The shocking truth behind AI's water consumption

China is intensifying its commitment to AI, channeling its resources towards the establishment of expansive new data centers - a process that requires an astonishing volume of water. Discover the shocking truth behind AI's water consumption and its impact on the environment. Learn about the water shortage crisis and how companies like Microsoft and OpenAI are using excessive amounts of water.
The process of training and maintaining AI models is notoriously energy-intensive, producing a significant amount of heat. To prevent data centers from overheating, companies employ water to cool the hardware. As per China Water Risk, by 2030, China could increase the number of its data centers threefold, amounting to approximately 11 million data center racks that accommodate servers and other equipment. However, it’s not just China. The surge in AI is already resulting in a colossal amount of water being utilized globally, including in the US. Last year, it was discovered that Microsoft, a partner of OpenAI, consumed a staggering 185,000 gallons of water just to train GPT-3, an amount sufficient to cool an entire nuclear reactor. Google also confessed in its 2023 Environmental Report that it had expended a massive 5.6 billion gallons of water in 2022. AI chatbots are significantly more energy-consuming than traditional online information retrieval methods, as they rely on specialized chips that use vast amounts of electricity.
A colossal Microsoft data center located in Goodyear, Arizona is consuming vast amounts of the desert’s water to bolster its cloud computing and AI initiatives, as reported by The Atlantic. An insider with knowledge of Microsoft’s Goodyear facility informed The Atlantic that it was specifically constructed for use by Microsoft and OpenAI, which receives substantial funding from Microsoft. Both companies refrained from commenting on this claim.
Microsoft’s data centers in West Des Moines, Iowa, consumed vast quantities of water last year, as reported by the Associated Press earlier this month, to maintain cool temperatures while training OpenAI’s ChatGPT-4, the most advanced publicly available large language model from the Microsoft-supported company. Critics highlight an additional troubling detail: this occurred during a drought that has lasted more than three years, further straining a water system that has been so parched this summer that outdoor enthusiasts couldn’t even paddle canoes in local rivers.
Beyond electricity, water is emerging as a potential constraint for nations and corporations seeking to expedite their AI research and development. Given that AI is still in its nascent phase and considering the impact of climate change, water resources are likely to become a contentious issue in the discourse on AI progress. It may also turn into a limited resource, potentially sparking political discussions between the US and China.
Thank you for watching!
Keywords: #artificialintelligence #microsoft #datacenters

[@AITalkingHead]

Water is going to be the next resource to fight for after electricity, and only the countries that have amply water resources can expand the development of AI. Perhaps, there will be new technology to cope with water thirsty AI systems. However, water will become an obstacle for AI progress. China and the US understand this very well but the development of AI is accelerating much faster than the water can produce. Is water going to be a determining factor with the future of AI? Can both China and the US solve this water resource issue soon without impacting the development of AI? Please share your thoughts down in the comment below. Thank you for watching!

[@GenerativeMind01]

China and the US are racing to find a new technology to cope with the heat problem in AI systems instead of relying water resources. It will be interesting to see how these two tech giants to come up with some ingenious ways to solve this problem.

[@GenerativeMind01]

Water is definitely a challenging resource in AI and the data center locations are going to be dictated by the water supply making it difficult to expand the availability of AI. Underwater data center is going to be a sensible option now.