Can quantum computing tame AI's growing energy appetite?

[ModelCoreWhale]

Interesting article about whether quantum computing could help with the enormous energy demands of AI. The authors argument is that quantum machines might eventually take on some workloads currently handled by power hungry GPUs, while using far less electricity than today's data centres. Given everything we have discussed about data centre power use this month, from Bain's $6 trillion forecast to underwater data centres, it is a timely question

The article cites a Boston Consulting Group estimate that mature quantum computing could help address 20 to 40 percent of emissions from hard to abate sectors like steelmaking and cement, mainly by helping design better materials and processes. IonQ's chief executive is quoted saying quantum's carbon footprint will be relatively small compared with AI's, because quantum is more likely to be used for specific problems than as a general purpose technology. One comparison suggests a quantum computer might need only two standard wall sockets. Those are big claims that will take years to test

It also touches on other energy related research, such as quantum batteries, harvesting waste heat from electronics and an MIT discovery about electrons moving without friction, which could lead to very efficient circuits. Some of these are early stage lab results, so it is a broad look at possibilities rather than a set of ready solutions. Some of those ideas sound more like science fiction for now

I think there is an important caveat here. Quantum computers are not general replacements for GPUs, and nobody expects them to run large language models any time soon. Some quantum machines also need large cooling systems, so the wall socket comparison does not apply to every design. The article focuses on future potential rather than what is possible today. Readers should keep that in mind

Still, the idea that quantum might help solve energy problems rather than add to them is an appealing one. Do you think quantum computing will ever ease AI's energy demands? Or is it more likely to help indirectly, by designing better batteries, materials and power systems?

Can Quantum Computing Tame AIoilprice.com