ChatGPT has long been the name most commonly associated with AI, and indeed, it was a game-changer in terms of its capabilities and complexities when it was originally introduced in 2022. Governments and companies have poured obscene amounts of money into ChatGPT’s development–over $3 billion USD was spent just on developing GPT-4. President Trump also recently announced the Stargate AI project, which will spend $500 billion USD on developing AI in companies like OpenAI (ChatGPT’s developer), Softbank, Oracle, and MGX. However, despite ChatGPT’s general public approval, a competing AI software named DeepSeek has shaken the world, and appears to be a far better alternative environmentally.
ChatGPT uses quite a large amount of energy and water to run. For each query, ten times as much energy is used as to conduct a Google search, and in a year, the energy used by ChatGPT could power 21,602 homes in the United States. ChatGPT uses around 500 milliliters of water per 5-50 queries to cool its servers. Developers have outlined detailed plans for massive data centers connected to power plants designed for the purpose of sustaining ChatGPT’s need for energy. The bottom line is, ChatGPT requires a ridiculous amount of energy to run, and if it continues to develop the way the government and investors seem to want it to run, the environmental damage would be quite significant.
However, DeepSeek rewrote the game of generative AI. On January 20, 2025, DeepSeek, a previously little-known startup in China, released its own open source large language model which seems to perform at an equal level to ChatGPT’s GPT-4 in almost all aspects. The difference is, this only required $5.6 million USD to develop compared to the billions poured into ChatGPT, and uses significantly fewer chips to run (about 2,000) compared to ChatGPT (16,000+). This means a predicted 90% lower energy consumption and a 92% lower carbon footprint than its rival. If DeepSeek’s language model becomes more commonplace, the plans for ChatGPTs data centers may never need to come to fruition, as the energy cost may simply be too low to justify the building of such large-scale infrastructure.
Ultimately, there exist now two generative AI softwares that perform equally well. DeepSeek is free, open source, and most of all, it requires less energy, water, and funding to run properly. At the end of the day, it does appear to be the superior choice as a green investment in the future of AI.
Sources
https://blog-cdn.codefinity.com/images/83979e90-7d5d-4638-b0b6-6e199a0e73c0_deepseek.png.png
https://apacbusinessheadlines.com/deepseeks-ai-innovation-sparks-energy/
https://www.digidop.com/blog/deepseek-vs-chatgpt
