Google Redefines AI with Carbon-Efficient Chips
Google steps into the limelight with a pioneering study highlighting the entire lifecycle emissions of its AI accelerator chips.
“The study found that innovation in our chip hardware design led to a 3x improvement in the carbon-efficiency of AI workloads over two generations and that decarbonising our electricity-related emissions will drive the biggest carbon reductions for our AI footprint,” explains Kate Brandt, Chief Sustainability Officer at Google.
“At Google, we know AI can drive transformative innovation in areas like information, optimisation and prediction.
“We also know it’s equally important to manage its environmental impacts, and we’re working to do that through efficient infrastructure, model optimisation and emissions reductions.
“This study is an important step in those efforts, unlocking critical insights for Google and others looking to reduce emissions across the full lifetime of AI hardware.”Adam Elman, Director of Sustainability EMEA at Google, says: “This is just the beginning with huge opportunities to continue optimising hardware and software for carbon efficiency.”
Focusing on the integration of AI in transformational areas like information, optimisation and prediction, Google also places a heavy emphasis on attenuating its environmental footprints.
Introducing Compute Carbon Intensity
A novel benchmark, Compute Carbon Intensity (CCI) has been crafted specifically for this study.
It stands as a promising tool for fostering transparency and kindling further innovation across the tech industry.
Delving into five Tensor Processing Unit (TPU) models, the study evaluates their entire lifecycle emissions while inspecting the influence of hardware design choices on carbon efficiency.
TPUs, which are specialised hardware accelerators, play a significant role in enhancing AI's environmental sustainability through heightened efficiency.
“Their efficiency impacts AI's environmental sustainability. This progress is due to more efficient hardware design, which means fewer carbon emissions for the same AI workload,” explains Robert Little, Sustainability Strategy Lead for gTech at Google.
CCI measures the carbon emissions of an AI accelerator chip per unit of computation, expressed in grams of CO₂ per Exa-FLOP.
A lower CCI score indicates reduced emissions from the hardware platform for any given AI workload, showcasing Google's commitment to increasing its TPUs' carbon efficiency.
Unpacking Google's research
Google's investigation uncovers a threefold enhancement in the CCI of its TPU chips across four years, spanning from TPU v4 to Trillium.
This significant progress suggests that opting for newer generations of GPUs can resultantly diminish carbon emissions for identical AI tasks.
It emerges that operational electricity emissions constitute more than 70% of the lifetime emissions of a Google TPU, underscoring the critical need to ameliorate both the energy efficiency of AI chips and the carbon intensity of the electricity that powers them.
Although manufacturing emissions also contribute to the total emissions, their proportion is set to increase as operational emissions are cut down.
Google harnesses detailed manufacturing Lifecycle Assessments (LCA) to hone its decarbonisation strategies in manufacturing, engaging actively with supply chain partners to curb these emissions as well.
“These findings highlight the importance of optimising both hardware and software for a sustainable AI future,” Robert says.
“It's important to remember where AI has important implications for reducing emissions and fostering sustainability.”
Make sure you check out the latest industry news and insights at ClimateTech Digital and be part of the conversation at our global conference series, Sustainability LIVE.
Discover all our upcoming events and secure your tickets today. Subscribe to the ClimateTech Digital newsletter.
ClimateTech Digital is a BizClik brand