The increasing use of artificial intelligence and generative AI tools like ChatGPT helped stoke a roughly 30% increase in Microsoft’s carbon footprint from 2020 until last year.
It’s a troubling development for the tech giant, which set an ambitious target of making its operations carbon negative by the end of this decade, removing more carbon from the environment than it emits each year.
“AI has made the path to 2030 more challenging,” said Brandon Middaugh, senior director of the company’s $1 billion Climate Innovation Fund. “The hill has gotten steeper.”
At a recent event for climate entrepreneurs hosted at the Washington Clean Energy Testbeds at the University of Washington, Middaugh discussed the strategies being implemented by the Redmond, Wash.-based company to address its climate impacts.
That includes investments by her fund into startups developing new clean energy and climate technologies, internal policies that penalize actions that generate carbon emissions, public transparency in Microsoft’s carbon cutting efforts, and other initiatives.
At the same time, the company has a leading role in promoting the use of AI and is partnering with OpenAI, which released ChatGPT in November 2022. The technology is driving the construction of new data centers that gobble energy to run their computations and to cool their heat-generating servers. Microsoft spent a record $19 billion on capital expenditures in its most recent fiscal quarter.
Despite the worrying rise in emissions, Microsoft leaders remain committed to their climate goals. At the UW event in Seattle, Middaugh offered three reasons she’s hopeful the targets can still be reached.
It’s early days in AI
“We’re in a period of foundational investment in AI, literally building the infrastructure and also building out the actual technology itself. And what we’ve seen in the past is that as you bring these new digital technologies online … and you start at a certain point, and then the efficiency improvements can be really significant,” she said.
“I’m seeing a ton of innovation and a ton of opportunity in improving the efficiency,” Middaugh added. “And that can be everything from the compute itself to thermal efficiency of the entire system.”
AI can help optimize the power supply
“As the AI tools themselves become stronger and accurate and reliable, we’re going to see a lot of opportunity for them to do some of the optimization of existing [power] resource use, and then that will begin to balance the source and use dynamic,” she said.
AI can generate novel materials for climate tech
“This goes beyond the horizon of our Climate Innovation Fund and much more into the research space, [which] is generating and finding novel solutions entirely,” Middaugh said. The emerging AI tools are leading to discoveries of new materials and new composite materials in particular, she added, and “that’s where I’m really excited long-term about where the research is leading.”
Microsoft and the Department of Energy’s Pacific Northwest National Laboratory (PNNL), for example, reported earlier this year that they used AI to discover a new material that could make batteries more efficient. AI could also be used to develop materials for carbon removal and low-emissions building materials.