Ilan Rakhmanov is the founder and CEO of StringGPT.
The advent of AI has ushered in transformative advancements across countless industries. But for all its advantages, this technology also has disadvantages. One of the biggest challenges in AI is the amount of energy needed to power the GPUs that train large-scale AI models. Computer hardware requires significant maintenance and upkeep, as well as uninterruptible power supplies and cooling fans.
A study found that training some popular AI models can produce about 626,000 pounds of carbon dioxide, the rough equivalent of 300 cross-country flights in the United States. A single data center may require enough electricity to power 50,000 homes. If this energy comes from fossil fuels, this can result in a huge carbon footprint. Already the carbon footprint of the cloud as a whole has surpassed that of the airline industry,
As the founder of an AI-focused company in the blockchain and cryptocurrency industry, I am keenly aware of the environmental impact of our business. Here are some ways we try to reduce this effect.
Environmentally friendly data centers
Moving your operations from large cloud companies like Google Cloud and AWS to companies that operate environmentally friendly data centers is an important first step. Sustainable data centers are designed to minimize their environmental impact by using renewable energy such as solar power, conserving water and reducing waste, while providing secure and reliable processing and storage.
These companies are often smaller and their services may cost more. And larger companies can provide grants, making it more affordable for young startups like us to run multiple servers. But choosing to use a sustainable data center can be an important step toward reducing your company’s carbon footprint.
Optimizing operations can enhance the benefits of this change. We are always working on making our AI models more efficient, which reduces training time and decreases GPU power demand.
Fight electronic waste
GPU technology is evolving rapidly, which means frequent hardware updates. Data centers typically replace their servers every few years. All those leftover computers, networking equipment, and cooling systems have to go somewhere, leading to a growing global e-waste problem.
We have launched an e-waste management program to ensure obsolete GPUs are handled responsibly. Many companies can implement an IT asset disposal program, which includes policies to reuse older hardware when possible and recycle it when not.
Charitable donations
Donating to organizations and charities that work to protect the environment and combat climate change allows businesses to offset the environmental impact of their operations. For example, we are committed to donating part of our profits to organizations fighting against air pollution. Donations to charities such as Binance’s “Protect Our Planet,” the first transparent blockchain-enabled donation platform, can easily be automated.
As more AI companies focus on sustainability, there will be increased pressure on big players like Google, Amazon, and Microsoft to make their operations eco-friendly. This collective demand is already bearing fruit: in 2022, AWS launched a tool which customers can use to calculate their IT carbon footprint.
There is a certain irony in the fact that while AI is used to optimize supply chains, monitor resource consumption and promote sustainable manufacturing processes, it also has its own significant environmental impact. As a technology company, it’s important that we keep this in mind and do everything we can to ensure the end result is positive for the planet.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Am I eligible?
