Communications of the ACM
GenAI: Giga$$$, TeraWatt-Hours, and GigaTons of CO2
For more than a decade, we have speculated about the impact of artificial intelligence (AI)/machine learning (ML) on the environmental sustainability of computing (see ACM2). It has become clear that Al's carbon emissions (scope 2), lifecycle carbon (scope 3), and other negative environmental impacts are growing explosively. Generative AI capabilities and applications exemplified and popularized in ChatGPT, DALL-E 2, Stable Diffusion, and Copilot, are the drivers. The evidence:
Giga$$$s of increased spending on AI computing equipment and infrastructure is driving a dramatic increase in infrastructure: AI computing silicon and datacenters.
Comments
Andrew Chien
On August 23, 2023, NVIDIA reported quarterly datacenter revenue at over $10B. This would reflect a 25% higher number than used for estimation in this article. Updating would produce a 1,200MW and 10.5 TWh and over 4 Gigatons of CO2.
At the same earnings call, NVIDIA guided for more than $12B in the upcoming quarter. This would correspond to 1,440MW, 12.5TWh, and over 5 Gigatons of CO2.
Analysts have quoted NVIDIA's AI chip market share at 70%, so allowing for the other vendors increase, this would produce estimates of 2,044MW, 17.85TWh, and over 7.1 Gigatons of CO2.
All of these extimates are based on next quarter estimates annualized. The AI datacenter market may well continue to grow, making these numbers an underestimate. This seems likely, as GPU vendors have indicated clear demand well into 2024.
The sustainability challenges for computing (and specifically AI) are growing rapidly!
Displaying 1 comment