Amazon's 100% Nuclear Bet: Renewable Energy and Carbon Credits will Reshape the AI Data Center Landscape
Talen, NuScale
Google said right now AI is representing 10% to 15% of their data center power use or 2.3 TWh annually. Per International Energy Agency, training a single AI model consumes more power than 100 households in a year.
The explosive growth of AI has necessitated a parallel expansion in renewable data center capabilities. The past 4-weeks have been incredibly busy in the expert network field, with a lot of our discussions focusing on data centers. A key topic of interest for many hedge funds and investment banks is the source of renewable energy for these data centers. This concern is particularly relevant for AI data centers, which use 5 times more power than traditional CPU-based data centers. In this blog, we'll go over some of the frequent questions we've encountered during these expert calls.
According to computer scientist Kate Saenko, the development of GPT-3 emitted over 550 tons of CO2 and consumed 1,287 MW hours of electricity. In other words, these emissions are equivalent to those generated by a single individual taking 550 roundtrip flights between New York and San Francisco. GPT4 is likely to be around 5X of that number. Given that GPT-3 required 1,287 MW for end-to-end training, and considering the factors mentioned above, I would estimate that training GPT-4 once from end to end could have consumed around 4,000 to 5,000 MW.
These power-hungry data centers are the backbone of the AI revolution, processing vast amounts of data at lightning speeds. The environmental footprint of this substantial energy consumption is increasingly worrying. Many leading tech companies are pledging to power their data centers entirely with renewable energy, with Meta having already achieved this goal.
In this article, we will go through:
The Future of Renewable Energy in Data Centers
Wind: Procurement options per State
Why North Dakota is getting disproportionate attention for AI Data Center?
Solar: Procurement options per State
Hyperscalers’ move with Nuclear Power & Conversion from Coal plants to Nuclear
Carbon Credits: A Tool for Sustainable Data Center Operations
Please note: The insights presented in this article are derived from confidential consultations our team has conducted with clients across private equity, hedge funds, startups, and investment banks, facilitated through specialized expert networks. Due to our agreements with these networks, we cannot reveal specific names from these discussions. Therefore, we offer a summarized version of these insights, ensuring valuable content while upholding our confidentiality commitments.