AI and the Next-Generation Internet: Insights from Alibaba Cloud VP Cai Dezhi at the 2024 Wuzhen Summit

The 2024 Wuzhen Summit’s “Next Generation Internet Forum” featured Alibaba Cloud Intelligent Group R&D Vice President and Infrastructure Network R&D Leader Cai Dezhi, who shared forward‑looking views on the integration of artificial intelligence (AI) with the next‑generation internet.

In his talk, Cai detailed two interaction directions: “Network for AI” and “AI for Network.” He introduced Alibaba Cloud’s innovative network architecture HPN7.0, which upgrades traditional “best‑effort” networks to a predictable, high‑performance fabric, dramatically improving compute resource utilization and providing stable support for large‑model training. He emphasized that achieving AI and network ubiquity requires protocol‑level innovations, such as upgrades to Ethernet, IPv6, and DNS, to make the network more efficient and cost‑effective.

Regarding AI‑enabled networking, Cai highlighted the potential of AI in network operations. Through an AIOps intelligent operations system, large AI models can significantly boost network operation efficiency, reduce costs, and ensure security and stability. AI can also optimize core network components like DNS systems and SDN controllers, enhancing reliability and software quality.

Looking ahead, Cai stressed the importance of openness and collaboration. Alibaba Cloud, as a key member of SONiC and the Ultra Ethernet Consortium (UEC), actively promotes international standardization of high‑performance network protocols and fosters a thriving open ecosystem. Domestically, Alibaba Cloud leads the formation of a high‑throughput Ethernet alliance, innovating with local chip characteristics to further enhance network performance and universality.

In summary, Cai noted that the high cost of AI—driven by expensive chips, networks, and storage—remains a major barrier. He advocated for bottom‑layer protocol innovation to lower technology costs, stronger cooperation with open‑source data‑center ecosystems, and active participation in IETF standards to extend AI‑centric network technologies from data centers to wide‑area networks, thereby building an AI‑ready next‑generation internet.