Federal agencies urged to invest now in accelerated hardware to meet looming AI workloads

Federal agencies face a critical juncture as the rapid adoption of artificial intelligence (AI) is expected to strain outdated IT infrastructure, according to a new report produced by Scoop News Group and underwritten by Hewlett Packard Enterprise (HPE) and Intel. The report underscores the need for immediate, strategic planning for  AI-ready hardware and computing capabilities to avoid falling behind in the delivery of government services.

The report highlights the growing expectation from citizens for seamless, frictionless government services, mirroring those offered by the private sector. The challenge for agencies is that current systems often lack the processing power, speed, and agility required to handle the evolving demands of AI workloads.

“We believe the principal workload for the future enterprise by 2029 or 2030 is going to be an AI workload — and that will require accelerated infrastructure capable of supporting AI applications,” Bill Burnham, former Chief Technology Officer for U.S. Special Operations Command and now U.S. Public Sector CTO at HPE said in the report. “The enterprise systems that agencies have built over the last 20 years are ill-equipped to handle AI workloads.  Agencies need to start looking now at their enterprise IT infrastructure and begin modifying and modernizing it for the AI age.”

The report emphasizes that a lack of proper planning today could leave the government on the back foot when AI becomes more prevalent in government workflows.

“You want to become both digitally modern and prepared for the AI era by making infrastructure procurements now that are AI-ready, AI-aware so that when you get further in your Al journey, you have an infrastructure ready to receive it,” added Cameron Chehreh, Intel Corp. Vice President and General Manager of Worldwide Public Sector, underscoring the importance of proactive planning.

Adding to the need for renewed infrastructure planning are factors like the increasing shift to edge computing and remote AI-driven data analytics. The associated costs and time involved in sending massive datasets over the internet to the cloud data and issues around data sovereignty, confidentiality, and security compliance are expected to increase demands for localized accelerated computing power.

Five key takeaways from the report

The report highlights several emerging themes agencies need to consider:

1. Hardware matters: Existing legacy systems are generally ill-equipped to handle the intensive workloads of AI. As agencies move forward with implementing AI applications, they must also assess their near- and longer-term hardware needs or risk falling behind.
2. Edge computing’s growing role: The rise of edge computing requires robust, efficient localized micro-processing capabilities to handle real-time AI tasks. Moving data to a central cloud for processing is becoming increasingly costly and time-consuming.
3. The five-year refresh trap: Traditional five-year hardware refresh cycles, where agencies typically upgrade a fifth of their hardware yearly, will likely leave agencies flat-footed as AI workloads dominate enterprise activities unless they begin procuring AI-ready technology.
4. Strategic chip procurement: Agencies must also consider the costs and benefits of deploying AI-ready CPUs and GPUs. Purpose-built acceleration CPUs for inferencing tasks may be more cost-effective and energy-efficient than GPUs. “People are coming to understand there is a spectrum of Al applications – from ‘augmented Al,’ which automates routine data tasks, to ‘autonomous Al,’ which facilitates full-scale automation. Not everybody needs GPUs. What people need is acceleration,” argued Intel’s Cameron Chehreh.
5. Factoring in data security and sovereignty: The origin and supply chain of processors remains critical to the security and trustworthiness of AI algorithms, adds Chehreh. Agencies must understand the security safeguards in today’s latest chips and the supply chain building them.

The report also highlights Hewlett Packard Enterprise’s involvement in building the world’s fastest AI-capable computers.  HPE, for instance, recently delivered an exascale supercomputer to the Department of Energy’s Argonne National Laboratory, capable of performing a quintillion (one billion billion) calculations per second. HPE systems now power 37% of the world’s top 500 supercomputers and seven of the top ten most energy-efficient supercomputers, according to company officials.

The report concludes that the future of government services hinges on agencies’ ability to innovate and adapt to the changing needs of citizens. By investing strategically in the right technology, government agencies can leverage the power of AI to enhance public services, improve citizen satisfaction, and remain competitive with private sector counterparts.

Download the full report here.

This article was produced by Scoop News Group for FedScoop and sponsored by Hewlett Packard Enterprise and Intel.