Artificial intelligence is changing enterprise networking in two fundamental ways. While much of the industry’s attention remains focused on building networks capable of supporting AI workloads, an equally significant transformation is occurring in how networks themselves are designed, managed, and operated.
At HPE Discover 2026, HPE expanded its vision for self-driving networks by extending AI-powered operations deeper into the data center while continuing to integrate networking with a broader portfolio that includes GreenLake Intelligence, OpsRamp, Morpheus, and Apstra. Together, these capabilities represent an evolution toward more autonomous infrastructure management designed to reduce operational complexity, improve reliability, and enable IT teams to focus on higher-value initiatives.
During a recent theCUBE Research conversation, Ben Baker, Director of Data Center Networking at HPE, discussed how AI is reshaping both network operations and AI infrastructure itself.
Complexity Has Outgrown Traditional Operations
For many enterprise IT organizations, the challenge is no longer simply keeping up with network growth. Modern environments now span physical infrastructure, virtual networks, cloud services, applications, multiple vendors, and increasingly AI workloads. Traditional operational models struggle to keep pace.
As Baker explained, “Operators are drowning in data, but starved for insights.” That observation captures one of the defining challenges facing enterprise networking today. IT teams have access to unprecedented volumes of telemetry but often lack the contextual intelligence needed to quickly identify root causes.
Baker grouped today’s operational challenges into three primary areas:
- Limited operational insight
- Insufficient speed and agility
- Reliability risks created by manual operations
Collectively, these issues force many infrastructure teams into a reactive operating model. As he noted, “Data center network operators are usually in reactive mode. They’re firefighting, constantly dealing with these emergencies instead of focusing on proactive strategic initiatives that really matter to your CIO.”
That finding closely aligns with theCUBE Research data. Organizations that have adopted AI-powered operations consistently report that while AI initially helps them identify and resolve problems faster, the longer-term benefit is giving engineering teams time to work on projects that directly support business objectives.
AI Workloads Raise the Bar Even Higher
Operational complexity is only increasing as organizations begin deploying AI applications and agentic workflows. Unlike traditional enterprise applications, AI environments introduce entirely new performance requirements around latency, congestion management, scale, and deterministic network behavior.
According to Baker “AI will continue to massively increase network traffic… especially as we get more and more agents running 24/7.” He also emphasized that AI inference changes user expectations dramatically, “Even milliseconds are going to add up and they’re going to matter.”
This reflects a broader industry shift. As AI moves from experimentation into production, networking increasingly becomes part of the application experience itself rather than simply the transport layer underneath it.
From AIOps to Self-Driving Networks
One of HPE’s primary themes at Discover was extending the AI operational capabilities many customers already associate with Mist and Marvis into the data center. Rather than describing self-driving networking as a future vision, Baker argued that many of these capabilities are already delivering value stating, “This is real. It’s working at our customers today.”
A key differentiator, according to Baker, is the use of the Apstra graph database, which provides contextual understanding of relationships throughout the infrastructure. He explained, “Context is very important. It’s foundational to what we do. It’s ingrained in the architecture, and it can’t be bolted on after the fact.” He added, “Graph databases capture and maintain information about the relationships among the nodes in a network. This gives us context for all the data.”
That context enables AI models to correlate operational events, application behavior, and network state more effectively than isolated telemetry alone.
Predictive Operations Replace Reactive Firefighting
Perhaps one of the most compelling discussions centered on predictive maintenance. Rather than waiting for components to fail, HPE’s Data Center Assurance platform continuously analyzes telemetry to identify devices likely to fail and prioritizes them by business impact.
Instead of dispatching technicians only after outages occur, operations teams can proactively schedule maintenance based on confidence levels and projected service impact. As Baker summarized, “This makes operations management incredibly efficient. Without it…you’re in firefighting mode, just reacting and responding to failures.”
This represents an important evolution in enterprise operations. AI is no longer simply helping administrators diagnose problems faster; it is increasingly helping them avoid problems altogether.
Integration Delivers Greater Operational Value
Another notable announcement at Discover was the pace at which HPE has integrated technologies following the Juniper acquisition. Rather than focusing solely on product consolidation, HPE appears to be emphasizing operational integration across networking, cloud management, automation, and IT operations.
Baker described this as creating “growth synergies” that extend well beyond cost savings. He noted: “HPE is unique in the industry in being able to bring together compute, storage, networking, hybrid cloud to deliver full-stack cross-domain IT infrastructure solutions.”
Two recently announced integrations illustrate this approach. Apstra Data Center Director now integrates directly with Morpheus, allowing virtual networking to be automatically provisioned alongside virtual machines. Similarly, Data Center Assurance integrates with OpsRamp to correlate infrastructure issues directly within broader IT operations workflows.
While these integrations may appear incremental individually, collectively they reduce manual handoffs between teams, shorten troubleshooting cycles, and simplify operational workflows.
Networking for AI Continues to Accelerate
While much of the discussion focused on AI operating the network, HPE also highlighted infrastructure designed specifically for AI workloads. Among the most significant announcements was the industry’s first fully liquid-cooled 1.6-terabit Ethernet switch.
Baker explained that HPE views AI networking across three dimensions:
- Scale-up networking within servers and racks including AMD Helios
- Scale-out networking across AI clusters including the new 1.6 Tb liquid cooled switch
- Scale-across networking connecting multiple AI data centers leveraging PTX routers
The new QFX5250 targets high-performance AI training environments, while the QFX5140 addresses inference deployments. Just as important, Baker highlighted how combining Juniper networking expertise with HPE’s long-standing liquid cooling capabilities accelerated platform delivery.
Why It Matters
The conversation reflects a broader transformation occurring across enterprise infrastructure. Organizations are no longer evaluating networking solely on bandwidth, latency, or hardware specifications. Increasingly, they are asking how networking platforms improve operational efficiency, reduce risk, automate repetitive tasks, and accelerate digital initiatives.
AI is becoming an operational multiplier rather than simply another application workload. Self-driving networks remain a journey rather than a destination, but the direction is becoming increasingly clear. As AI deployments continue to expand, infrastructure operations will need to become more predictive, more autonomous, and more integrated across networking, compute, storage, and cloud management.
For enterprise IT leaders, the question is no longer whether AI will become part of network operations. The more strategic question is how quickly organizations can evolve their operational models to fully capitalize on it.
