Asia Morning Snapshot: First Comparative Climate Data for AI and BTC Drops – Controversy Incoming?

New data from Mistral AI has set off a fresh wave of comparisons between artificial intelligence and Bitcoin, reigniting debate over which technology poses the greater environmental challenge — and what that really means.

In a rare move for the AI industry, Mistral disclosed the environmental costs of its top large language model, Mistral Large 2. Over 18 months, the model emitted 20.4 kilotonnes of CO₂, used 281,000 cubic meters of water, and consumed 660 kg of antimony-equivalent minerals. In practical terms, a single response from its chatbot Le Chat produces just 1.14 grams of CO₂, 45 mL of water, and 0.16 mg of minerals — a small footprint for day-to-day use.

By comparison, Bitcoin’s environmental toll appears far more substantial. A single transaction emits between 600–700 kg of CO₂, consumes 17,000+ liters of water, and creates over 130 grams of electronic waste. In 2023 alone, the entire Bitcoin network emitted 48 million tonnes of CO₂, used over 2 billion liters of water, and generated more than 20,000 tonnes of e-waste, according to the Cambridge Centre for Alternative Finance.

But the Full Story Is More Complicated

Bitcoin critics often cite such figures to push for mining bans — but not all energy sources are equal. A March 2023 survey by crypto fund Batcoinz found that more than 40% of Bitcoin’s energy comes from renewables like hydro (23.1%), wind (13.9%), and solar (5%). Nuclear adds another 7.9%, while gas and coal represent 44%.

AI models may benefit from greener energy by default — especially in Europe, where nuclear accounts for over 22% of the grid. Mistral’s relatively low emissions are partly due to its location. If trained in fossil-heavy U.S. regions, those numbers would look very different.

Usage vs. Design: Who Wins?

The underlying models drive vastly different energy behaviors. Bitcoin consumes power continuously to maintain its proof-of-work system, regardless of usage — every 10 minutes, a block is mined. This creates a fixed, time-based energy cost.

AI models, on the other hand, consume power based on usage. Training frontier models like GPT-4 or Gemini is resource-intensive, but day-to-day chatbot responses are comparatively light. That means AI’s marginal cost is low — but its peak usage can still demand significant compute and cooling.

Mistral’s report helps define new benchmarks for evaluating digital sustainability. And while AI is far more efficient per interaction, Bitcoin’s defenders point to its utility: securing a decentralized monetary system, where energy use supports global financial infrastructure.

Still, both industries face pressure to go greener. Bitcoin’s halving cycles and shift toward renewables are gradually improving its energy profile. AI providers, meanwhile, are racing to reduce the carbon and water costs of model training.

As computation becomes central to every part of the global economy, understanding — and improving — the sustainability of these technologies will only grow more urgent.

Market Snapshot – July 23, 2025

• Bitcoin (BTC): Trades at $119,500, slipping from last week’s peak of $123,100. Bearish retail pressure on Binance has dragged Net Taker Volume under $60 million, hinting at waning bullish momentum.
• Ethereum (ETH): Down 3.1% to $3,696, with signs of an incoming correction despite steady institutional buying.
• Gold: Rose nearly 1%, reaching a five-week high of $3,430.41, as traders hedge against trade uncertainty and softening U.S. bond yields.
• Nikkei 225: Gained 1.71% on news of a new U.S.-Japan trade deal that includes a 15% tariff shift — boosting export expectations.
• S&P 500: Inched up to a record 6,309.62, buoyed by earnings optimism despite mixed broader market sentiment.

About The Author

ctleditor

See author's posts