AI DC

  • The Three Puzzles of AI Data Center Fabric (Scale-Up, Out, Across)

    In our last post, we discussed the story of the birth of Ultra Ethernet Transport (UTE), which emerged as an alternative to the sluggish TCP and cumbersome RoCEv2 to handle the explosive traffic of the AI era. However, many of you often ask, “The concept is great, but is this really a feasible technology? When will it be implemented in our data centers?” Global consulting group McKinsey &…

  • [Discover 2026 In-depth Analysis #1] Next-Generation AI Data Center Autonomous Driving Operations Strategy

    In the previous post, we briefly touched upon 'Autonomous Driving Networks,' the core theme of the HPE Discover 2026 keynote. Today, we intend to delve deeply into 'AI Data Center Networking,' the area where the most intense technological innovation is taking place. Today, enterprise data centers have truly reached an operational 'Breaking Point.' Amidst the flood of data and complexity, engineers…

  • The Birth of Ultra Ethernet (UET) and Next-Generation AI Fabric

    When training AI models, networks are no longer mere pipes. In a collective communication environment where tens of thousands of GPUs exchange data, a single packet loss becomes a critical bottleneck that delays the entire Joint Training Completion (JCT). At last year’s HPE Discover Barcelona, HPE and VIAVI Solutions presented a remarkable demonstration signaling a fundamental transformation of Ethernet…

  • [AI Data Center] The Heart of Massive Data Pipelines: Why Routers Again?

    With the recent surge in generative AI led by ChatGPT, the construction of 'AI Data Centers (AIDCs)' has emerged as a major topic in enterprise IT. When thinking about AI infrastructure, it is easy to focus only on tens of thousands of GPUs and the internal switch fabric connecting them. However, there is a key point that should not be overlooked: the fact that AI never remains confined 'inside' the data center. Successful AI…