By Dan Gearino for Inside Climate News
On Sept. 21, President Donald Trump responded to rising public concerns about artificial intelligence with a post on Truth Social.
“The same people that said, ‘We’ll all be dead in twelve years because of Global Warming,’ a name since ‘reborn’ to Climate Change, because the Planet was cooling, not warming, and nobody was dead, are the people that are now saying that AI is going to kill us, that robots are going to attack us, and everything is a DISASTER,” he wrote on Truth Social.
He then declared fear of AI “a hoax.”
I found it notable that he injected climate denial into the discussion of AI, as if that would help his argument, and that he did so on the first day of Climate Week NYC.
The annual event has included corporations touting how AI could help the climate. For example, the chip maker Nvidia highlighted how its AI products are being used to advance clean energy, such as operational improvements at a nuclear power plant and energy-storage projects that utilize used batteries. Josh Parker, Nvidia’s head of sustainability, expanded on this in an interview Wednesday with The New York Times at Climate Week. He said AI could help reduce emissions reductions for transportation, manufacturing and buildings, with benefits that would outweigh AI’s climate footprint.
Looking at Trump’s comments alongside some of the hopeful messages coming from Climate Week, I see an evolving AI discourse that mixes climate denial and techno-optimism, neither of which is helpful.
Trump was touching on—and dismissing—concerns that AI would lead to superintelligent computers that would kill people. Yes, that’s a reason to worry, but the more immediate concern is AI’s vast need for water, electricity and investor dollars.
So what’s the reality? The most useful encapsulation of the peril and opportunity of this moment, in my view, was the paper published in a Nature journal in August that finds the increase in carbon emissions from powering AI data centers and using AI to help optimize fossil fuel production is much more than the savings from using AI for end uses related to clean energy. (My colleague Arcelia Martin wrote about the paper.)
The co-authors include Will Alpine and Holly Alpine, a couple who worked on AI at Microsoft before starting a nonprofit, the Enabled Emissions Campaign, which calls on technology companies to be accountable for how AI will boost fossil fuel use.
Their paper’s findings align with statements from organizations such as the International Energy Agency, which has said the potential climate benefits from AI-enabled systems are still on the horizon, while some concerning ramifications will come sooner, such as increased electricity demand for data centers.
Related | Americans’ hatred of data centers is about so much more
I asked Holly Alpine what she makes of the talk of AI at Climate Week.
“What we’re seeing at Climate Week is a familiar pattern: companies spotlighting a handful of clean energy pilot projects while the overwhelming majority of AI’s actual applications are aimed at increasing fossil fuel extraction and production,” she said in an email. “Nvidia’s post is a good example—highlighting a renewable energy partnership is easy and photogenic, while conveniently omitting how much AI is being deployed to find, extract, and produce more oil and gas.”
This isn’t complicated. If AI were powered mostly by clean sources and its emissions-cutting uses outweighed its climate-harming uses, then there would be no need to mislead about the effects.
Jonathan Koomey, an independent researcher who has been studying data centers for decades, is wary of companies describing AI as beneficial for the climate while not acknowledging the potential downsides.
I spoke with him recently following a speech he gave in my city, Columbus, Ohio, about the challenges of accurately forecasting the scale of the data center boom— a particularly relevant topic in Columbus, one of the country’s leading data center markets.
“Projecting short-term trends for even 10 or 12 years is a recipe for disaster, in my experience,” he told me.
Koomey’s larger point is that the scale of the data center boom remains uncertain. One of the most harmful scenarios is that utility companies build new power plants to serve anticipated data centers, only for many, if not most, of those projects to never materialize. If that happens, households and businesses would be left to cover the costs through their utility bills. Regulators and lawmakers need to take steps to protect consumers, he said.
For now, one of the most troubling aspects of the AI data center boom is how the rush to build as fast as possible has led developers to rely on some of the least efficient and most expensive power technologies, such as natural gas combustion turbines.
“It’s madness,” Koomey said, noting that many of the turbines are essentially jet engines, built for short-term use. “No sane [utility] would buy electricity from combustion turbines running 24-7. It just wouldn’t happen.”
At some point, economic gravity will assert itself, he said. Data centers will need to recoup their development costs and determine how much customers are willing to pay.
This reality doesn’t mesh with Trump’s all-caps defense of AI, and feels far removed from talk of how AI will help the climate. It portends an industry downturn that may be coming soon.
One more thing: I didn’t want to get waylaid by fact-checking Trump’s statement, but one part demands some context.
When he refers to people warning “we’ll all be dead in twelve years,” he’s likely referring to the media coverage of a 2018 report by the United Nations’ Intergovernmental Panel on Climate Change that said the world had 12 years to reduce emissions enough to avoid a temperature threshold that would lead to severe heat and drought.
My colleague Bob Berwyn covered the report at the time and wrote a followup story putting the timeline in context.
No credible climate scientist has said we’ll all be dead by 2030 because of climate change. To say otherwise is—sorry—a hoax.