For decades, computer networking has basically had one job: move the traffic.
Your switches, routers, fiber, access points and firewalls form the digital highway underneath virtually everything a business does. When traffic backs up, something fails or performance starts deteriorating, the network produces telemetry and alerts – and humans figure out what happened.
That model is beginning to change. The next generation of networks will not simply carry artificial intelligence. Artificial intelligence will increasingly help operate the network itself.
The network is getting better at understanding what is happening inside the network.
IMAGINE A NETWORK WITH ITS OWN TRAFFIC CONTROLLER
Picture a busy six-lane interstate. Traditional networking is incredibly good at moving cars fast. But when traffic starts stacking up three miles ahead, the equipment normally follows rules engineers configured ahead of time.
Now imagine an intelligent traffic controller watching every road, entrance ramp, accident and traffic pattern at once. It sees congestion developing, spots a route that is degrading, notices another path has room, and starts making adjustments.
That is increasingly what networking manufacturers are building.
Broadcom’s Tomahawk 6 switch family delivers 102.4 terabits per second of Ethernet switching capacity on a single chip. Its Cognitive Routing 2.0 features include advanced telemetry, dynamic congestion control, rapid failure detection, global load balancing and adaptive flow control – capabilities designed for enormous AI clusters.[1] Broadcom said in March 2026 that Tomahawk 6 had moved into production-volume shipping.[2]
BUT IS AI ACTUALLY DECIDING WHERE EVERY PACKET GOES?
Not exactly – and this distinction matters.
Modern switching chips make forwarding decisions extraordinarily quickly. You do not want ChatGPT sitting inside a switch pondering the existential meaning of Packet #7,482,193 before deciding which port it should leave.
The switching silicon still does what it does best: move enormous amounts of data at line rate. AI is more useful as the intelligence above and around that machinery – analyzing telemetry, spotting unusual behavior, anticipating congestion, identifying likely failures and deciding what changes should be made.
Cisco now describes its Crosswork platform as an AI-enabled platform for autonomous networking, with capabilities for self-healing operations, risk detection, root-cause analysis and AI-guided closed-loop remediation.[3]
THE PHYSICAL NETWORK IS CHANGING TOO
The change is not stopping at software. AI infrastructure is pushing innovation into switching silicon, network adapters and optics. NVIDIA’s Spectrum-X architecture, for example, uses telemetry-based congestion control and dynamic load balancing so network paths can adapt to real-time conditions rather than relying only on traditional static behavior.[4]
WHY THIS MATTERS NOW
Technology environments are becoming too complex for people to manually watch every component all the time: thousands of devices, millions of conversations, cloud applications, remote offices, wireless networks, cybersecurity systems and increasingly demanding AI workloads.
The natural evolution is infrastructure that increasingly observes itself, detects trouble, understands what is happening, makes adjustments and verifies whether the change actually worked.
We are not at a point where every business network is fully autonomous. But the ingredients are clearly arriving: deeper telemetry, adaptive routing, AI-assisted operations, software-defined control and faster remediation loops.
THE STRATEGIC TAKE
For years, businesses have thought about AI primarily as something they use: an assistant, chatbot or productivity tool.
But another transformation is happening underneath all of that. The infrastructure itself is becoming intelligent.
Networks will increasingly be able to recognize congestion, predict failures, optimize traffic and help repair themselves – sometimes before the people using the network realize there was a problem.
That does not eliminate the need for skilled technology professionals. It changes their role. Instead of chasing every individual alert, teams increasingly define the policies, security boundaries and business rules that intelligent infrastructure operates within.
The network of the future will not just move your data. It may understand what is happening to it while it does.
SOURCES
- Broadcom, “Broadcom Ships Tomahawk 6: World’s First 102.4 Tbps Switch,” June 3, 2025. Broadcom product release.
- Broadcom, “Broadcom Now Shipping World’s First 4 Tbps Switch in Production Volume,” March 12, 2026. Broadcom investor release.
- Cisco, “Crosswork Network Automation,” accessed August 14, Cisco.
- NVIDIA, “Optimize Large-Scale AI Workloads with NVIDIA Spectrum-X,” technical NVIDIA.