Uploaded June 2026 | Updated September 2026, 3 hours ago
Every major AI capability—from distributed training to immediate inference—depends on the network beneath it, and that dependency is structural rather than incidental. As systems scale, limitations in the current internet architecture are becoming harder to ignore, with address exhaustion, operational complexity, and fragmentation creating constraints in production environments. If the future of AI is inseparable from the future of the internet, what does that future internet need to look like? Can it be built on a dual-stack architecture not designed for global, AI-driven workloads? This session examines whether the successful global deployment of a single-stack IPv6 internet is not simply a network-engineering question, but a condition for the next phase of AI development.
Every major AI capability—from distributed training to immediate inference—depends on the network beneath it, and that dependency is structural rather than incidental. As systems scale, limitations in the current internet architecture are becoming harder to ignore, with address exhaustion, operational complexity, and fragmentation creating constraints in production environments. If the future of AI is inseparable from the future of the internet, what does that future internet need to look like? Can it be built on a dual-stack architecture not designed for global, AI-driven workloads? This session examines whether the successful global deployment of a single-stack IPv6 internet is not simply a network-engineering question, but a condition for the next phase of AI development.
