TL;DR:
- The data center AI networking market is projected to exceed US$25 billion by 2028.
- Growth is driven by Ethernet, InfiniBand, and 400/800G Optical Transceivers.
- Regional variances in vendor performance observed in the market.
- By 2028, 1 in 5 Ethernet switch ports in data centers to be AI/ML related.
- Record levels of 800 Gbps ports and optics are anticipated in the initial 18 months of AI/ML adoption.
- Hyperscaler CAPEX for 2024 to adjust with AI/ML projects moving towards production.
- Customers exploring diverse networking topologies for AI/ML requirements.
Main AI News:
A fresh report from 650 Group forecasts a robust surge in the data center AI networking market, projecting a staggering growth to over US$25 billion by 2028. This expansion is chiefly propelled by Ethernet, InfiniBand, and 400/800G Optical Transceivers, with the market expected to hit over US$15 billion by 2024, according to the study. The report unveils substantial regional variations in vendor performance, accompanied by numerous record-setting achievements.
Key Insights:
- By 2028, it is anticipated that 1 out of every 5 Ethernet switch ports deployed in data centers will be dedicated to AI/ML and accelerated computing.
- The advent of AI/ML is poised to trigger unprecedented levels of 800 Gbps ports and Optics within the initial 18 months of adoption.
- Hyperscaler CAPEX in 2024 is poised to experience ongoing adjustments as AI/ML initiatives progress towards production stages.
- Customers are actively exploring diverse networking topologies to cater to the evolving demands of AI/ML networking.
“Following a year of remarkable growth in vendor revenue associated with AI/ML networking, 2024 is poised for a significant surge as many Proof of Concepts (PoC) transition into full-fledged production networks,” remarked Alan Weckel, the founder, and technology analyst at 650 Group. Weckel emphasized the substantial bandwidth requirements imposed by AI/ML on networks, identifying it as a primary catalyst for the burgeoning demand in data center switching over the forthcoming five years.
“As AI’s bandwidth demands escalate, the segment of Ethernet switching linked to AI/ML and accelerated computing will evolve from its current niche status to a substantial market segment by 2028,” Weckel continued. “We are on the brink of witnessing record shipments in 800 Gbps-based switches and optics as soon as these products achieve scale in production to meet the demands of AI/ML.”
Conclusion:
The forecasted surge in the data center AI networking market signifies a substantial shift in the industry landscape. With AI/ML driving unprecedented demand for high-speed networking solutions, vendors and stakeholders must adapt swiftly to capitalize on emerging opportunities and meet evolving customer needs. This trajectory underscores the critical importance of innovation and scalability in navigating the dynamic data center ecosystem.
