We’re on day three of HPE Discover, and the tone is shifting. The first few days focused mainly on the fundamentals: networking, data, and computing. Today, Fidelma Russo delivered a keynote that conveyed a clear message to me. The question is no longer whether AI will transform your organization, but how to manage that transformation and make it operational.
For those who don’t know Russo: she is the Chief Technology Officer at HPE and heads the Hybrid Cloud division there. With more than thirty years of experience in the industry, having previously worked at companies such as EMC and Sun Microsystems, she approaches AI with the perspective of a builder. That makes her story very practical. She looks beyond the hype and focuses on how to actually make it all work.
In my previous two blog posts, I wrote about the fundamentals and about the network as a platform. This third part brings it all together where it ultimately matters: in your operations.
Workflows are giving way to systems that act on their own
We’re clearly entering the next phase. AI is no longer just a standalone assistant that helps out here and there. It’s evolving into what HPE calls a“distributed agentic enterprise.”Behind that term lies a simple idea: intelligence is everywhere—in your applications, your workflows, your teams, and your infrastructure—and it’s starting to make decisions and take action on its own.
The shift that stuck with me the most concerns how work is done. In the past, a person would make a decision and the system would carry it out. Now, systems are emerging that observe, reason, and act on their own. This happens through AI agents that collaborate with one another. And that’s where the challenge lies, because that intelligence isn’t centralized. It’s spread across silos, data, and platforms.
The challenge today lies in something other than building AI. It revolves around orchestrating, managing, and ensuring its reliable operation.
Closed-loop operations: AI in a controlled loop
HPE’s response to this is called closed-loop operations. Systems monitor, analyze, take action, and verify that the result is correct—all continuously and in real time.
That only works if three key elements are in place: reliable data, efficient infrastructure, and intelligent operations. Each of these was mentioned in the keynote.
Data is becoming an ongoing building block
What I found impressive: data is no longer viewed as simply the input for a model. It’s a component that remains integral to the entire process. With HPE Data Fabric 8.2, HPE aims to make that a reality. Data is available everywhere, whether you’re working at the edge, in the cloud, or in your data center. Governance and security are built in by default, and identity and access are managed automatically.
The underlying principle is simple. Without reliable and accessible data, you’ll never build AI that truly scales.
Token economics: AI costs money—and a lot of it
One of the most concrete sections dealt with token economics. Each AI agent consumes tokens, continuously reasons, and performs actions. As a result, inference is no longer a one-time workload, but a constant cost that keeps ticking away.
One figure really stood out: examples of $13,000 per agent per month. That immediately puts things into perspective. AI economics thus become infrastructure economics. Success depends on efficiency, scale, and consumption—and by no means solely on the model itself.
HPE illustrated this with an internal example, a platform codenamed Mindstone. They built it on-prem, with private cloud AI. The result is an environment that runs about thirty times cheaper and saves nearly $100,000 per month. The lesson is clear. Those who control their data and infrastructure gain an advantage in both cost and governance.
Storage becomes memory
Another thing that caught my attention: storage is evolving into active memory. With the HPE Alletra Storage X10000, featuring KV Cache acceleration and NVIDIA certification, the time to first token is up to twenty times faster, and throughput is seventeen times higher.
Why does that matter? Without that memory, agents have to rebuild their context over and over again. That’s a pure waste of computing power, resulting in rising costs. It’s typically the kind of detail that has a huge impact at scale.
Infrastructure grows along with it; it doesn’t shrink
Some people think that AI reduces the size of your infrastructure. In practice, the opposite is true. More agents mean more API calls, more database queries, and more load on your GPUs and CPUs. AI creates a multiplier effect across your entire IT landscape.
HPE is therefore positioning its private cloud portfolio—which includes the PC1000, 3000, and 7000—along with Private Cloud AI as the foundation for keeping that manageable.
Intelligence in Your Operations
This is where it really resonated with me personally. Through its software stack, HPE brings AI directly into operations. Morpheus handles orchestration and automation, OpsRamp handles observability, and Zerto handles resilience.
The real heart of the system lies in GreenLake Intelligence. That’s where central governance for agents comes together, along with identity and policy management and orchestration via a “planning agent.” In addition, there are copilots for compute, orchestration, and observability. In practice, this means you can manage, automate, and troubleshoot your infrastructure using natural language and AI support.
I found the link to ServiceNow interesting. That integration paves the way for an autonomous AI workforce. So you don’t just gain insights—you also get automated service delivery.
A Reality Check from Customers
What I appreciated were the customer testimonials. AMD views AI infrastructure as an opportunity to evolve from a token consumer to a token generator. Point32 Health emphasizes governance, reusable AI platforms, and AI skills within the organization. Both confirm the same thing. AI is just as much about organization and processes as it is about technology.
What I’m taking with me
If I were to summarize Day 3 from my own perspective, it boils down to five things:
- AI has become an operational model; it is no longer just a project.
- Token economics is becoming a key factor in every business case.
- Data governance is becoming a prerequisite.
- Your infrastructure is becoming more critical and, at the same time, more complex.
- And above all: You need to control AI, not just build it.
The real shift lies in how you integrate AI. It’s less about adding AI to what you do, and much more about weaving AI into the way your organization operates. That’s a bigger undertaking than a new use case, and it determines whether all those use cases will ever deliver value.
I’m curious to know how you view this. For you, is the challenge mainly in building AI applications, or do you also find that managing them is becoming the real challenge?
Be sure to also read Part 1 about the foundation and Part 2 about the network.
About the author
Frank Dierckx is a Partner Alliance Manager at Xylos and tracks developments in infrastructure, partner ecosystems, and emerging technologies. His expertise helps clients make technology choices that are technically sound and economically sound.
