HP Z and AMD lead the Federal AI infrastructure revolution. Scalable and resilient AI infrastructure starts here.

HP Z and AMD lead the Federal AI infrastructure revolution | 2

AI infrastructure is mission-critical. Whether you’re in research, operations or business development, it’s the foundational element that drives business mission needs and when paired with emerging technologies, it can help Federal agencies exceed public expectations.

During a recent webcast titled “Building Smarter Solutions: Exploring In-House AI Development and Infrastructure,” leaders from the National Oceanic and Atmospheric Administration (NOAA), HP and AMD spoke about the importance of AI infrastructure and what agencies can do today to get it right.

Future-proof AI infrastructure with HP Z and AMD.

With edge data growing in importance, organizations are shifting their efforts from cloud computing toward distributed racked solutions. At CES 2025, this idea was front and center, with the HP Z and AMD team showcasing how they’re liberating AI workloads by making it easier to access secure insights at the mission’s edge.

“The HP Z AI workstations that are introduced this year are going to have an [50+ TOPS] AI engine in them, so more of these workflows are going to come to the edge, whether they’re using the integrated AI engine or high-performance graphics on endpoint devices,” said Andy Parma, Director of MNC Workstation Processors at AMD.

Platforms like these also empower users who may not be familiar with the complex coding languages associated with large language models (LLMs). James Olguin, AI & Data Science Business Development Manager at HP, forecasts that by 2026, users will be able to interact with digital agents

“The HP Z AI workstations that are introduced this year are going to have an extremely high- performance AI engine in them, so more of these workflows are going to come to the edge.”

— Andy Parma, Director, MNC Workstation Processors, AMD

https://events.govexec.com/building-smarter-solutions-exploring-in-house-ai-development-and-infrastructure/ https://events.govexec.com/building-smarter-solutions-exploring-in-house-ai-development-and-infrastructure/

HP Z and AMD lead the Federal AI infrastructure revolution | 3

to deliver a semi-complete story. While agentic capabilities are forthcoming, capabilities like this promise to provide productivity gains, informing users of potential issues or helping direct resources to serve local communities.

But as teams increasingly talk about AI infrastructure, managing costs will be crucial to shifting from a Data Center Operational Expenditure (OpEx) model to a Capital Expenditure (CapEx) model with procurement purchasing, leasing or financing the hardware.

“As our customers buy infrastructure that they’re going to keep for three to four years, it’s really important to buy the most capable AI products now because 12 to 24 months from now, the capabilities of AI software and infrastructure are going to be dramatically different,” Parma said. “You want to make sure that the hardware you buy today is going to scale with that evolving software and workflows, so you don’t have to do an off-cycle refresh.”

Start small: Rapid AI deployment relies on trust and focused use cases.

Despite the need for rapid procurement and deployment, Federal agencies face a complex landscape, with developments in AI on a near-daily basis making it difficult to do the right kind of planning. Considering this, Frank Indiviglio, Chief Technology Officer at NOAA, emphasized the importance of prototypes, not production. Start with a narrowly focused, well-defined project. Work alongside a group of key stakeholders and make sure to put technology into the hands of employees so that they can start to engage with and test the technology.

Define “AI success rates” to gauge operational effectiveness.

Work with your stakeholders to define the program’s key performance indicators

Before new models support Federal workloads, technologists often ask hundreds of questions about their perceived operational effectiveness.

• What’s the latency?

• How quickly does the model develop an answer?

• How relevant is the answer to the specific prompt or task at hand?

• How large of a workload can the hardware handle?

• What’s the performance per watt?

“Give them room to run and be flexible. We’re in an age of exploration.”

— James Olguin, AI & Data Science Business Development Manager, HP

While each of these questions plays an important role in AI, it’s important to remember that the end goal of AI is to drive tangible benefits for the end user.

“We’re looking for a return on investment. Is this actually helping people be more effective? Are they using their time better?” asked Olguin.

As organizations tie AI into existing systems, simplified KPIs could help technologists measure the impact of a specific model through “AI success rates”: measuring if the model lives up to its intended purposes and drives positive outcomes for local constituents.

Communicating the return on investment that AI systems could create will be crucial moving forward, and by simplifying KPIs, Federal leaders can improve that understanding with key stakeholders.

HP Z and AMD lead the Federal AI infrastructure revolution | 4

Build a layered approach to AI infrastructure with HP Z and AMD.

Generative and traditional models shouldn’t operate in a siloed manner. As Public Sector organizations explore different use cases, building data pipelines across workloads can help increase understanding.

“An organization where data flows, is an organization that’s well-informed.”

— Frank Indiviglio, Chief Technology Officer, NOAA

“An organization where data flows, is an organization that’s well-informed,” said Indiviglio. “But there’s also the governance part. You’ve got to think about the four pillars of governance. What’s the purpose? Why are we doing this with AI?”

As Public Sector organizations build up this AI infrastructure, asking questions about each layer and how it’s governed lays the groundwork for future efforts. If a program isn’t delivering on predicted results, this information could be used to inform stakeholders and customers about anticipated project changes.

“It’s a way for your organization to interact with their customers and stakeholders in a much more meaningful way,” said Indiviglio.

Ready to drive impact with HP Z powered by AMD?

Contact a government expert to learn more.

https://reinvent.hp.com/AIforGovernment