- CO/AI
- Posts
- The AI Energy Question
⚡️ The AI Energy Question
As generative AI has been embraced around the globe, the corresponding energy demands are skyrocketing.
Data centers globally account for approximately 1 to 1.5 percent of global electricity usage per the International Energy Agency
📈 Looking Forward
Experts are Predicting a dramatic increase in energy usage due to AI, estimating that NVIDIA alone could be shipping 1.5 million server units annually by 2027.
These servers, if running at full capacity, could consume over 85.4 terawatt-hours of electricity each year, surpassing the annual energy consumption of many small countries.
Training AI models is particularly energy-intensive. GPT-3 consumed approximately 1,300 megawatt-hours of electricity, equivalent to the annual power usage of 130 U.S. homes.
👨🔬 Interesting Solutions: Crusoe
Company Overview: Crusoe Energy captures excess natural gas from operations, which would otherwise be flared into the atmosphere (set on fire!) and uses it to power data centers and cryptocurrency mining.
How: Crusoe has developed modular, portable data centers that can be brought on-site to capture this excess energy and power AI’s growing energy needs.
Why: It is easier to transport data than natural gas; all the computing work done on-site is instantly online.
Reply