|
Getting your Trinity Audio player ready...
|
Several years ago, Google announced an ambitious target to achieve net zero carbon emissions by 2030. But the reality is that instead of shrinking, Google’s emissions have skyrocketed, increasing nearly 50% since 2019. Similarly, Microsoft reported this year that its carbon emissions have grown nearly 30% since 2020.
What’s causing this sudden increase in tech companies’ greenhouse gas emissions? The explosive growth of AI technology.
The rise of AI is keeping us tied to fossil fuels
Generative AI (such as ChatGPT, DALL-E, and similar tools that generate content in response to a request) uses enormous amounts of energy. A ChatGPT prompt requires about 10 times as much energy as an ordinary Google search. As AI tools become more complex and their usage more widespread, energy usage is increasing. According to the World Economic Forum, AI’s computational power demand doubles roughly every 100 days. This means that by 2028, “AI could be using more power than the entire country of Iceland used in 2021.”
Tech companies are building new data centers as quickly as possible to supply all this energy. According to the Los Angeles Times, PG&E recently reported that it had received 26 applications for new data centers, which would use 3.5 gigawatts of power – enough to support nearly 5 million homes. Researchers from Goldman Sachs estimate that by 2030, the data centers that power AI technology will use 8% of the total energy in the US, as compared with 3% in 2022.
This massive surge in energy demand is, quite simply, keeping us tied to fossil fuel usage. As much effort as California is making to invest in renewable energy like solar and wind, the simple fact is that we can’t generate enough renewable energy quickly enough to supply AI’s voracious needs. If we ever want to wean ourselves off fossil fuels for good, we’re going to have to put the brakes on our embrace of AI.
AI’s energy demands will impact ordinary people
It’s also probable that the rise of AI technology will raise electricity rates for ordinary people and increase the strain on the energy grid. Those new data centers need to be connected to the energy grid, which means PG&E and other power companies must build new transmission lines and other infrastructure to serve the new data centers. The cost of constructing this infrastructure – billions of dollars, according to a recent statement from PG&E – is then passed on to ratepayers. Meanwhile, affordable housing projects have to wait months for PG&E to turn their lights on, and the threat of blackouts during heat waves is ever-present.
Increased water usage due to AI
AI doesn’t just use massive amounts of energy – it uses far more water than non-AI technology. This is because all that energy radiating inside the data centers that power AI technology creates a lot of heat, and data centers use water to cool off. Both Google and Microsoft have reported sharp increases in their water consumption due to AI. (The tendency to build data centers in hot desert areas doesn’t help, either.)
What can we do about it?
First, users should be able to “opt out” of AI. You may have noticed recently that when you do a Google search, something called an “AI overview” often pops up at the top of the screen. If you use Google Chat to communicate with co-workers, helpful little summaries of everyone’s comments sometimes appear amid the conversation. We never asked for this AI content to be provided to us, and we were not given the opportunity to decline it.
If you use ChatGPT or another AI tool to help you perform tasks at work or for your personal use, do you know how much energy or water that query is using? Would it make a difference in how often you use generative AI if you did know? Many people who choose to ride a bike instead of driving, who choose to purchase an EV, or who make sure to turn off the lights when they leave a room would probably say yes. Now, experts who study the energy and water consumption of AI technology are calling for tech companies to provide this information to users.
Second, regulators should put the brakes on the recent explosion of new data centers. At the very least, government officials should be asking where all the energy these new data centers will consume is coming from and whether this massive increase in energy consumption is actually sustainable. California law requires regional planning entities to develop sustainability strategies to reduce greenhouse gas emissions, and many California cities have enacted climate action plans to achieve those strategies. Before a city approves the building of a new data center it should analyze whether that data center’s energy usage and resulting greenhouse gas emissions will move the city in the wrong direction for its own climate goals.
AI technology can help our society address problems and perform more efficiently, but we should not be blind to the environmental cost of ubiquitous, unregulated AI usage. By raising awareness of the energy and water consumption of generative AI and by demanding that our leaders not look the other way on this issue, we can get back on track to addressing climate change.
