At a time when generative AI is no longer a promise but an operational reality in enterprises, Hewlett Packard Enterprise CEO Antonio Neri delivered a powerful message from the Discover event in Las Vegas: we are facing one of the greatest technology platform shifts in history. And HPE is not just observing it but capitalizing on it with a comprehensive transformation of its portfolio, from networking to storage, computing, and operations governed by AI agents.

The Network as the Backbone of AI

For HPE, the network is the foundation on which any AI infrastructure is built. "Every byte, every token, every decision, everything goes through the network," Neri emphasized. The company has structured its network offering into four layers: scaling within a rack, scaling between GPU clusters, data center interconnection, and inference routing at the edge. The new QFX switches, the PTX 12,000 routing platform with 800G capacity, the SRX 4700 firewall with quantum security at 1.44 Tbps, and the MX 301 edge router with Juniper's sixth-generation Trio silicon are the key pieces of this strategy.

Latency is the silent enemy in AI environments. Neri illustrated this with a revealing example: "Multiplying a small delay across hundreds of thousands of GPUs over weeks of training can mean the difference between training a model in 90 days or 30." In a sector where innovation speed determines leadership or following, these network optimizations are critical.

Computing: ProLiant DL 394 Gen 12 and Agentic Scalability

Computing is not far behind. HPE has introduced the new ProLiant DL 394 Gen 12, specifically designed for AI agent workloads and long contexts. This server is the heart of the new Private Cloud AI configurations, which now scale up to 256 GPUs with multi-node inference. According to Neri, this allows training models with a quarter of the GPUs compared to the previous platform and performing inference at a tenth of the cost per million tokens.

The company structures its portfolio into three levels of AI Factory, targeting enterprises, service providers, and sovereign deployments. A unified gateway provides a single API to access advanced and open-source models, while a shared cache reduces the cost of the first token. "Private Cloud AI can now serve larger models across multiple systems with multi-node inference, so capacity grows with compute," Neri stated.

Unified Storage for AI Data

AI agents are only as intelligent as the data that supports them. HPE has responded with the Alletra MPX 10,000, which becomes the storage layer for Private Cloud AI, unifying files and objects in a single architecture. This system adds real-time metadata enrichment and native MCP support, allowing agents to access structured and unstructured data without custom preparation. HPE claims it accelerates time-to-value by 7 to 12 times compared to custom environments.

"Your AI agents are only as smart as the data they are trained on," Neri said. "Traditionally, that data required custom preparation for each use case and months of pipeline building, but that is no longer the case."

Agentic Operations: Governance for the New Era

The proliferation of AI agents in enterprises, often managed by developers outside IT control, poses governance and scalability challenges. HPE has integrated a governed agent layer into Private Cloud AI that allows registering agents from any framework, applying security controls over API calls, identity, and encryption without code changes. A three-tier identity model verifies the user, governs the agent, and requires human approval for sensitive actions.

Additionally, integration with Nvidia Open Shell provides isolated environments with policies, NeMo Cloud for designing governed workflows, and Zerto for reverting to clean states when agents make mistakes. This approach is key for enterprises to adopt agentic AI without compromising security or control.

CloudOps and Hybrid Unification

HPE CloudOps unifies virtualization, data protection, and cloud management into a single hybrid operational layer. The Unleash AI program has expanded to over 60 validated partners, facilitating the integration of AI solutions in enterprise environments. This platform allows companies to manage their AI workloads consistently, whether on-premise, at the edge, or in the public cloud.

For those interested in automating business processes with n8n and AI, HPE's infrastructure provides a solid foundation for implementing agentic workflows with security and scalability guarantees.

The Energy Challenge

Neri did not avoid the elephant in the room: energy. "Every model, every workload, every agent depends on energy, because essentially an AI factory converts electrons into tokens," he explained. He warned that the U.S. faces a 19-gigawatt deficit by 2028, with data centers accounting for nearly half of electricity consumption through 2031. "As AI scales, the future will not be determined solely by computing. It will be determined by how efficiently we can power, cool, and connect it."

This challenge opens opportunities for energy efficiency and advanced cooling solutions, areas where HPE is investing significantly. The company is also exploring advanced home automation for offices as part of its sustainability strategy.

Implications for Enterprises

HPE's message is clear: traditional infrastructure is not ready for the era of AI agents. Companies that want to harness the potential of agentic AI need low-latency networks, scalable computing, unified storage, and robust governance layers. The integration of Juniper Networks into HPE's portfolio reinforces this vision, offering a complete platform from edge to data center.

In Neri's words, "AI today is about moving faster from ambition to outcome, accelerating time to token, reducing execution risk, and ensuring environments ready from day one." For IT professionals, this means AI adoption is no longer just about algorithms but about infrastructure.

As we have seen in how AI reviews code better than humans, intelligent automation is transforming all aspects of development and operations. HPE's proposal provides the foundation for these capabilities to be deployed securely and efficiently at enterprise scale.

Original source: ComputerWorld. Analysis and adaptation by ForgeNEX.