen World Exhibition & Convention Center, bringing together companies and professionals from across the electronics manufacturing ecosystem to explore emerging technologies, applications and opportunities in the AI era.
Key sectors covered at the show include:
- PCB Assembly & SMT
- Semiconductor Packaging & Testing
- Optical Communications & High-Speed Connectivity
- Smart Factory & Industrial Automation
- Robotics & Machine Vision
- AI Hardware & Electronics Manufacturing
By connecting electronics manufacturers, equipment suppliers, technology providers and industry professionals, NEPCON ASIA provides a platform to discover advanced manufacturing technologies, identify new supply chain partners and explore business opportunities.
As AI moves beyond simply increasing computing power toward faster and more efficient connectivity, next-generation electronics manufacturing technologies will become an increasingly critical foundation for scaling AI.
NEPCON ASIA 2026 invites electronics manufacturing professionals from around the world to explore the technologies and solutions shaping the next era of AI-driven manufacturing.
Date: 27–29 October 2026
Venue: Shenzhen World Exhibition & Convention Center (Bao’an)
Location: Shenzhen, China
Discover the technologies powering the next era of electronics manufacturing at NEPCON ASIA 2026.
Register Now:
https://ali2.infosalons.com.cn/reg/CNZ26/web/#/en/login?track=962K8L
Please visit official website for more information: https://www.nepconasia.com/en-gb.html
Contact for overseas visitor enquiries
Walden Li
Mobile/WhatsApp/WeChat: +86 13651251335
E-mail: walden.li@rxglobal.com