One hour of Claude code— well, I’d guess it would be comparable to an hour of driving an electric car. How to know?
I'm not sure how many queries is equivalent to an hour of Claude code use, but maybe 5 seconds, which means an hour of continuous use = 216 Wh, or ~50x less than an electric car.
OP has a longer article about LLM energy usage: https://hannahritchie.substack.com/p/ai-footprint-august-202...
For a long input of n tokens from a model with N active parameters, the cost should scale as O(N n^2) (this is due to computing attention - for non-massive n, the O(N n) term is bigger, which is why API costs per token are fixed until a certain point and then start to rise). From the estimates from [1], it's around 40Wh for n=100k, N=100B. I multiply by 2.5 to account for Opus probably being ~2.5x larger than gpt-4o, and also multiply by 2 to pessimistically assume we're always close to Opus's soft context limit of 200k (it's possible to get a bigger context for extra cost, but I suspect people compact aggresively to not have to use it). That gets me 7.2J/t, which at a rough throughput estimate of 20t/s gives me power of 144W. Like a powerful CPU or a mediocre GPU, and still orders of magnitude lower than a car.
[1] https://epoch.ai/gradient-updates/how-much-energy-does-chatg...
"A lot of energy used for cooling": hyperscale data centers use the least cooling per unit of compute capacity, 2-3x less than small data centers and 10-100x less than a home computer.
"Water consumption is enormous": America withdraws roughly 300 billion gallons of fresh water daily, of which IT loads are expected to grow to 35-50 billion gallons annually by 2028. Data center water demands are less than a rounding error.
"distributed and does not suffer from the same problems": technically correct I guess but distributed consumption has its own problems that are arguably more severe than centralized power consumption.