It’s Only Common Sense: AI Makes Preparation Easier and Excuses Harder
Happy’s Tech Talk #50: Start Your Digitalization Strategy With Automation
The Chemical Connection: A Few Things That Still Scare Me
TrendForce's latest AI infrastructure research indicates that the continued expansion of AI server clusters and surging accelerator bandwidth requirements are fueling rapid growth in the global AI networking and interconnect market. Market revenue is projected to climb from US$37.9 billion in 2024 to US$279.8 billion in 2027, representing a CAGR of 95% over the period.
Back-end network switches and optical transceivers have traditionally been the primary growth engines. However, market momentum is increasingly extending to high-speed accelerator interconnects and network interface cards. Looking ahead, the shipment mix and network topologies adopted by different AI chip platforms will play a pivotal role in determining both the market's overall size and how value is distributed across the supply chain.
Switches are expected to remain the largest revenue contributor, generating US$105 billion in 2027 and accounting for nearly 38% of the market. Network interface cards and data processing units (NICs/DPUs) are also poised for robust growth, supported by higher deployment volumes and bandwidth upgrades across next-generation platforms. Revenue from these products is forecast to more than double YoY to US$57.4 billion in 2027.
Meanwhile, continued migration toward higher-speed specifications is expected to lift pluggable optical transceiver revenue to US$84.1 billion in 2027. Together, switches, NICs/DPUs, and pluggable optical transceivers will account for approximately 88% of total AI networking and interconnect revenue.
Among interconnect segments, high-speed connections between AI accelerators, known as scale-up networking, are seeing the fastest revenue growth. As GPU platforms evolve toward rack-scale architectures, accelerator interconnects are expanding beyond server motherboards into dedicated switch trays, creating additional demand for switching equipment and copper cables.
Based on AI cluster deployment estimates, TrendForce projects that scale-up networking's share of AI networking and interconnect revenue will grow from 11% in 2024 to 24% in 2027. Copper cables are expected to remain the dominant interconnect medium within GPU racks in 2027. Google, however, is already scaling its TPU clusters with optical interconnects through its Inter-Chip Interconnect (ICI) technology, alongside optical circuit switches (OCS).
Scale-out networking, which connects servers across AI clusters, remains the largest interconnect segment by revenue. Ongoing bandwidth upgrades are expected to shift the primary revenue driver for pluggable optical transceivers from 800G to 1.6T products in 2027.
Despite the continued decline in the average selling price (ASP) of 1.6T transceivers as production scales up, these products are still expected to command more than twice the price of their 800G counterparts. Consequently, the growing share of higher-speed products in the shipment mix will continue to bolster overall market revenue. TrendForce forecasts that pluggable optical transceiver revenue will surge 76% YoY in 2027, outpacing shipment growth, with 1.6T products contributing approximately 56% of pluggable transceiver revenue.
Importantly, growth in AI chip shipments will not translate proportionally into networking revenue. The networking value associated with each accelerator varies according to the interconnect topology, switch configuration, and NIC bandwidth requirements of the underlying AI platform.
Emerging optical connectivity technologies and network architectures are also lifting overall market revenue. Co-packaged optics (CPO) and OCSs, for example, address distinct interconnect requirements. By integrating optical engines and switch chips within a single package, CPO shifts a portion of the optical component value traditionally captured by pluggable transceivers into switching equipment. OCS, meanwhile, enables large-scale cluster interconnectivity through optical circuit switching.
Overall, TrendForce expects the distribution of value across the AI networking and interconnect supply chain to continue evolving as AI platform configurations diversify and optical interconnect architectures advance.
