close
close

Analyst: Proposal for AI testbed for Department of Energy is a good first step


Analyst: Proposal for AI testbed for Department of Energy is a good first step

To make an immediate impact, the report’s authors say, “the Secretary should bring together energy suppliers, data center developers and operators, and other key stakeholders to begin an active dialogue on how to overcome current power supply constraints” and “develop strategies to generate and deliver the power needed to maintain AI leadership well into the future.”

The recommendation to create a test bed, Beran said, “is the first step for the Department of Energy to understand what infrastructure is being used and how much energy it is using. And once we have that baseline data, how can we improve? It really reminded me of the old saying, ‘You can’t improve what you can’t measure.’ You have to start somewhere and establish a baseline, and that’s it for me.”

Develop solutions

He said the hyperscalers the working group asked for their input were facing “unresolved issues in addressing the energy requirements of AI workloads. It’s not that the industry has solved the problems or challenges, it’s more like, ‘We’ve identified challenges in AI workload and energy profile requirements and now we need to start developing some solutions for them.'”

These solutions, Beran said, ranged from changing the architecture of the data center facilities to system changes to meet the workload profile.

Particularly important, he said, is the need to improve energy efficiency. He added that while sustainability “has been such a critical factor in recent years, it has now taken a back seat to some of the requirements of AI growth. It’s important to try to balance the two.”

In addition, Thomas Randall, head of AI market research at Info-Tech Research Group, said via email: “As AI models grow larger and require more computing power, the amount of energy needed to support this market will also increase. Without a more comprehensive energy strategy, countries where the data centers used by AI companies are located will face continued carbon emissions.2 Emissions problems, growth constraints and opportunity costs of using energy elsewhere.”

Leave a Reply

Your email address will not be published. Required fields are marked *