Energy and e-waste – the AI tsunamis

Key Takeaways:

– The record-breaking uptake of ChatGPT and other Large Language Models (LLMs) has generated significant interest and investment in generative AI.
– Generative AI has the potential to bridge the linguistic gap between humans and machines, leading to automation and improvement in various aspects of our lives.
– Data center businesses play a crucial role in providing the necessary infrastructure for running generative AI applications.
– High-density power, modular architecture, high-bandwidth connectivity, and advanced cooling are critical factors for data centers supporting generative AI.
– The biggest challenge in supporting generative AI is the surge in power loads, as generative AI models require significantly more energy than traditional CPUs.
– Forecasting the power requirements of generative AI is challenging, but it is expected to significantly increase data center capacity in the coming years.
– The rise of generative AI will also result in a wave of e-waste due to the rapid refresh rate of AI equipment.
– E-waste is a fast-growing waste stream, and proper recycling and disposal of electronic equipment are crucial.
– The industry will need to address the challenges of power consumption and e-waste while also meeting zero emission targets.
– Low-to-no-carbon power sources, such as renewables and microgrids, will be crucial in addressing the power challenges of generative AI.
– Recycling, remarketing, and secure disposal of old AI equipment will be essential for efficient performance and reducing environmental impact.
– The infrastructure industry that supports generative AI will undergo significant changes and opportunities for growth.

TechRadar:

The record-breaking uptake of ChatGPT has raised huge interest- and investment – in generative AI. The ability of ChatGPT and other Large Language Models (LLMs) to bridge the linguistic gap between humans and machines has caught the popular imagination and raised awareness of the potential to automate and improve many aspects of our lives. This piece will look at the phenomenal new technology from the point of view of the IT infrastructure it will need to succeed, and how two key challenges – energy and e-waste – will require particular attention.

Step change

Generative AI will power more and more applications over the coming decade. ChatGPT, DALL-E, GitHub Copilot and Stable Diffusion, are just the first generation, creating and sorting images, answering complex questions, creating websites, and making programming accessible to all. Data Centre businesses provide the infrastructure for many High-Power Compute (HPC) configurations running generative AI and have developed specialist facilities that meet their needs. High-density power, modular architecture, high-bandwidth training (input) and inference (output) connectivity and advanced cooling are all critical factors for customers. This fascinating process supports new applications that promise to accelerate innovation and even save lives, starting the step change in infrastructure design that our industry will have to make.

Mark Kidd

EVP & GM at Iron Mountain Data Centres & Asset Lifecycle Management.

Power surge

Source link

AI Eclipse TLDR:

The record-breaking success of ChatGPT and other Large Language Models (LLMs) has sparked significant interest and investment in generative AI. These models have the ability to bridge the gap between humans and machines, leading to the automation and improvement of various aspects of our lives. However, this new technology also presents challenges in terms of energy consumption and e-waste.

Generative AI is expected to power more applications in the coming decade, with models like ChatGPT, DALL-E, GitHub Copilot, and Stable Diffusion already demonstrating their capabilities in creating images, answering complex questions, creating websites, and making programming accessible to everyone. Data centers play a crucial role in providing the necessary infrastructure for running generative AI, requiring high-density power, modular architecture, high-bandwidth connectivity, and advanced cooling.

The main challenge in supporting generative AI is the significant surge in power loads. These models use graphics processing unit (GPU) chips, which consume 10-15 times more energy than traditional CPUs. Many models have billions of parameters and require efficient data pipelines during their training phase, which can take months to complete. For example, ChatGPT 3.5 has 175 billion parameters and was trained on over 500 billion words of text, requiring 300-500 MW of power. This is a substantial increase compared to the typical power requirements of a data center, which range from 30-50 MW.

Forecasting the future power requirements of generative AI is challenging, but it is expected to significantly increase current demands. Analysts predict that global data center capacity could double in five years and quadruple in 10 years, with a potential 25% compound annual growth rate (CAGR) if generative AI is included. This rapid growth in power demands will coincide with the need to address the climate crisis and transition to low-to-no-carbon power sources, such as renewables, microgrids, batteries, hydrogen, and nuclear power.

Another challenge posed by generative AI is the increase in e-waste. The constant innovation in AI chips and GPUs will lead to the disposal of older equipment, contributing to the already growing e-waste stream. Proper recycling and remarketing of these components will be essential to ensure efficient performance and minimize environmental impact. Although the industry has been slow to integrate sustainable practices, the uptake of IT asset lifecycle optimization, recycling, remarketing, and secure disposal is expected to accelerate.

In conclusion, the success of generative AI has highlighted the need for robust IT infrastructure to support its growth. The challenges of increased power demands and e-waste must be addressed through the use of low-to-no-carbon power sources and effective recycling and remarketing practices. The generative AI revolution will not only transform the industries that utilize its applications but also revolutionize the infrastructure industry that supports it.