The Colossal Power Demands and Implicit Danger of AI

­Welcome to the May issue!

Lately, it seems like my “Final Thought” (the editorial bookending each issue) has become an unofficial AI column, with 3 of the last 5 features having to do with artificial intelligence.

And for good reason – AI is far more than simply the latest topic du jour. For better or worse, it’s infiltrated every corner of the industry, and we’ve definitely “passed the point of no return.

AI will become a ubiquitous part of our daily lives. The only questions are when and how much control humans can maintain. When a self-aware system that knows everything about us and can think faster and perform tasks quicker and more efficiently than our own biological limitations is with us 24/7, how do we prevent ourselves from becoming the redundant forms of life?

That said, our industry’s efforts to develop and refine AI at these relatively early stages gives me hope that science fiction will remain fictitious, and that’s where May’s issue – “Powering AI” – comes in.

The latter is becoming more and more relevant – while AI consumes less energy than, say, cryptocurrency (Bitcoin, alone, is projected to devour around 91 terawatt-hours of electricity annually), the rapid growth of the former will see it greatly exceed the latter in the near-future.

If we think training and operating generative AI programs is an energy hog, just wait until we can’t sneeze without an AI-optimized widget taking notice.

This is backed up by one of this month’s contributors, OmniOn Power’s Vito Savino, who points to Goldman Sachs research estimating that data center power demand will grow by 160% by 2030, while “the Electric Power Research Institute (EPRI) estimates that data centers could grow to consume up to 9% of the United States’ annual electricity generation.”

Guess which area is increasingly AI-optimized….

Indeed, “the rapid evolution of graphics processing units (GPUs) used for AI applications” and “Balancing thermal efficiency, cost, and performance in the face of rapid change” presents significant challenges for data center operators.

Of course, data center power demands and thermal management challenges predate AI by a significant margin, but AI has intensified all the normal concerns (and created several new ones).

As he explains, data center operators can address these power and thermal management hurdles by taking steps like adopting higher voltage architectures, strategizing data center locations and footprints, and optimizing cooling solutions.

And sticking with our theme, I’d like to briefly call-out a recent, AI-centric blog – wherein I explain how AI, amongst all its benefits and hazards, could give a significant boost to cybercrime.

Enjoy the May issue!

Best Regards,

Jason Lomberg

North American Editor, PSD