Renewable Energy, Edge Computing, and Heat Reuse

Heat Reuse

Artificial intelligence uses a lot of energy, and almost all of that energy used to train and power AI is converted into heat. To help increase sustainability, this heat can be captured and reused in heat commericial and residential buildings surrounding the data centers. This reused heat can reduce the needs of heat production by up to 10 percent. Capturing and supplying this heat from data centers to other buildings can also be up to three times cheaper than producing that heat.

Due to the benefits found by heat reuse in data centers, European governments are passing new laws to encourage this practice. For example, heat recovery integration is now mandated for new data centers in Germany and the Netherlands, and the EU energy efficiency directive requires all large data centers to implement waste heat recovery if feasible. (Heat Pumping Technologies)

Renewable Energy Powering AI

The creation of more renewable energy sources can address some of the problems of powering the training and use of AI systems. Some large companies in AI are investing billions of dollars towards clean energy. One such example of this is Microsoft investing 10 billion dollars into clean energy such as wind and solar to help power their AI training data centers. (Straight Arrow News)

Despite billions of dollars being spent towards renewable energy, it is important to note this still doesn't solve all energy issues. The CEO of OpenAI has expressed many concerns about this issue, and has stated a breakthrough would be necessary to have this be effective enough to offset the issue. (Forbes)

Edge Computing

Currently, most large AI systems are trained at data centers that are not centralized to where the data is sourced, this requires large amounts of data transferring and is very inefficient. Edge computing is the practice of storing and processing data near the device that created the data. When data is processed near the source, that means less data is being sent, and the processing is faster. It is also easier to intelligently distribute processing tasks on edge devices compared to non-edge devices. By implementing edge computing at scale, this could result in drastically less data transmitting and therefore less CO2 consumption, it also reduces the need for inefficient and harmful data centers. (Digi)