Spirent Communications has released its first-ever report on the high-speed Ethernet (HSE) market, revealing that advancements in artificial intelligence (AI) are reshaping data center, telecom, and enterprise networks. AI’s transformative impact on these infrastructures is surpassing that of traditional cloud applications, creating unprecedented demand for higher Ethernet speeds and increased capacity. The report, titled “The Future of High-Speed Ethernet Across Data Center, Telecom, and Enterprise Networking,” draws on insights from over 340 engagements that Spirent supported in 2023, offering a comprehensive analysis of key drivers, market impacts, and predictions for the future of HSE. The report is available for download at Spirent’s website.
According to Spirent’s findings, the growth of AI is accelerating the demand for faster Ethernet solutions and more robust network architectures. In 2023 alone, more than 70 million HSE ports were shipped, and the market is poised for exponential growth, with over 240 million ports expected to be deployed between 2024 and 2026. To meet the requirements of AI-driven applications, data centers and networks are migrating rapidly to 400G, 800G, and even 1.6Tbps Ethernet speeds. In parallel, testing and validation practices are evolving, as AI workloads create unique and significant stress on data center performance, requiring new testing methodologies to ensure reliability and scalability.
Telecom operators, enterprises, and hyperscalers are not waiting for future developments but are instead actively rearchitecting their networks to handle AI’s growing influence. This shift is particularly visible at the edge, where early upgrades in transport networks are becoming necessary. Spirent’s report points to far-edge sites requiring 25-50Gbps, mid-edge sites moving to 100-200Gbps, and near-edge facilities advancing to 400Gbps, with 800Gbps expected to follow in the coming years. Additionally, RoCEv2 (Remote Direct Memory Access over Converged Ethernet) is gaining traction in back-end data centers, enabling low-latency AI interconnect fabrics that are critical for high-performance networking.
• High-speed Ethernet port shipments: Over 70 million ports shipped in 2023, with 240 million projected by 2026
• Growing speeds: The market is rapidly migrating to 400G, 800G, and 1.6Tbps Ethernet to accommodate AI-driven applications
• AI’s impact on infrastructure: AI workloads are reshaping data centers and interconnects, surpassing the demands of traditional cloud applications
• Testing AI networks: New, cost-efficient testing strategies are required to generate realistic xPU workload traffic for AI data centers
• Edge capacity upgrades: Far-edge locations need 25-50Gbps, mid-edge sites require 100-200Gbps, and near-edge facilities are migrating to 400Gbps with a roadmap to 800Gbps
• RoCEv2 adoption: The rise of Remote Direct Memory Access over Converged Ethernet (RoCEv2) is facilitating low-latency AI networking in back-end data centers
Further insights from the report indicate that the demand for Ethernet speeds of 800G and beyond will soon be driven by the need for faster AI inference, particularly at the network edge. Edge locations, where significant amounts of AI traffic will be processed, are expected to experience early capacity upgrades to keep pace with the evolving AI ecosystem. Predictions suggest that even smaller edge locations could require speed grades of 25-50Gbps, while larger edge sites will need 100-200Gbps, and near-edge sites may push beyond 400Gbps. This rapid acceleration in demand for high-speed Ethernet will likely lead to a faster refresh cycle, potentially introducing 1.6Tbps Ethernet technology within the next few years.
In addition, the report points to the increasing adoption of RoCEv2 as a crucial enabler of AI networking. RoCEv2 technology allows for direct memory access between devices over Ethernet, improving performance and reducing latency in AI interconnect fabrics. As AI models continue to grow in complexity and size, technologies like RoCEv2 will be vital for maintaining the high levels of performance required by modern data centers.
• AI inference: Significant AI traffic will be processed at the edge, leading to early capacity upgrades in access and transport networks
• Ethernet refresh cycles: Faster refresh cycles are expected as 1.6Tbps Ethernet technology emerges to meet growing AI demands
• RoCEv2 for AI fabrics: Adoption of Remote Direct Memory Access over Converged Ethernet (RoCEv2) is increasing in data center interconnect fabrics to support AI workloads
• Impact on telecom and enterprise networks: Telecom operators are adopting new strategies to meet AI-driven requirements, while enterprises are not waiting for future developments to start implementing next-gen networking solutions
“As the market focuses on AI’s potential, the demand for faster speeds and innovation has never been greater,” said Aniket Khosla, Vice President of Wireline Product Management at Spirent. “AI is driving an inflection point in the market, and there is strong demand to understand and get ahead of the trends shaping this transformation.”
The Spirent impact report The Future of High-Speed Ethernet Across Data Center, Telecom, and Enterprise Networking is available for download now.