The Jevons Paradox in AI Infrastructure

⚡ Why Efficiency Gains Will Accelerate Energy Demand

Good Morning,

I’ve been in talks with various Datacenter nerds and environmental writers thinking about the impact of AI Infrastructure projects on the future of energy. Energy and compute have a weird relationship in 2025. You can expect more articles on this theme in due course from this publication.

Yesterday we learned that Google’s (Alphabet’s) Capex is actually going up more than we thought, to $75 billion in 2025, much of that going to their AI efforts. This means BigTech’s capex in 2025 will be over $300 Billion, a stupendous amount and a big increase from 2024.

BigTech and Wall Street have been in damage control due to DeepSeek’s innovations, citing the Jevons Paradox, the idea that as the cost of using a resource falls, demand will go up — not down.

The core idea behind Jevons Paradox is that improvements in efficiency lower the cost of using a resource, which can lead to increased demand.

• What does DeepSeek mean for energy and climate anyways?

• DeepSeek serves as a contemporary example of Jevons Paradox, impacting both AI Infrastructure capex and Nvidia’s stock.

I asked Grace Shao of AI Proem, for her take on Jevons Paradox, since she is one of the Datacenter nerds I had mentioned that I respect. Morgan Stanley or Satya Nadella, a bit less so. So is less more? Susan Su makes some good arguments:

“Having a much cheaper model won’t change how retail users like you and I use ChatGPT, but for developers who are looking to embed AI into other use cases where businesses must eek out additional margin on top of what they’re already paying for inputs, this is a game changer.” - Susan Su

It’s still not clear how large language model, GPU, hardware, datacenter cooling technologies and energy innovation and their improved efficiency might impact the future of energy required for all the compute of that supposed higher adoption. Nor when or if BigTech will see legit ROI on their AI Infrastructure efforts of the mid 2020s. BigTech spending on compute is a strategic bet, but there are bulls and bears here too.

Indian AI Infrastructure Supremacy

Over in India, Mukesh Ambani's Reliance wants to build an AI datacenter campus 5x the biggest by Microsoft. It could be the world’s largest, is expected to reach a total capacity of 3 gigawatts, significantly boosting India’s current data center capacity, which is estimated at under 1 gigawatt. So it’s not just American BigTech and Cloud players involved preparing for a compute bonanza of the future AI.

While DeepSeek claims to use far less energy than its competitors, but there are still big questions about what that means for the environment. Meanwhile Meta is investing in Solar for its Datacenter projects. Meta recently signed a deal with Spanish renewable developer Zelestra for 595 megawatts of solar power in Texas, just two weeks after signing a separate solar deal with utility company Engie.

Goldman Sachs Research estimates that data center power demand will grow 160% by 2030.

This is why we need to understand Jevons Paradox more deeply in relation to AI Infrastructure, the future of energy and the future of artificial intelligence more deeply.

AI Proem Top Picks 🔥

Read

Grace Shao

on AI Infrastructure deep dives.

• Everybody Losing Sleep Over DeepSeek: Industry Implications to LLMs and AI Infrastructure

• AI Arms Race Far From Over: Chips is Only Half the Game, Infrastructure is the Other

• DeepSeek V3 puts China AI on the global map: consumer use and capital expenditure implications

• Robot Everything: Robot Girlfriends, Robot Firefighters, Robot Dogs, Robot Maids and Beyond.

Grace is a great source for ideas around Chinese innovation, physical AI, AI Infrastructure and deep musings on innovation. She also often writes op-ed pieces on AI, tech, and corporate governance for Fortune, The Diplomat, EIU, and other international publications.

The Jevons Paradox in AI Infrastructure: Why Efficiency Gains Will Accelerate Energy Demand

When steam engines became more efficient in the 1800s, coal consumption skyrocketed. As AI gets dramatically cheaper, history is about to repeat itself.

DeepSeek’s R1 release last week sent shockwaves through the U.S. power and technology sectors, causing significant declines in stock prices and raising doubts about future energy demands for AI infrastructure and the direction of AI capex.

Specifically, many power companies heavily linked to the surge in data centers within the tech industry experienced substantial drops in their stock prices. Despite being among the top performers in the S&P 500 earlier this year:

● Constellation Energy: Down over 16%

● Vistra Corp: Fell by over 16%

● GE Vernova: Dropped approximately 18%

● Talen Energy: Decreased by more than 15%

However, Microsoft CEO Satya Nadella was one of the first to say that this might be the wrong reaction to increased efficiency.

• For less than $2 a week, get access to my best content.

• Read deeper with AI Proem:

If you enjoy analysis on Datacenters, AI Infrastructure and energy, this article is for you.