Why AI's Energy Use Isn't Demonized Like Bitcoin's
Last month, bitcoin mining company Core Scientific (CORZ) signed a 200 megawatt (MW) artificial intelligence (AI) deal with cloud computing firm CoreWeave, the former agreeing to modify some of its existing infrastructure to host the latter’s GPUs for high performance computing operations.
Miners, bitcoiners, and technologists have been talking about the overlap of AI and Bitcoin for a while and this Core-to-Core deal marked the official collision of these two (potentially) over-hyped and over-frothed industries. The union makes perfect sense: Bitcoin miners have built out robust data centers, complete with attractive energy contracts, and as bitcoin mining trends towards lower profitability, providing AI companies with infrastructure is an obvious and straightforward way to bridge the gap (until things get better).
That said, just like with Bitcoin, not everyone is totally psyched about AI. And AI detractors have valid concerns: bias, transparency, privacy, safety, validity, and (worst of all) stealing my bad art to make even worse art.
But to someone who has at one time or another been in the throes of bitcoin reportage, there’s something very obvious mostly missing from the AI hysteria which was Bitcoin’s political Achilles heel: energy use.
AI, if it is to grow as proponents believe it should, will require a lot more energy to power the data centers which make AI possible. Investment bank Goldman Sachs predicted that data centers will use 8% of the U.S.’s total power supply by 2030 (up from 3% in 2022), of which AI is a strong driving force. Additional research from French energy company Schneider Electric suggests that AI’s share of data center energy demand will rise to 15% -20% by 2028 (up from an estimated 8% in 2023). There are countless other projections and estimates out there and none that I’ve found suggest anything except more.
Now whether this energy demand is “worth it” is a decent question for another day, but why is it that there are mountains and mountains of articles and thought pieces about Bitcoin using an “XYZ-country’s amount of energy” and not for AI?
I have some ideas.
Money, power, respect (the last one is best)
Money: The route of all people.
There are hundreds of billions of investment capital pouring into AI and its speculative future—as a proxy, just look at AI-chip-maker Nvidia (NVDA) which is up 175% this year. Bitcoin didn’t have that when the hype started and even now, near its peak, no one is actively searching for ways to invest hundreds of billions into Bitcoin-related ventures (aside from maybe one of the Bitcoin ETFs).
With all this money flowing in come the big companies: Google, Microsoft, Amazon, Meta and with it the power of influence they possess. These four companies are ubiquitous, trillion-dollar giants and masters of PR. How many people hear the word “data center” and think “Oh my, what a waste! All that energy!”? It’s not all too common.
And with all that money and influence comes respect for the big-brained intellectuals in favor of AI. Microsoft CEO Satya Nadella and Meta chief Mark Zuckerberg say AI is world-changing and good. That comes with cachet. When someone with a wacky X handle tweets that Bitcoin is world-changing and good, that comes with cachet too, but a different kind.
The worst-case scenario is pretty bad
Humans are really good at constructing realities at logical extremes, especially on the bad end. Here’s a bad case for the world with AI: AI gets really good at art and so instead of doing art we give up writing and drawing and painting and we all get to focus on work.
Sounds pretty awful.
But oh, it can get worse. Imagine the AI becomes sentient and it takes over the world and begins “harnessing humans’ heat and electricity as an energy source.” Who has the time to worry about AI’s energy cost when the stakes are so much higher?
As for Bitcoin, the worst case scenario is … what? It collapses and some people you know lose a lot of money? Or maybe it succeeds and topples the Federal Reserve and Bitcoin becomes the world’s reserve currency?
Both scenarios are longshots. In the meantime: Have you seen how much energy Bitcoin uses?
Plus people have more pressing things to worry about. Their livelihoods, for instance.
‘This affects me’
If Bitcoin succeeds, then some people get rich and, even though “fix the money, fix the world” is a common bitcoin saying, will our lives really change that much if bitcoin wins?
Meanwhile, the main narrative around AI is that it will … obsolete my job? No thank you, we must stop this at all costs.
Who cares how much energy AI uses? I have bills to pay and I need a job to make money. My only fear is that there won’t be enough of us to stem the tide.
On the other hand, lots of regular people are using AI to make art they share all day on social media, cheat on term papers or generate Wojak memes. No doubt many are using AI for salubrious purposes as well. The point is, AI’s utility is obvious to everyday people, mitigating any concern they might have about energy use, in a way that bitcoin’s value isn’t. (Censorship-resistant payments or hard-to-confiscate assets don’t matter until the day you need them, at which point they’re the only things that matter.)
AI and Bitcoin: Same, but different
To be sure, there are some who raised the energy use alarm for Bitcoin who are raising the same alarm for AI. I have written articles (and research reports) in defense of Bitcoin’s energy use and, while I won’t write any in defense of AI’s energy use, I am waiting for the day when its vast energy use becomes a main argument against AI.
Though I suspect I’ll be waiting a while, because, whether warranted or not, the relative lack of loud noises around AI’s energy use can be explained swiftly and easily: AI and Bitcoin are different.
Note: The views expressed in this column are those of the author and do not necessarily reflect those of CoinDesk, Inc. or its owners and affiliates.