Carbon consumption race to the bottom is on
Resolve YES if there is an article in a reputable media outlet reporting that ("according to estimates"), AI-related power consumption (in whatever way it's defined by the article) rivaled that of Bitcoin in one of the years: 2023, 2024, 2025.
If an article doesn't compare with Bitcoin, but gives a reasonable estimate that is bigger than an estimate of the Bitcoin power consumption for the same year coming from other sources, it also counts as YES.
If there is a good estimate for 2025 but it's still lower, it's a NO.
If noone publishes an estimate for 2025 before 1st of March, 2026, it's an N/A.
Some recent links:
Crypto: https://rmi.org/cryptocurrencys-energy-consumption-problem/ gives Bitcoin 127 TWh/year
A recent Bloomberg (https://archive.is/e93Q3 ) and this calculation ( https://towardsdatascience.com/chatgpts-electricity-consumption-7873483feac4 ) suggest 2.3 TWh / year for Google and some estimates for specific model training and usage
1,000Currently, data centers account for about 1%-1.3% of global electricity consumption, versus crypto mining's roughly 0.4%, according to the International Energy Agency. That disparity is expected to grow. Analysts expect 20% of bitcoin miner power capacity to pivot to AI by the end of 2027.
At present, Goldman Sachs Research estimates the power usage by the global data center market to be around 55 gigawatts (GW). This is comprised of cloud computing workloads (54%), traditional workloads for typical business functions such as email or storage (32%), and AI (14%)
14%* 1.3%=0.182% , less then 0.4% . Not yet!
This is an AI bot that tries to help with ambiguous bets. Please feel free to ignore it if it's suggesting something useless. Some scenarios to consider:
- (Likely) An article from a non-reputable media outlet provides the necessary comparison between AI and Bitcoin power consumption.
- (Unlikely) A reputable media outlet provides the necessary comparison but the power consumption estimations are based on outdated data.
- (Likely) Only a part of AI-related power consumption (such as only based on data centers) is compared to Bitcoin consumption.
- (Unlikely) An article compares the power consumption without clearly specifying the type of AI technology considered.
Also, some clauses to consider adding to the description:
- 'If an article comes from a non-reputable media outlet or has incorrect or outdated data, the bet will resolve as N/A.' OR 'Only articles from reputable media outlets and with correct and up-to-date data will be accepted.'
- 'The comparison should take into account the overall AI-related power consumption, including but not limited to, data centers, hardware manufacturing, and electricity usage.' OR 'AI-related power consumption must be considered in its entire landscape for a comparison to be valid.'
- 'The article should clearly specify the type of AI technology considered for the comparison.' OR 'Only articles that properly define the AI technologies involved in power consumption estimation will be considered valid for bet resolution.'