Cleaning up the technology industry was supposed to be easy.
Powering server racks and personal computers isn’t nearly as energy-intensive as making concrete or steel.
So until recently, eliminating planet-warming emissions in the tech sector was expected to be relatively straightforward. Tech companies positioned themselves as climate leaders and boasted of their eco-friendly bona fides.
But the sudden emergence of artificial intelligence is casting doubt on those assumptions.
Big tech companies’ bets on a new generation of power-hungry computer chips and data centres are forcing them to significantly increase their power usage.
Microsoft, Google and Amazon are investing in nuclear power. Energy demand, including from Google and Meta, is helping to keep running a coal plant in Nebraska. Natural gas providers are hatching ambitious expansion plans as they prepare to meet the new demand from the tech sector.
At the centre of the AI boom is Nvidia, the chip-making juggernaut that has, over the past few years, become one of the most valuable companies in the world.
Nvidia’s chips are incredibly power-hungry. As the company rolls out new products, analysts have taken to measuring the amount of electricity needed to power them in terms of cities, or even countries.
There are more than 5,000 data centres in the US, and the industry is expected to grow nearly 10% annually. Goldman Sachs estimates that AI will drive a 160% increase in data centre power demand by 2030.
Dion Harris, Nvidia’s head of data centre product marketing, acknowledged that AI is creating a huge spike in power usage. But he said that over time, that demand would be offset as AI makes other industries more efficient.
“There is sort of a myopic view on the data centre,” he said, “but not really an understanding that a lot of those technologies are going to be the main way that we’re going to innovate our way to a net-zero future.”
What AI might do for climate
The notion that AI can solve climate change is deeply alluring.
Global warming is one of the biggest challenges we face as a species, and the promise of a new technology magically making an existential problem evaporate holds deep appeal.
So far, however, there is scant evidence that AI will deliver a miracle fix.
Incremental gains are possible. Harris told me that already, Nvidia is helping companies in a range of industries make their operations more efficient. It recently helped Foxconn, a major producer of electronics such as iPhones, use AI to design a new factory, reducing energy demands by 30%.
“They’re doing things that they used to do before, and they’re figuring out a way to do it much faster, much more efficiently and actually use less energy,” he said.
Last year, we reported on Google’s use of AI to create a system for reducing aircraft contrails, which can lead to additional global warming. Utilities are looking to use AI as they work to keep the lights on after storms.
Harris also said that while Nvidia chips and data centres might be straining power grids today, a handful of developments could make the AI industry more efficient in coming years.
A new generation of chips can break down complex tasks, ultimately using less energy. And companies are looking at putting data centres used to train AI models close to remote energy sources, such as hydroelectric or geothermal facilities.
“The important thing to realise about AI is that it doesn’t necessarily care where it learns,” Jensen Huang, Nvidia’s CEO, said last month in Washington. “We can build a data centre near where there’s excess energy and use the energy there.”
A ‘confluence’ is the cause
The surge in power demand from the technology sector is just one part of a broader spike in power use.
For the past 15 years or so, the United States experienced relatively modest growth in energy demand. Offshoring sent many energy-intensive factories overseas. Growth in the US economy was coming from tech and finance, two industries with relatively low emissions.
But now manufacturing is coming back to America. Electric cars and heat pumps are adding demands to the power grid. At the same time, the tech industry is no longer looking so efficient.
Even before AI began straining the grid, there were signs that the efficiency gains that powered the past few decades of technological growth were faltering.
One of the fundamental selling points of the biggest tech companies was that they were scalable – producing web products that reach larger audiences at marginal cost. But that dynamic has broken down in recent years.
Cryptocurrency brought the advent of Bitcoin mining, which requires huge amounts of electricity. And now we have the AI boom.
“You have this confluence that is leading to a spike in demand,” Harris said.
For tech companies, which have long pledged to rapidly draw down their planet-warming emissions, the AI revolution has disrupted their climate ambitions.
Microsoft said its emissions had soared 30% since 2020 because of its expansion of data centres. Google’s emissions are up nearly 50% over the past five years because of AI.
Eric Schmidt, former CEO of Google, recently said the artificial intelligence boom was too powerful, and had too much potential, to let concerns about climate change get in the way.
Schmidt, somewhat fatalistically, said that “we’re not going to hit the climate goals anyway,” and argued that rather than focus on reducing emissions, “I’d rather bet on AI solving the problem.”
The AI boom is leaving the tech industry in an awkward spot. After years of claiming to be environmentally friendly, some of the biggest companies in the world are making hard trade-offs between increasing profits and lowering emissions. It’s the kind of conundrum that often leaves executives wanting to have their cake and eat it, too.
“AI is a problem,” Harris said, “and a solution.” – ©2024 The New York Times Company
