Amazon Web Services (AWS) has modified a key component of its infrastructure strategy by making a flat network design the default for most new data center installations worldwide, with AWS makes flat network design now central to its approach. This shift meaningfully changes the long-standing hierarchical “fat-tree” model that has defined large-scale cloud networking for decades.

Since late 2024, the new system, called Resilient Network Graphs (RNG), has been progressively implemented in AWS facilities. The architecture is already commonplace for the majority of new data center designs, according to AWS, reflecting a larger movement to enhance performance, resilience, and efficiency in hyperscale environments.

Traffic flows up and down a hierarchy of routers and switches in traditional data center networks, which are based on layered topologies. Although reliable, this model requires a large number of networking devices and can create congestion points. According to AWS, RNG replaces it with a flatter architecture that connects routers in a more distributed pattern, allowing data to travel through multiple alternative routes.

According to the company, the concept reduces networking hardware requirements by over 69%, increases throughput by up to 33%, and cuts energy usage by nearly 40%. These improvements are made possible through a quasi-random connection approach that reduces bottlenecks and improves overall network resource utilization.

AWS introduced supporting technologies, such as a passive optical device called ShuffleBox that helps manage complex interconnections without adding operational overhead and a routing system called Spraypoint, to make such a design feasible at scale.

AWS says the architecture was first validated in production environments in Europe before being expanded globally. Importantly, the move does not require any changes from customers, as the network operates entirely beneath existing cloud services.

In an environment where network efficiency has become just as important as compute and storage capacity, hyperscale cloud providers are under increasing pressure to optimize infrastructure for AI workloads and growing data traffic demands.

Thus, Business Fortune is of the view that AWS’s design shift signals a major step toward more efficient hyperscale networking infrastructure.