Discover how China copes with AI heating problem

AI systems, especially those located in massive data centers, are major consumers of water, mainly for cooling purposes. Discover how China copes with AI heating problem in this insightful video. Learn about the innovative methods and technologies being used to keep these systems running efficiently. If you're interested in the future of AI technology, this video is a must-watch!
To give you an idea of the scale, the training of a single AI model, such as GPT-3, could demand up to 700,000 liters of water. This amount is comparable to the annual water usage of multiple households. Furthermore, by 2027, the global water demand for AI could equal half of that of the United Kingdom. For instance, in West Des Moines, Iowa, a large data center cluster that hosts OpenAI’s most advanced model, GPT-4, accounted for about 6% of the district’s water usage in July 2022. Given these factors, the question arises:
How can China manage these water resource demands without hindering the progress of AI?
China Launches a 1,400-Ton Commercial Underwater Data Center
Highlander, a company specializing in underwater data centers, has established a commercial facility in the sea near Hainan Island, China. The system, weighing 1,300 tonnes (1,433 tons), is located 35 meters below the sea surface and uses the ocean for cooling its computing components. Although specific details were not disclosed, the company stated that the module can process over four million high-definition images within 30 seconds, which is equivalent to the simultaneous operation of 60,000 conventional computers.
Highlander aims to install 100 such modules at the site, which it claims would conserve 68,000 square meters of land, 122 million kilowatt-hours of electricity, and 105,000 tons of freshwater annually. China Central Television reported that the system was launched on Friday as the first commercial underwater data center. However, Highlander had previously asserted that it had deployed a commercial system in Hainan in April. Highlander initially announced an underwater test in January 2021, in the port of Zhuhai in Guangdong. Throughout 2021, additional tests were conducted, and several Chinese regions incorporated the concept into their five-year plans. The commercial Underwater Data Center (UDC) was announced in early 2022 and was constructed by COOEC in the waters just off Lingshui Li Autonomous County, Hainan Province. Highlander has been recognized as a “little giant” (a rapidly growing, innovative company) by China’s Ministry of Industry and Information Technology. Highlander reports that it has received orders from companies such as China Telecom and SenseTime, a Hong Kong-based AI software/surveillance company.
Before Highlander’s initiative, Microsoft was the pioneer in testing the concept of UDCs, submerging its first Project Natick data center off the US Pacific coast in 2015 and serving Azure cloud loads from the seabed. This initial test was followed by a two-year trial off the Orkney Islands in Scotland, which concluded in 2020. Microsoft found that underwater data centers enhanced the reliability of IT hardware by shielding it from oxygen and accidental damage.
Microsoft pioneered the concept of undersea data centers and successfully conducted a test in 2015. However, despite driving OpenAI, they did not continue to invest, experiment, and deploy this technology. They were aware of the increasing demand for electricity and water, yet chose not to take action to mitigate the impact on natural resources.
In contrast, China has been actively testing and deploying undersea data centers to prepare for the anticipated surge in data center deployment and energy resource needs. Did China foresee the consequences that the US overlooked? As the pace of AI development accelerates, the demand for electricity and water will inevitably increase, potentially leading to adverse environmental and societal impacts. China is taking responsibility for minimizing these negative impacts while continuing to drive AI development. Highlander has become the first to commercialize Undersea Data Centers, making significant progress and already receiving orders from various AI companies. Could undersea data centers be the solution to cooling AI? We will continue to monitor this emerging technology and provide updates as they become available.
Thank you for tuning in.
Keywords: #ainews #datacentersolutions #microsoft

[@AITalkingHead]

Undersea Data Center is going to be an alternate way to cope with AI cooling and China has a long term plan to do this. Therefore, the energy and water resources could be managed in the long run without sacrificing the development of AI. Thus, this will also bring down the costs for AI operations. Would this new technology become a mainstream in AI water cooling ? Do you think that Microsoft would bring this technology back to live knowing that OpenAI has already been proven a thirsty system? Tell us what you think down below in the comment. Thank you for watching!

[@GenerativeMind01]

As energy costs become a bottleneck for AI, it is imperative that China and the US must resolve this in order to expand the utilization rate. However, this is a long-term issue and not easy to resolve overnight.

[@asiabusinesscircle4265]

AI is definitely expanding and energy is king and whichever country that has the natural resources are gonna win in this AI race. This is as simple as that! 🙂

[@GenerativeMind01]

Underwater Datacenter is the future to resolve massive energy hungry, and China is certainly prepared for this as it is the leading manufacturing nation. Although this idea was initiated by Microsoft, China, once again, take it to real-life. Congrats!

[@asiabusinesscircle4265]

Underwater Datacenter is a cool idea and nice to know that China has implemented it successfully and this will certainly reduce the need for land and the energy to cool down the racks. Great stuff!

[@peteryu1053]

The energy requirement for AI will only go up and increasing in a geometric rate as the utilization becomes more and more widespread across different industries and communities. If this is not addressed quickly, it will limit the growth of use cases and utilization.

[@julianezman4458]

All these comments be AI

[@AITalkingHead]

No, you wrong! We are very much human and monitoring the channel and comments 24x7.

[@RusticRaver]

lol, AI making videos about AI, robots have balls!