No, your internet habits are not killing the planet.
It's probably less than the alternatives, it is not that big a problem, and there is a clear pathway to reducing them.
On the wonderful Heated website, Arielle Samuelson asks, Are your internet habits killing the planet? She has a nagging feeling, worrying about bingeing on Bridgerton because it’s on Netflix, which is hosted by Amazon Web Services, which stores it on servers in data centres that are huge energy hogs and carbon emitters.
“Because of this, the internet’s contribution to climate change already sizable. The Royal Society estimates that digital technologies—our gadgets, the internet and the systems supporting them—are responsible for anywhere from 1.4 to 5.9 percent of global greenhouse gas emissions. The scientific academy admits that it’s “an intricate challenge“ to find an exact number, but the estimate puts our web habits on par with the aviation industry. And at least one expert I spoke with said she thinks those estimates are too low.”
A spread of 1.4 to 5.9 is pretty wide, but I looked into the footnotes of the Royal Society report and found they were from a few different studies of varying ages. One, dating back to 2015, predicts that “electricity usage could contribute up to 23% of the globally released greenhouse gas emissions in 2030.” All of these studies projected wild increases in power consumption and carbon emissions.
Except it hasn’t happened. Over the last decade, the energy consumption by data centres in the USA has increased minimally. As David Sterlace of ABB notes, “this minor uptick in energy demand, which has held steady at approximately two percent of overall US energy usage, has occurred in the context of a vast proliferation in the number of smart devices and the expansion of online culture to the point of near ubiquity.”
The International Energy Agency puts emissions from data centres and networks at 0.6% of total emissions and notes that “emissions have grown modestly despite rapidly growing demand for digital services, thanks to energy efficiency improvements, renewable energy purchases by information and communications technology (ICT) companies and broader decarbonisation of electricity grids in many regions.”
When I was writing my book, Living the 1.5 Degree Lifestyle, I worried a lot about the carbon footprint of my internet usage, especially since I was writing during the depths of the lockdown and the number of gigabytes everyone was using was increasing exponentially. But it didn’t amount to much. I wrote:
“Energy is a major operating cost, so the companies have been ruthless in their hunt for efficiencies. The servers and the hardware have followed a Moore’s Law- like increase in efficiency and reduction in energy consumption per gigabyte handled. It really had to, or Google and Amazon would be sucking up every kilowatt in the country. Cooling the data centers was one of the biggest consumers of electricity, so they located many of them in cooler places and switched to chips that put out much less heat. Meanwhile, the data companies got greener. Apple claims to run the iCloud on 100% renewables, Google claims to be carbon- neutral, as does Microsoft. Netflix “offsets and buys renewable energy certificates.” Amazon, by far the largest cloud service, promised to be 100% renewable but is really only about 50% now and has been backsliding.”
A few years later, I can say that Apple is not quite telling the truth about being 100% renewable, Microsoft is missing its deadlines, Netflix’s offsets are questionable, and Amazon is still backsliding. But the fact remains that these are big companies with big electrical bills, and it is in their interest to drive efficiencies and reduce energy consumption and heat generation as much and as fast as possible.
What about AI?
Artificial intelligence takes a lot of power. According to Samuelson in Heated, “An AI-assisted search requires 10 times more power than a traditional search.” But I use Perplexity for AI-assisted searches and find I am doing way fewer of them, often with all the results laid out after one search. It is the power per useful result that matters. (I asked Perplexity how much energy it used per search, and it waffled, noting that “The energy consumption for a single search would depend on the complexity of the query, the number of sources consulted, and the specific AI models employed.”)
Open AI’s CEO Sam Altman has said that “an energy breakthrough is necessary for future artificial intelligence, which will consume vastly more power than people have expected.” He is personally investing $375 million in fusion power. But it’s possible that his AI dreams may be no more realistic than his fusion dreams.
Christopher Mims of the Wall Street Journal writes that the pace of innovation in AI is slowing, its usefulness is limited, and the cost of running it remains exorbitant. He suggests that we are caught in the classic hype cycle and that everything “seems to be predicated on the idea that AI is going to get so much better, so fast, and be adopted so quickly that its impact on our lives and the economy is hard to comprehend. Mounting evidence suggests that won’t be the case.” I have followed Mims for years, and he is often right.
Stop comparing the Internet to aviation
Samuelson is not the first to compare the carbon emitted by the internet to the emissions of the aviation industry. This is unfortunate for a number of reasons; for one, those AI chips are going to get more efficient by the day and they run on electricity, which gets cleaner every day. There is a clear path to follow. Aviation runs on fossil fuels, and there is no clear path to decarbonization.
Furthermore, it is like comparing water and single-malt whisky; one is a necessity shared by everyone, while the other is mostly a treat for rich people. The wealthiest 1% generate half the world’s global aviation emissions, while 80% of the world’s population has never been on a plane.
By 2019, 56% of the world’s population was connected to the internet. According to the Borgen Project, it has lifted people out of poverty, improved access to education, improved communication, and even improved crop efficiency. These are good things.
And, of course, Arielle Samuelson and I owe our livelihoods to it.
Carbon emissions from the internet (other than cryptocurrency and really bad DALL-E illustrations on Substack posts) are not the same as aviation emissions. We should feel guilty about flying and try to do less of it. I do. But Samuelson should not feel guilty about bingeing on Bridgerton. I don’t.
I will agree with you on this - TA-DA! - I use Perplexity as well. Not that it is the best but I find that it hallucinates the least (eg, returns a lesser number of made-up answers) - I would counsel people to double-check answers, on some basis of your own, returned from your queries.
I'm going to disagree with the final premise of this post. In the avenue of power consumption, I will quote the unknown wag: "You ain't seen nothin' yet!". You are correct in that part of the tech race is to make the millions (billions??) of GPUs and NPUs more efficient, and the machine learning processes MAY get more efficient, utterly MASSIVE data centers to hold incomprehensible yottabytes necessary for that data's organization and subsequent queries against it.
I forecast, from my retired seat in the IT sidelines, that the number of planned and built data centers will continue to ramp up, for a while, exponentially. And each one will be bigger than the previous ones by all of the major players in this field.
After all, a thumbdrive isn't going to hold all the data necessary to train all these task specific and generalized models...
Interesting comments on Perplexity – that's something I look forward to trying.
Given how often I agree with almost everything you write, it’s unusual that I find something to object to in nearly every paragraph of this post. Here’s a few …. First, I find the emphasis on the low operating energy consumption of phones and average computers to be nearly pointless. So much of the energy use occurs in the transmission/cloud, and equally if not more significant, in the production of the frequently-replaced equipment. While the electricity used to power a phone in some cases is provided through relatively clean generation, the mining, manufacturing and long-distance shipping will be much harder and take longer to decarbonize.
Second, from what I read, the energy demands of the cloud have been going up rapidly for years. Perhaps these sources are wrong, but here's one such reference:
“Amid explosive demand, America is running out of power – AI and the boom in clean-tech manufacturing are pushing America’s power grid to the brink. Utilities can’t keep up.”
https://www.washingtonpost.com/business/2024/03/07/ai-data-centers-power/?_pml=1
For example, the article says “A major factor behind the skyrocketing demand [for electricity] is the rapid innovation in artificial intelligence, which is driving the construction of large warehouses of computing infrastructure that require exponentially more power than traditional data centers. AI is also part of a huge scale-up of cloud computing.”
Re AI chips, you write “those AI chips are going to get more efficient by the day and they run on electricity, which gets cleaner every day.” But we have a great example of Jevon’s Paradox – while individual AI chips get more efficient, the overall energy consumption skyrockets. AI farms may buy renewable energy, but in many places that simply means there’s not enough clean electricity left to run heat pumps or electric vehicles.
You make a good point that aviation is primarily a global elite service, internet usage is far more widespread. True, but: I suspect access to and use of multiple streaming services is heavily slanted to the world’s wealthier residents; ownership and frequent replacement of multiple devices (phone, tablet, laptop, watch, large-screen TVs) also reflects global inequality. I would expect that responsibility for internet-related carbon emissions is heavily weighted to the rich world, even if that weighting is not quite so extreme as in aviation.
I certainly agree that internet usage CAN BE less environmentally impactful than many other standard activities in western society. If we are mindful about what we buy and how we use it, the benefits of internet usage might even outweigh the harms. But the whole internet industrial complex pushes toward greater consumption of nearly everything. So, taken as a societal whole, I would say our internet habits ARE killing the planet.