· 2 min read
illuminem summarizes for you the essential news of the day. Read the full piece on Deutsche Welle or enjoy below:
🗞️ Driving the news: Chinese AI firm DeepSeek has launched a new chatbot that it claims is more efficient and less power-intensive than models from U.S. giants like OpenAI and Google
• This could challenge assumptions that AI advancements will require massive expansions in energy-hungry data centers
• If true, DeepSeek’s technology could significantly reduce AI's environmental footprint
🔭 The context: AI models currently demand vast amounts of energy and water, with over 8,000 data centers consuming 1–2% of global electricity
• Projections suggest that data center power consumption could grow by 160% by 2030, as AI adoption increases
• Some companies, including Microsoft, are exploring nuclear power for AI, but fossil fuels still dominate energy sources for data centers
🌍 Why it matters for the planet: The energy and water consumption of AI is a growing concern, with AI model training consuming millions of liters of water
• Without energy-efficient alternatives, data center emissions could double by 2030
• If DeepSeek’s AI can deliver high performance with less energy, it could set a new standard for sustainable AI
⏭️ What's next: If DeepSeek’s model is truly efficient, it may reduce the need for large-scale data center expansion
• However, increased efficiency could also lead to greater AI usage, potentially offsetting sustainability gains due to the Jevons paradox
• The broader industry must balance innovation with environmentally friendly AI development
💬 One quote: "AI has a huge, a ferocious appetite essentially, for energy." — Paul Deane, senior lecturer in clean energy futures at University College Cork, Ireland
📈 One stat: AI could withdraw 6.6 billion cubic meters of water annually by 2027—six times Denmark’s yearly consumption
Click for more news covering the latest on green tech