The Reality of AI Power Usage
AI power usage generates significant controversy. Headlines paint it as an environmental catastrophe waiting to happen. The reality proves more nuanced and potentially more optimistic than these dire warnings suggest.
A ChatGPT query uses 10 times more energy than a Google search. This sounds alarming until one realises it equates to running your hairdryer for six seconds. The entire data centre industry, including all AI operations, accounts for just 1.5% of global electricity consumption.
Here is a rundown of the more pressing issues with AI power usage.
We can power AI with clean energy
Data centres power everything from email to streaming services to artificial intelligence. The total energy footprint of all data centres in the world remains remarkably modest at 1.5% of global electricity use. This figure includes all AI operations, from training to inference.
To put this in perspective, AI operations consume approximately 54 TWh annually, representing just 0.2% of global energy output. By comparison, Bitcoin mining alone uses 175 TWh, nearly triple AI usage, accounting for 0.6% of global energy. Unlike AI which drives technological progress and can help solve climate challenges, cryptocurrency mining provides questionable utility while consuming significant resources.
Another comparison: 1.5% roughly matches aviation industry emissions. However, aviation remains heavily dependent on fossil fuels, whereas data centres can transition to renewable energy sources. The tech industry has unprecedented opportunities to lead in sustainable operations.
Solving climate challenges requires abundant clean energy rather than solely focusing on reduced consumption. Human behaviour and technological progress make consumption reduction an unrealistic primary strategy. Instead, we must focus on transitioning to sustainable power sources and optimising resource usage.
We can use AI to solve climate challenges
AI serves as a powerful ally in our climate fight. Energy companies utilise artificial intelligence to optimise renewable energy systems, maximising their efficiency. AI can aid its own power transition to sustainable energy sources - no other use of power has this potential.
Other environmental applications already show promising results. Teams use AI systems to detect methane leaks from space, enabling rapid response to emissions. Agricultural researchers employ machine learning to breed heat-resistant crops, securing our food supply against climate change. Materials scientists harness AI to design new carbon capture compounds.
The real challenge is not power, but water
The environmental impact of AI extends beyond electricity consumption. Modern data centres require substantial water resources for cooling, and many new facilities are being constructed in regions already experiencing severe drought conditions. This presents a more immediate environmental challenge than power consumption.
Unlike electricity, which increasingly comes from renewable sources and benefits from rapidly advancing solar technology, fresh water remains a finite and unevenly distributed resource. The suggestion to reopen closed coal power plants for AI operations fundamentally misunderstands the challenge: power generation will become increasingly sustainable, but water scarcity cannot be solved through technological advancement alone.
The Path Forward
The solution lies not in abandoning AI development but in pursuing it responsibly through several key strategies:
Sustainable Power Sources
Our data centres must run on clean, renewable power sources. This transition requires strategic planning and investment but remains technically feasible. The rapid advancement of solar technology makes this increasingly achievable.
Strategic Water Management
Data centre locations need careful selection based on sustainable water availability. Regions facing drought conditions cannot support additional cooling requirements. This geographical constraint may shape the future distribution of AI infrastructure more than power availability.
Efficiency and Purpose
Continuous improvement in model efficiency reduces resource consumption while maintaining capabilities. This includes both training and inference optimisation. More importantly, we must prioritise AI applications that directly combat climate change. The technology possesses unique capabilities to address environmental challenges at scale.
AI power usage warrants attention but not panic. The technology presents both environmental challenges and opportunities. Our task involves maximising its potential for climate solutions while minimising its resource footprint. The real challenge lies not in power consumption, which renewable energy can increasingly address, but in responsible water usage and ensuring AI development serves meaningful environmental purposes.
I would like to thank Roisi Proven, Romilly Cocking, Tadas Tamošauskas, and Aimee Rivers for their thoughtful contributions to this discussion. Additional valuable insights came from Ajay Prabandham and Guillaume Roy.
More articles
Building the Future
Something has been on my mind for months. The rapid evolution of AI agents has opened up possibilities I cannot ignore.
We are witnessing the emergence of semi autonomous agents that will fundamentally reshape how we work and communicate. The opportunities in this space are extraordinary. I am diving deeper into this world of AI agent development and product creation.
My newsletter is evolving. Instead of dispensing tips from a position of authority, I invite you on a journey of discovery. I will document my experiences building with AI, how to apply my tech experience in a new world, and navigating the inevitable struggles and setbacks.
Read on for several key areas I am exploring.
Read moreAI Therapists: Self Reflection With AI
“I am not ready for this conversation.” We all know that feeling. The endless mental rehearsal. The anxiety building with each imagined scenario.
But what if I told you that AI could be your practice partner for both difficult conversations and deeper self-reflection?
Coding with AI: How To Do It Well And What This Means
I am shipping AI-first production code every day. Not experimental features. Not throwaway prototypes. Real, deployed, mission-critical code powering Cherrypick’s tens of thousands of users.
Social media overflows with “vibe coding” demonstrations. These flashy but superficial examples show AI apparently conjuring perfect code in seconds. The reality of professional AI-assisted development runs much deeper. Real production work with AI is messier, more nuanced, and demands rigorous thinking, but very effective.
This is not about magical code generation. It is about a new way of thinking about development. It requires substantial real-world development experience to do well: the onus is upon those of us with this experience to teach the next generation how to harness these tools effectively.
This is how I am doing it, what it all might mean, and how we can help others find the way.
Read morePrompting Sucks (And What We Can Do About It)
Prompting sucks. If you have spent any time working with LLMs, you already know this.
It is not just that prompting is difficult - it is fundamentally broken as an approach to working with AI. It is brittle, model-specific, and endlessly repetitive.
Here is why we need to move beyond prompting, and what we can do about it.
Read moreHow to Build a Robust LLM Application
Last month at Cherrypick we launched a brand new meal generator that uses LLMs to create personalized meal plans.
It has been a great success and we are pleased with the results. Customers are changing their plans 30% less and using their plans in their baskets 14% more.
However, getting to this point was not straightforward, and we learned many things that can go wrong when building these types of systems.
Here is what we learned about building an LLM-based product that actually works, and ends up in production rather than languishing in an investor deck as a cool tech demo.
Read more