Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

How AI Data Centers Are Shaping The Future Of Ethernet

Mar 31, 2025 - forbes.com
The high-speed ethernet (HSE) industry is experiencing significant growth driven by the increasing demands of AI infrastructures. As data centers invest heavily in GPUs and accelerators, the need for faster and more efficient networks is paramount. Shipments of 400G and 800G ethernet ports are rapidly increasing, with projections indicating that ethernet will surpass InfiniBand in AI clusters by 2028. The evolution of port speeds is crucial, with 800G ethernet expected to dominate AI back-end networks by 2025 and 1.6-terabit ethernet by 2027. This shift is largely due to the extreme demands AI workloads place on data center networks, requiring high scalability, bandwidth, and low latency to optimize performance.

Testing and validation of these networks have become essential as packet loss and latency can significantly impact AI workload efficiency and cost-effectiveness. A 1% packet loss rate can degrade performance by 30% or more, translating to substantial financial losses. To address these challenges, the industry is developing lossless ethernet technologies like the ultra-ethernet transport (UET) specification and new AI network testing solutions. These innovations enable organizations to optimize network performance and validate designs at a lower cost, ensuring that stakeholders across the ethernet ecosystem can capitalize on the AI revolution.

Key takeaways:

  • The high-speed ethernet (HSE) industry is experiencing significant growth driven by the increasing demand for AI infrastructures, with ethernet port shipments expected to overtake InfiniBand by 2028.
  • AI workloads are placing extreme demands on data center networks, leading to the deployment of GPUs and accelerators, and the construction of scalable back-end ethernet networks using spine-leaf architectures.
  • Testing and validation of AI network designs are crucial due to the potential performance degradation caused by packet loss and latency, which can result in significant financial losses.
  • New AI network testing solutions are emerging, allowing organizations to optimize network performance and identify bottlenecks at a lower cost, benefiting stakeholders across the ethernet ecosystem.
View Full Article

Comments (0)

Be the first to comment!