What’s next for Azure networking?
Network World ·

AI workloads change how traffic moves inside a data center. GPUs exchange data with other GPUs, servers, and storage, which shifts traffic from north-south to east-west . AI agents add another layer. They span compute, GPUs, the network, and data centers, and demand can move from one place to another overnight. At its Ignite conference this week, Microsoft detailed Azure networking updates for this shift, including new services for enterprises as well as how Microsoft itself is using agentic AI for networking. (It also previewed Windows Server 2028 , its largest platform update since Windows Server 2025.) Networking-related announcements made at Ignite include: Azure Multicloud Interconnect: Provisions cross-connects between Azure and other clouds through APIs. Amazon Web Services (AWS) is generally available. Google Cloud Platform (GCP) is in preview, as is Azure Last Mile Interconnect with Lumen. AT&T is coming soon. Cross-cluster networking for Azure Kubernetes Service (AKS) : Peers clusters through Azure Kubernetes Fleet Manager with failover and one network policy. Azure Route Server route maps: Filter and summarize routes and apply autonomous system (AS) path prepending. Agent Grid: Enforces identity-based microsegmentation, isolation, and traffic policies for AI agents on the network. Azure Firewall AI add-on: Runs two inline small language models on Azure Firewall Premium to detect malware and exploit requests. Azure Adaptive Virtual Patching: Maps published vulnerabilities to workloads on virtual machines and applies inline mitigations until the actual patch arrives. Microsoft points to the shifts in traffic , location, and security that AI creates. “So, with all of these shifts, the network is no longer about connectivity. It is really becoming the intelligent control for the agentic world,” said Igor Sakhnov , corporate vice president and general manager of Azure Networking, in a session at Microsoft Ignite. Azure flattens its fabric and adds hollow core fiber As part of the AI shift, Microsoft is moving its data center network from three tiers to a flatter two-tier design. In the older design, servers sit under fixed branches with shared uplinks. Sakhnov explained that a path that took five hops can now take three. He added that servers can push full throughput without being limited by the network, and that the new fabric is more than 20% more power efficient than the old one. “What you also probably know is that in the current AI world, power is the scarcest resource to get. So, the network giving power back to compute and GPU is a huge deal,” Sakhnov said. The other shift that Microsoft is taking with its own networking connectivity is cabling. “Step outside of the data center, and physics becomes the next constraint,” he said. “For 50 years, we used to go and run light through the solid glass in the fabric networks.” Microsoft has now shifted to using hollow core fiber that moves light through air instead of solid glass. Microsoft says this cuts latency by about one-third and requires fewer amplifiers. Data centers can sit farther apart without a latency penalty, which matters because power allocation heavily drives site selection. Microsoft runs about 15,000 kilometers of hollow core fiber today and plans 10 times more by the end of 2028. width="512" height="304" sizes="auto, (max-width: 512px) 100vw, 512px"> Microsoft GPU capacity is also moving outside Azure regions to partner data centers, extended zones, and enterprise locations. Microsoft connects that capacity to Azure through AI points of presence (PoP). Private traffic runs from each PoP over Azure ExpressRoute, and purpose-built security devices attached to the PoP provide a security overlay. “Outside of the region does not mean outside of the network, and does not mean outside of control,” Sakhnov said. Azure runs its network with more than 100 AI agents While extending the network to support AI is critical, so too is using AI to keep the network running. Microsoft operates about 1.6 million network-attached devices. More than 100 AI agents already work on that network. Microsoft calls them digital coworkers, each with its own identity and a manager. Over the last year, the agents handled more than 30,000 cases and reduced the backlog by about three quarters. Sakhnov detailed one agent, named Miles Dyson, that specializes in fiber cuts. ( Editor’s note: If that sounds familiar, it’s the name of the fictional founder of Skynet from the Terminator science fiction franchise ). On June 7, an excavation cut a fiber east of Jakarta. About 20 minutes later, Miles had used device telemetry to locate the fault to within a meter. It negotiated two-hour updates with the local partner repair crew and followed the crew through the repair. When the crew did not return in the expected time, Miles brought Microsoft engineers onto the bridge. After the crew resolved the problem, Miles verified the link from both ends and closed the ticket. The agents absorb repetitive work as the network grows. Microsoft says network scale expands exponentially while manual work stays flat. “And that also means that without AI, we are not going to be able to do it. AI has to go and carry the load,” Sakhnov said. Azure targets operations and security gaps in AI networks Beyond improving its own network, Microsoft also announced customer-facing services that improve networking for the AI era. Many customers already run multiple virtual networks (VNets) that peer with each other and connect to on-premises sites through ExpressRoute and VPN gateways. “So, the problem is not about how to set up this network. It’s how you operate it,” said Narayan Annamalai , vice president of product management for Azure Networking. Azure Multicloud Interconnect addresses one part of that. Annamalai said a cross-connect between clouds that once took days or weeks now takes minutes. Setup runs through APIs. Each connection is quad resilient, with two sites and two connections at each site. IPv4 address space is another limit. Virtual network peering with IPv6 is in public preview. Teams reuse the same IPv4 ranges in every VNet and give each dual stack VNet unique IPv6 addresses. Traffic between VNets runs over IPv6. Annamalai said Microsoft runs its own network this way. Route maps for Azure Route Server handle routing at scale. Teams can filter routes, summarize on-premises routes into fewer prefixes, drop routes, and apply AS path prepending to select a source. Cross-cluster networking for AKS removes the manual stitching between clusters. Teams peer clusters through Azure Kubernetes Fleet Manager. Traffic shifts to the other cluster if backends fail, and network policy is defined once. Security controls move into the network as well. Agent Grid enforces identity-based policy for AI agents on the network. Agent code does not change. Each agent connects out to other agents, models, Model Context Protocol (MCP) servers, and the internet. “If you want to protect each one of these, we have to create gateways at every single point. It has become very complex,” Annamalai said. The new services address connecting, scaling, and securing networks for AI workloads. “You can have the best AI model in the world, but if your network goes down, it’s an expensive way of saying 404 not found,” Annamalai said.