Artificial intelligence is not only transforming enterprise applications but is completely redefining the role of data networks. That was the powerful message from Rami Rahim, former CEO of Juniper Networks and current Executive Vice President of HPE Networking, during his speech at HPE Discover. This year's event had special symbolism: for the first time, Juniper Networks, now fully integrated into HPE, was part of the provider's strategic discourse. Rahim took the stage to issue a direct warning to networking professionals: traditional models are no longer sufficient to support the demands of AI.

The Network: From Silent Infrastructure to Strategic Platform

One of the key points of his speech was the changing role of networking within the organization. According to Rahim, the network has ceased to be an invisible component and has become a critical element for business. To illustrate this, he used a striking example: the Millennium Tower in San Francisco, a skyscraper that began to tilt because its foundations were not prepared for the passage of time. The analogy is clear: current networks are not designed to support the loads of AI.

Massive data movements, continuous inference, real-time response times, and extreme scalability are factors that stress infrastructures designed for a different paradigm. Hence, he warned that “you can invest millions, even billions, in GPUs, but if the network introduces latency or bottlenecks, you are limiting performance.”

For network engineers, the message implies a shift in focus: designing with east-west traffic, minimal latencies, and deterministic behaviors in mind, rather than traditional north-south models. And above all, aligning technical metrics with business outcomes, such as model training times or user experience.

The End of Manual Networks: Towards Autonomous Operations

“The old, static, manual, and reactive model cannot handle today's speed or complexity,” Rahim stated. The alternative lies in AI-native networks with autonomous operations, supported by platforms like Aruba Central and Mist, and technologies like Marvis or the so-called “Marvis Minis,” within an agentic AI approach.

Sunalini Sankhavaram, Vice President of Product at HPE, reinforced this idea by explaining how the company is evolving towards experience-centric AI. This model is based on real usage data —“every user, every minute”— enriched with digital twins and validated with support cases. In one of the demonstrations presented, the system automatically detected degradations in user experience, identified the cause in saturated access points, and applied a correction without human intervention.

“The result is that the network identifies the problem, understands the root cause, and resolves it before the user complains,” Rahim summarized. This approach implies a profound change in the role of professionals: from reactive operators to supervisors of autonomous systems, responsible for defining policies, limits, and SLAs.

Unification: A Single Architecture for the Entire Edge

Another strategic announcement was the progress towards a unified platform that integrates Juniper and Aruba technologies. The vision involves combining Mist and Aruba Central, wired and wireless networks and routing, and a single AI engine. “We are innovating on both platforms to offer a coherent autonomous network experience,” explained Sankhavaram, who highlighted the use of microservices to develop functionalities once and deploy them across the entire ecosystem.

Among the new features is the integration of Marvis into Aruba Central, including functionalities like the Marvis Trust List, which allows automating complete actions —for example, recovering a failed device— without human intervention. In parallel, HPE has already launched hardware prepared for this approach, such as dual-platform access points, and is working to extend its CX portfolio to environments managed by Mist. According to Rahim, “the mission is simple: bring the best innovations to all customers, regardless of the platform they use.”

A Clear Roadmap for Network Professionals

Beyond technology, Rahim's speech sends a direct message to industry professionals: networking enters a new stage where AI will be the central axis. This implies designing networks specifically for AI workloads, adopting real AIOps models (not just monitoring tools), speaking the business language based on user experience, and defining policies for autonomous systems instead of operating manually.

In short, networking ceases to be a support and becomes a direct enabler of business value. And, as Rahim made clear, those who do not evolve with this change risk being left behind: legacy networks are simply not prepared for what is coming.

Designing for Platform Flexibility and Automation

Beyond architecture, HPE's strategy introduces a key concept: platform optionality. That is, assuming that, throughout the hardware lifecycle, the management layer can change. For network teams, this means betting on infrastructures capable of operating with different ecosystems, such as Mist or Aruba Central, without requiring complete replacements.

In parallel, the use of digital twins gains prominence. Rahim and his team emphasized that these should no longer be limited to vendor demonstrations but integrated into real pre-validation and change management processes, through synthetic tests and experience simulations. Another pillar is automation. HPE insists on an API-first approach, which facilitates programmatic access to both data and network actions. In practice, this means professionals will need to strengthen skills in areas such as Python, CI/CD, or infrastructure as code, as discussed in our guide on server virtualization with Proxmox.

Networks and Security: Inevitable Convergence in the AI Era

One of the most repeated messages by Rahim was the growing convergence between networking and cybersecurity. According to the executive, both areas can no longer operate independently. “Attackers are already using the network as their primary tool,” he warned, adding: “And with AI making threats faster, smarter, and more sophisticated, defenders need to use the network as part of the defense.”

This vision aligns with market evolution, where more and more network engineers are taking on security responsibilities. Customers themselves corroborate this. Marlon Drummond, from the Royal Bank of Canada, was blunt: “Security is our top priority. We have no other job than to protect our customers' data.” In his case, detection relies on the network layer, using technologies like SD-WAN and DPI to build user behavior profiles and detect anomalies.

At the product level, HPE took advantage of the event to announce a unified SASE orchestrator, integrating EdgeConnect SD-WAN with its SSE stack in a single console. Additionally, it presented an AI-aware firewall concept, capable of distinguishing between approved, unauthorized, or tolerated applications, applying specific controls on workloads, prompts, or keywords. In Rahim's words, the goal is clear: to allow organizations to “see, govern, and protect the use of AI without slowing down the business.”

For network professionals, this scenario implies assuming a broader role: managing zero trust policies and AI governance within the network fabric itself, using the network as the primary security sensor by analyzing telemetry and lateral movements, and ensuring that automated decisions respect segmentation and security policies. In short, the role evolves towards a hybrid profile between network engineer and security architect, similar to what is discussed in our article on configuring secure VPNs and firewalls.

Experience-Centric Networks at Real Scale

Customer cases were probably the most illustrative part of the presentation, showing where complexity and scale push networks to the limit. This is the case of Ohio State University, whose network operates as if managing a small city: 66,000 students, thousands of faculty, tens of thousands of access points, and massive events that concentrate hundreds of thousands of people. In this context, the use of AIOps reduces incidents from hours to minutes.

In the healthcare sector, Tom Johnson from Sentara Health highlighted how network performance directly impacts patient care. “We move enormous amounts of data, and when it gets delayed, so does care.” AI technologies that generate clinical notes in real time are already in production. In his view, this demands reliable, secure, low-latency networks.

Disney, for its part, illustrated the challenge in the content industry. According to Ben Croy, global head of networks, a single production can generate massive data volumes, with hundreds of simultaneous projects worldwide. In this context, the network must be “fundamental, but ideally invisible,” so that creative teams can focus on content.

These examples reinforce a common idea: traditional metrics are no longer sufficient. What matters is measuring user experience and application performance, from videoconferencing to clinical systems or visual effects pipelines, and optimizing the network based on these indicators. Additionally, tools like digital twins and synthetic tests allow anticipating problems before they occur, validating complex deployments or large-scale events.

Conclusion: Networking Enters a New Era

Rahim closed his speech with a clear message: the traditional model of network operations has reached its limit. “The scale is too large, the complexity too high, the pace of change too fast, and AI is accelerating everything,” he stated. In this context, autonomous networks cease to be a future promise and become a real necessity.

For industry professionals, the challenge —and the opportunity— lies in leading this transformation: designing, governing, and evolving these new infrastructures before someone else does. Because, as the speech implicitly concluded: AI will not replace network engineers, but those who know how to leverage it will. In a landscape where giants like SpaceX, Nvidia, and Apple are already setting the pace, or where Qualcomm buys Modular to break monopolies, network modernization is an unavoidable step.

Original source: ComputerWorld. Analysis and adaptation by ForgeNEX.