Are Bill Gates’ energy expectations for AI optimism or realism?

As both a major funder of climate change initiatives as well as one of the largest economic beneficiaries of Open AI, Bill Gates’ claim that climate-focused concerns of AI energy usage are being overblown has been understandably met with some skepticism.

Is Gates just providing cover to protect his economic interests, or is AI actually capable of playing a critical role in guiding the transition from fossil fuels to sustainable energy?

The answer seems to be both; AI requires significant energy use, but it is also capable of providing transformative technological power that can shape our energy future.

It’s true: AI’s computing needs require massive amounts of electricity, which is partially why our world is in a constant grapple with the technology. Individuals like Gates are seemingly prescribing a treatment that has the potential to exacerbate the underlying problem. 

If AI threatens to dismantle hard-fought climate goals and our progress toward them, the question goes, shouldn’t we take a step back? 

While these are legitimate questions to ask during the rollout of a revolutionary technology like AI, concerns about its ultimate energy impact are overblown. The likely benefits of AI as a powerful tool in the fight against climate change can more than offset the increase in emissions its computing power requirements create. 

The power usage element of AI is real: The average AI query uses almost ten times the energy that a traditional Google search does, and just training the systems requires almost as much power as using them. The burden weighs heavily on America’s outdated power infrastructure. That means that utility companies across the country are starting to need to increase capacity to keep up with demand.  

But AI isn’t the only thing exacerbating energy demands; we’re all tethered to electronics that didn’t exist just two decades ago, driving electric vehicles that are being incentivized by state and federal laws, using heating elements that used to depend on natural gas and oil. When combined with the superfluous, not to mention similarly enormous computing and water power demanded by cryptocurrency mining — the demand on the electrical grid has never been higher. 

So doom and gloom over AI’s contributions to global emissions is a dramatic oversimplification of a complex web of increased energy needs. 

It’s not to say there aren’t many valid energy concerns when it comes to AI use, but rather that they’re being overhyped and under-examined. Grandiose exaggerations of AI gobbling up a quarter of all the electricity in the US, are more clickbait than genuine predictions, and fail to take one major variable into consideration: the energy market. As more and more energy is demanded, and there is less and less excess supply, costs will increase. 

AI is already a loss leader for most AI companies, like Open AI. While their war chests may feel infinite, they are decidedly not. If (and as) energy becomes more expensive, AI companies have increased incentive to drive power consumption down — already one of the reasons many AI focused companies are leaning into manufacturing their own chipsets.

In the interim, however, there are valid concerns that increased AI use will resurrect dead and dying fossil fuel plants to keep up with energy demand. This is a genuine risk — one that can’t be solved by depending on either the goodwill of consumers and businesses or painstakingly slow government regulation to rein in.

A more apt, and more immediate solution, would be to ensure that the carbon impact of this energy production is baked into the price per kilowatt hour. Such regulations would apply pressure on AI companies to create more energy efficient models, as well as to begin establishing price points for AI use which will begin to wind down frivolous and wasteful uses, ensuring that the tool is used on problems important enough to bear its increased cost.

Some of the problems that warrant the increased cost? Leveraging AI as an ally in the battle against climate change. 

The International Energy Agency calls AI and energy “the new power couple,” and foresees extensive usage of AI to optimize power production, monitor and manage distribution grids for efficiency and reliability, and unlock better consumer information that can help utility customers minimize usage. 

AI is also a powerful tool for researchers looking to drive game-changing innovations in energy production and conservation. AI can drive superior wind tunnel research for turbine testing and blade design. AI can help chemists analyze materials to identify binding agents with CO2, allowing the accelerated investigation of reactive agent permutations for carbon capture plants. It can also help design better and more energy-smart transportation and heating/cooling systems, two of the largest energy uses in our society. 

Yes, AI is increasing the demand on our energy grid, but it’s just part of a much larger story of electricity demand growth. Ultimately, we have to shift our energy use from fossil fuel to more sustainable sources, and AI is likely to be a critical tool in helping us make our grid as efficient as possible.

Bill Gates will be happy when Open AI eventually generates big returns. Let’s hope that continues to fuel the Gates Foundation’s climate giving, and introduces the world to some new tools we can use as we take on the work of solving our climate crisis.

Chris Yoko is the founder of Yoko Co and Carbon Off.