🔋✨ Energy and AI … A Quick Q&A with science writer Michael Puttré

'Fortunately, the US seems a lot more pragmatic on the energy front, even with the constant drumbeats of climate change and green energy.'

Since OpenAI made ChatGPT available to the public in November 2022. Since then, AI’s energy implications have become increasingly apparent to developers, lawmakers, and environmentalists alike. AI isn’t slowing down, and, as the economy incorporates AI into nearly every industry, neither will our energy demand. As science writer Michael Puttré wrote for Discourse magazine recently:

The advent of AI threatens to upend all the assumptions of green power and energy transition advocates who now call the tune for the climate change bureaucracy. AI-assisted queries on the internet are already having a tremendous impact on electricity consumption. The current thinking now is that an AI-assisted query requires 10 times the amount of electricity as a pre-AI search. This is just the nature of AI, which throws computation processing at a problem, like chess-playing computers able to defeat human Grand Masters by generating a million possible moves. Processing requires electricity—and lots of it. As with essentially all promising technologies in recent history, experts and policymakers are proceeding with their preferred outcomes with either a blind eye toward or willful ignorance of the practical effects of their mandates.

Following up that piece, “AI Confounds Assumptions of a Clean Energy Future,” I asked Puttré a few quick questions about the US’s energy infrastructure outlook. We touched on how facilitating nuclear power could be the key to both environmental sustainability and a techno-optimist future.

Puttré is a writer, editor, and journalist. He is currently a contributing writer and editor for Discourse, as well as a correspondent for PV Magazine. He was formerly chief editor of the Journal of Electronic Defense and Solar Industry, and is the author of science fiction novel Outre Mer.

---

1/ Should we be holding off on AI development until we can find a way to meet the impending energy demand? How big is the shortfall looking to be?

I don’t know that there is a way for AI developers to be restrained even if we wanted that to happen. Unlike electric vehicle mandates, it’s not government incentives or preferences prompting demand. People out there want to make money with AI, whether it’s hedge funds seeking an edge in weather forecasting, media companies wanting faster, cheaper content creation, or tech giants trying to hold users with better search results. 

I’m sort of agnostic on the benefits of AI because I don’t understand all the things you can do with it now, let alone going forward. Think of the first airplanes. All I know is that people are investing huge amounts of money in it, and data centers have become the largest growth area for electricity demand. The Boston Consulting Group says data center demand is going to increase between 15–20 percent annually through 2030, leading to a possible shortfall of 80 GW of 24/7 generating capacity by then. Demand is growing a lot faster than government agencies predicted.

In a way, we’ve sort of dodged a bullet in the short term because EVs are just not that popular. Back in 2020, Pew was saying that electricity consumption was going to increase 38 percent by 2050 because of EVs, and AI wasn’t really on anybody’s mind then. Government can make mandates all it likes, but I bet we’ll see some flexibility on those. But AI seems to be coming on strong, ready or not.

2/ What happens if we allow AI to develop without reconfiguring our current energy infrastructure?

We’ve been living in an interesting couple of decades in that electricity demand since around 2000 has been relatively flat in this country. Overall economic growth has continued at some level since then. But the US Energy Agency says demand didn’t change much at all. Residential demand went up a little bit, but commercial and industrial demand actually decreased. Energy efficiency of lighting and HVAC and appliances are playing a role. But a lot of this is because we just don’t manufacture as much domestically as we used to — all of that offshoring. 

This had even led some environmentalists and energy-transition advocates to say economic growth was decoupling from energy demand. The thinking was you can have economic growth because of the emerging information economy without manufacturing and production using all that electricity. The advent of AI and cloud computing, which are becoming inextricably interlinked, has sort of turned that assumption on its head. 

The relatively flat demand for the last 20 years has encouraged a lot of the optimism and experimentation with renewable energy sources, particularly photovoltaic solar and wind generation. The demand flatness gave policymakers the breathing room to incentivize the deployment of clean energy at all levels, from homeowners’ rooftops up to utility-scale wind and solar farms. In many ways, this has been revolutionary and has enabled many utilities to take their dirtiest coal and oil plants offline.

Increasing electricity demand means we can’t just swap out certain classes of traditional generation sources for wind and solar at our leisure. We need renewables and traditional generation: more of all of the above. They won’t say it out loud, but New York is probably sorry it took the Indian Point nuclear power plant offline. California has already changed its mind about taking Diablo Canyon offline. Nuclear is making a comeback. Natural gas has a new lease on life. I’m sure this makes energy transition advocates grumpy for now.

3/ Do you anticipate us increasing dependence on traditional energy sources as electricity demand grows before we can make the shift to clean sources? How is government reacting?

If by clean sources you include new technology nuclear and natural gas along with renewables, then the shift to such generation can continue apace and even accelerate. Solar in particular is relatively easy to install even at utility scales compared to natural gas and especially nuclear projects. The technology is well understood and the components are solid state. A solar project can go up in a fraction of the time, as long as land can be found and grid interconnections can be scheduled. These are the only real roadblocks. Municipalities are finding productive solar locations on big commercial flat roofs, capped landfills and even fallow industrial sites.

Wind, as we are finding out, is a little dicer. The travails of Vineyard Wind off Nantucket are showing offshore wind can have huge technical challenges. But a lot of the country gets a significant proportion of its electricity from wind, particularly in the Midwest, from Texas up through Iowa and the Dakotas. Everything in its place. 

Which means, yes, traditional generation sources are going to be with us longer than the most optimistic and enthusiastic energy-transition advocates would like. In Germany, the actual political authority of the Greens enabled them to take aim and shoot their country in the foot by shuttering nuclear plants entirely, even accepting Russian natural gas pipelines in the bargain, all to fulfill a hard ideological goal. 

Fortunately, the US seems a lot more pragmatic on the energy front, even with the constant drumbeats of climate change and green energy. Perhaps the most promising thing I’ve seen in ages is President Biden signing the so-called ADVANCE Act in July easing restrictions on building new nuclear plants and reforming rules on existing ones. This new law passed both houses of Congress with overwhelming majorities and reinforces my faith that government can act pragmatically and with foresight, at least some of the time. 

4/ Can AI developers and nuclear plant engineers form some kind of alliance, practically speaking, to achieve their mutual energy goals?

An AI systems developer I talked to says he could see new data centers with AI-processing chips needing 150 MW, and that these would benefit from having their own nuclear reactors on site; one of the new smaller, modular ones they are planning. This would be neat. However, those modular new-style reactors don’t exist yet, and data centers are going up now. They need the electricity now. 

One renewable energy company I know of is working with data center builders to design and develop integrated campus facilities with large solar farms on site. This would take a lot of land and so would not necessarily work in all regions. However, it is an example of the private sector looking at a potential problem as an opportunity to develop new business arrangements. I’m sure that the reality of AI demand will spark a number of such partnerships going forward. Maybe involving wind, maybe involving new-generation gas plants. 

In any event, data center builders will shop around to find places where utilities have the capacity to meet their electricity demands. This means a lot of utilities are going to be reluctant to retire their existing traditional generation plants and may even be motivated to build new gas-fueled plants. 

5/ Do you see AI becoming politicized by environmentalists as the huge energy demands become more apparent?

I get the sense that environmentalists are still coming to terms with the implications of AI’s electricity demands, as are we all. Generally, Americans are only dimly aware of AI’s increasing power requirements and how these might eventually affect their own utility bills and even access to electricity. Energy transition advocates might be more plugged in, so to speak, on the issue. But Big Tech has been a strong ally of environmentalists ever since it overshadowed HP and IBM. The Google homepage has that little footnote saying it’s been carbon neutral since 2007. It’s going to be hard for environmentalists to call them out. 

Big Tech has been able to get by pointing to carbon offsets like planting trees and buying renewable energy credits from wind and solar generation elsewhere to compensate for their energy usage. These arrangements are very popular and widely accepted by environmentalists, but their actual impact on emissions, let alone the climate, is dubious. What is indisputable is that more data centers mean more demand from all sources and some percentage of this will increase emissions. Unless AI is actually a bubble, it is on course to be responsible for an increasing share of emissions going forward. You won’t be able to plant enough trees. 

I don’t know how true-blue environmentalists are going to cope with this. I’ve seen some predictions that AI will help us design more efficient technologies and cleaner forms of energy. Possibly. But it will be interesting when this AI electricity demand really starts to hit the grid. 

6/ What is the optimistic end game here?

Well, assuming Colossus doesn’t emerge and start bossing us around, I view AI as an example of a new technology that could enable visionaries and entrepreneurs to expand our capabilities and create more prosperity. Or, as I said, it could be another tech bubble. I doubt the latter, but I’m not an expert nor a venture capitalist.

I am a technology optimist, though, and in the short term I will be very interested to watch how AI developers work with other industries, like power generation, to increase the energy supply. As the supply increases to accommodate data centers, developers will be able to increase the supply for us all. And as someone who supports an all-of-the above energy strategy, I’m looking forward to more solar and wind — where appropriate — and new-technology nuclear reactors and cleaner-burning natural gas plants. Ultimately, I think we can have our prosperity-generating energy and lower emissions, too.

And who knows? Maybe AI will help mankind get to Mars in my lifetime.