Arm: Why Strategic Buying is Crucial to Sustainable AI

As AI adoption accelerates, procurement professionals are uniquely poised to advance both operational excellence and sustainability by embracing edge AI and energy-efficient infrastructure.

With national priorities in the US shifting towards resilience, now is the moment for strategic procurement to enable innovation while controlling costs and carbon footprints.

Strategic partnership addresses AI sustainability crisis

These topics are central to a joint paper from Arm, a leading player in the semiconductor industry, and the Special Competitive Studies Project (SCSP), a non-profit, non-partisan initiative established to strengthen America's long-term competitiveness in AI, biotechnology and other emerging technologies.

The paper, Smarter at the Edge: How Edge Computing Can Advance U.S. AI Leadership and Energy Security, looks at the challenge which is plaguing companies across the globe: how to scale AI sustainably.

In a blog post, Vince Jesaitis, Head of Global Government Affairs at Arm, outlines the issue – as AI adoption quickens, a strain is being felt by the infrastructure which allows it to thrive, with growing power and resource demands.

Vince says: "Without new approaches, the energy cost of progress risks becoming unsustainable. Investment and research into edge AI, alongside data centres, can enable faster, more secure, more cost-effective AI, while easing the pressure on national grids and cloud infrastructure.

"The paper builds on Arm's continued work to promote efficient-by-design IP and compute architecture as the defining measure of progress for the AI era."

Edge computing: The answer to AI's energy crisis

AI tools, like large language models (LLMs) and generative AI, have fundamentally reshaped how humans engage with technology. But with every text enquiry or demand to generate an image or video, energy is needed.

Vince highlights that, while the training of massive models often captures headlines, inference (the process of generating outputs for users) now accounts for the majority of AI's energy use.

Vince believes that, although the efficiency obtained is very real, AI energy efficiency is not keeping up with its explosive growth. He adds: "If left unchecked, AI's power demands could outstrip available grid capacity, driving up costs and constraining innovation."

But there are ways to counter this. Edge computing – processing AI closer to where it's used – offers a path to AI energy efficiency, easing pressure on power grids while strengthening US competitiveness and energy security.

While edge AI is not a direct replacement, it is there to support it. Vince outlines how the frontier models can train in large data centres, but inference – where AI meets the real world – can increasingly happen on devices, in factories, hospitals and local networks.

"By combining specialised, energy-efficient hardware with optimised software architectures, edge computing can reduce energy consumption by up to 60% for equivalent workloads," he continues. 

"It also provides tangible operational advantages: lower latency, enhanced privacy and reduced network dependence. These advances make edge AI development and proliferation a strategic advantage."​​​​​​​

Each use case brings AI closer to the user while reducing the need to move vast amounts of data across energy-hungry networks.

Policy and investment: Building the US' AI future

Vince outlines how the US could play a central role in accelerating this transition. The White House AI Action Plan, the CHIPS and Science Act, the Department of Energy's EES2 initiative and AI Science Cloud all lay the groundwork for a more efficient AI ecosystem.

However, there must be a match of continued investment, smart procurement incentives and edge AI testbeds for public and critical infrastructure-sector missions like wildfire monitoring and grid management.

The future of AI leadership hinges on our ability to strategically distribute computing power throughout the complete infrastructure – from massive data centres down to the countless edge devices woven into everyday life.

Nations worldwide have unveiled bold initiatives, positioning themselves for the coming wave of AI competition. For the US to maintain its competitive edge, it must adopt a comparable approach that connects advances in hardware and software with intentional policy frameworks supporting energy-efficient AI computing.