As AI creates opportunities for business growth and societal benefits, we’re working to reduce their carbon intensity through efforts like optimizing software, improving hardware efficiency, and supporting our operations with carbon-free energy.
At Google, we’re committed to understanding the entirety of our environmental impact so we can apply the best, boldest, and most holistic solutions. In this post, we’ll talk through an assessment technique called Life Cycle Assessment (LCA) to understand the complete picture of carbon emissions.
Measuring environmental impact with Life Cycle Assessment
LCA is a process-analysis method for evaluating the environmental impact of a product-system or service throughout its entire life cycle. This includes everything from raw material extraction and processing, manufacturing, transportation, use, and end-of-life treatment (recycling, disposal, etc.). LCA enables us to measure emissions along every step of our hardware manufacturing, find the sources of those emissions, identify ways to reduce them, and track our progress towards global net-zero emissions.
The Google Cloud Carbon Footprinting team has developed a best-in-class LCA approach to evaluate the embodied carbon emissions associated with the supply chain of our data center hardware1, including AI/ML accelerators, compute machines, storage platforms, and networking equipment.
Figure 1. LCA stages and system boundary
The approach is consistent with global LCA standards, ISO 14040/14044, and is specifically tailored to Google Cloud’s data center technology portfolio and underlying manufacturing production processes. In addition, Google Cloud’s LCA methodology has been critically reviewed by Fraunhofer IZM, ensuring completeness, accuracy, and adherence to industry standards. This enables Google to accurately account for emissions that come from the manufacturing of various types of data center hardware, all the way down to the smallest components that compose the fleet.
- aside_block
- <ListValue: [StructValue([(‘title’, ‘Try Google Cloud for free’), (‘body’, <wagtail.rich_text.RichText object at 0x3e0ed9683a00>), (‘btn_text’, ‘Get started for free’), (‘href’, ‘https://console.cloud.google.com/freetrial?redirectPath=/welcome’), (‘image’, None)])]>
Driving the industry forward
By collaborating closely with our supply chain partners, academic leaders, and industry peers, we’re pioneering the development of highly configurable Life Cycle Inventory (LCI) models. This innovative approach empowers us to move beyond generic assessments, unlocking the potential for detailed, customized environmental insights for vital components like semiconductors, hard disk drives, PCBAs, and thermal management solutions.
To achieve unparalleled accuracy, Google Cloud is transforming LCA data collection by partnering directly with suppliers to gather primary data. This means capturing the direct flows (i.e., material and energy transactions with the natural environment) that occur throughout manufacturing. These custom LCIs are powerful tools, enabling us to precisely measure our environmental impact and accelerate our journey towards net-zero.
Figure 2. Process-level environmental transactions
In addition to driving accuracy, Google is driving standardization in the hardware industry by participating in a collaborative effort to develop consistent LCA guidelines. This initiative aims to create Product Category Rules (PCRs) that facilitate primary data collection and improve comparability across product assessments. By building on established ISO standards and aligning with GHG protocol and Product Environmental Footprint (PEF), this collaboration seeks to enhance the accuracy and transparency of environmental accounting efforts.
In a recent LCA study, we evaluated the environmental impact of our Tensor Processing Units (TPUs) throughout their entire lifespan. The introduction of a new metric, Compute Carbon Intensity (CCI), helped uncover findings showing that over two generations, more efficient TPU hardware design has led to a 3x improvement in the carbon-efficiency of AI workload. LCA studies like this are crucial for understanding and reducing the carbon footprint of hardware across the ecosystem.
Advancements in LCA
At Google, we believe that informed action is essential, and that requires a foundation of accurate measurement. Through our advancements in LCA and by fostering collaboration within the global community, we’re driving meaningful, measurable progress towards a more resilient future.
To learn more, visit these resources:
1. Upstream supply chain activities are also defined as cradle-to-gate or Scope 3