Why the future of network security is the convergence of SASE and firewalls

In technology, the pendulum rarely stays at one extreme for long. When Secure Access Service Edge (SASE) entered the enterprise networking lexicon a few years ago, the industry quickly pivoted to it. The prevailing consensus held that traditional on-premises security hardware was headed for extinction. Industry watchers envisioned a world where even corporate headquarters and massive branch offices would shed their physical appliances, plugging directly into cloud-delivered SASE POPs. It was a clean, compelling story: Offload processing to the cloud, eliminate local box management, and let security follow the user wherever they go.

Except, like many tech hype cycles, reality intervened. This is a movie we’ve all seen before. The cloud was supposed to eradicate on-premises workloads, but that didn’t happen. Voice was declared dead at one point, but it’s alive and kicking. And software was going to eat the world, and all that would remain is commodity hardware.

We are now seeing the SASE pendulum swing in a different direction. As enterprises aggressively roll out edge computing architectures, IoT deployments, and, in particular, real-time and agentic AI applications, the physical edge isn’t disappearing—it’s becoming dramatically more complex. Today, we are seeing a clear course correction in network security strategy. The debate is no longer about choosing between a SASE-first cloud architecture and an on-premises firewall model. Instead, we are entering an era driven by the unavoidable convergence of firewalls and SASE into a single, cohesive framework.

The AI and edge reality check

Why is the cloud-only SASE model showing its limitations at the physical site? The answer largely boils down to the physics and economics of data traffic, especially in the age of edge computing and AI.

As executive voices across the tech industry, most notably Nvidia’s Jensen Huang, have repeatedly emphasized, we are moving into the era of inference. The massive models trained in…