Addressing AI’s Insatiable Energy Needs

Welcome to another edition of the Sustainable Intelligence newsletter.

This is where I try to explore the intersection of artificial intelligence and environmental sustainability. My mission is to share my observations across these core themes: AI for Good, Ethical AI, Sustainable AI and AI in Sustainability. I hope to inform, inspire and engage readers interested in the potential of AI to drive positive change for both people and the planet.

Here’s what’s in stock for you today:

• News Bite: Meta launches its most powerful AI model to date.

• Deep Dive: Addressing AI’s Insatiable Energy Needs.

• Tech Trend: Divergent views in the use Gen AI

News Bite: Meta launches its most powerful AI model to date.

Meta has officially launched the Llama 3.1 405B, an advanced open-source AI model, along with its 70B and 8B versions, while also upgrading the existing 8B and 70B models. This marks the first openly available model that rivals the leading AI models, offering state-of-the-art capabilities in general knowledge, steerability, mathematics, tool usage, and multilingual translation.

Deep Dive: Addressing AI’s Insatiable Energy Needs

Globally, data centres consume 460 terawatt-hours (TWh) of electricity annually, equivalent to Germany's total usage. In the U.S., data centres used 2.5% of the nation's electricity in 2022 (~130 TWh), projected to triple to 7.5% (~390 TWh) by 2030, equating to the power used by 40 million U.S. homes.

As AI models grow larger and smarter, their power demands will increase. Reducing the power requirements for data centres is crucial to achieving AI's potential. Addressing this power issue is urgent. Organizations are exploring various solutions, and while no single fix exists, the growing focus on this problem is promising. Below are some approaches under investigation to help solve AI's energy problem.

Better Tech: Arm, originally known for battery-operated products, is now developing chips to meet AI's growing demands. Nvidia's latest AI chip, Grace Blackwell, uses Arm-based CPUs that run generative AI models on 25 times less power than previous versions. This technology focuses on reducing power use by improving compute efficiency, often described as “more work per watt.” While promising, this alone won't solve the AI energy crisis.

Alternative Energy Investments: OpenAI CEO Sam Altman has invested in several alternative energy ventures, including a solar startup with shipping-container-sized modules and nuclear startups Oklo and Helion. Microsoft plans to buy Helion’s fusion electricity starting in 2028. Google has partnered with a geothermal startup to power a large data center, and Vantage, the data centre operator has built a 100-megawatt natural gas plant in Virginia to keep one of its data centres off the grid.

Cooling Servers Down: Innovative cooling technologies are being developed to improve data centre efficiency. The Electric Power Research Institute supports methods like air-assisted liquid cooling and immersion cooling. Although liquid cooling is already in use, immersion cooling is still being developed and implemented in a few new data centres.

Flexible Future: Flexible computing is a new approach to building AI data centres, optimising computation based on electricity availability and cost. The key idea is to compute more when electricity is cheaper, more available and greener, and less when it’s more expensive, scarce and polluting. This method requires advancements in hardware, software, and grid-data centre coordination.

Direct-to-Chip Cooling Microsoft is exploring direct-to-chip cooling with Cold Plates, which offer more effective heat exchange than traditional air cooling. This system involves redesigning data centers to include innovations like sidekick—a liquid cooling system that draws heat away from the racks, similar to a car radiator.

Tech Trend: Divergent views in the use Gen AI

Global media companies hold differing views on the use of Gen AI for covering the Paris Olympics. US broadcasters are open to adopting Gen AI for Olympic broadcasting, while their European counterparts remain skeptical, arguing that Gen AI is still too nascent for roles like sports commentary. This may signal the start of regional differences in Gen AI adoption. Stay tuned.

That’s all I have for you today.

That’s all I have for you in this edition of Sustainable Intelligence.

If you have any questions or feedback please use this form. I will try my best to respond to all your questions and feedback.

If want me to feature an article in the News Bite or Tech Trend sections, please get in touch and I will see what I can do to feature it.

If you are not yet subscribed to this newsletter, please subscribe here and forward it on to all your friends who you think will benefit from it.

Thank you!

Emeka Ogbonnaya