Amazon is stepping up its environmental efforts by testing a groundbreaking carbon-removal material for its data centers. The company, which is tackling the growing emissions linked to the artificial intelligence systems powering these centers, has partnered with Orbital Materials, a startup that used AI to design the innovative substance.
Jonathan Godwin, CEO of Orbital Materials, explained that the new material acts like an atomic-level sponge, with cavities precisely sized to capture CO2 without interacting with other elements. This targeted approach could be a game-changer in carbon filtration.
One of the appealing aspects of the new material is its cost-effectiveness. Godwin estimates that the material could account for just 10% of the cost associated with renting a GPU chip for AI training, significantly less than the price of traditional carbon offsets.
Meanwhile, the demand for energy in data centers is rising, as AI’s rapid development requires more power and cooling solutions. This surge poses a challenge for Amazon, which is committed to achieving net-zero carbon emissions by 2040.
Amazon Web Services (AWS), the world’s largest cloud provider by revenue, plans to begin piloting the AI-designed carbon removal material in one of its data centers starting in 2025. This initiative is part of a three-year collaboration with Orbital, which will also gain access to AWS’s technology and open-source AI tools for further development.
Howard Gefen, General Manager of AWS Energy & Utilities, stated that the partnership would promote sustainable innovation, but financial details remain undisclosed. Orbital, with offices in Princeton, New Jersey, and London, began its journey about a year ago by setting up a lab to synthesize AI-designed materials. The startup aims to work with AWS to test additional AI-generated solutions, addressing water usage and cooling requirements in data centers. Godwin co-founded Orbital, which currently employs 20 people and is supported by investors such as Radical Ventures and Nvidia’s venture arm. Before this, Godwin contributed to materials science work at Alphabet’s DeepMind until 2022.