Velocity by Booz Allen

The IT systems we design today that use cloud and AI technology will be the subject of regulator assessments tomorrow. Organizations can prepare for the requirement to report on Scope 3 carbon emissions by gaining visibility into cloud carbon metrics reporting from the data generators. Due to the shared responsibility model between cloud providers and cloud consumers, both sides need to work together to tackle the cloud carbon data visibility challenges. Cloud providers can evolve existing carbon footprint tools to support real-time reporting services with APIs; publicly available and repeatable emissions calculations methodologies; and granular tracking of carbon emissions per service, region, and workload with the ability to tag resources based on customer-defined groupings across projects and business units. Consumers of cloud computing, especially systems integrators for federal entities, can relay carbon reporting requirements to CSPs for feature enhancements and embed sustainability metrics, thresholds, and goals into cloud and AI algorithm design. Any investment in education, strategy, and implementation of sustainable cloud and AI practices can lay the foundation for a greener future and avoid costly rework of applications, cloud infrastructure, and AI models to comply with federal GHG reporting requirements. What Can Organizations Do Today? Curbing cloud and AI carbon emissions begins with awareness of the connection between them and understanding every person’s role in the solution—from enterprise leadership to the end user. Based on our on-the- ground experience navigating enterprise sustainability, here are specific areas of focus for key roles that can start to make an impact today: • Senior Executive and Agency Leader. Identify sustainable cloud and AI as an organizational priority, incorporate objectives into the

corporate experience in solicitations.

Auditor. Require clear, repeatable, and publicly available methodologies for carbon emissions calculations from federal and commercial organizations. Project Manager. Identify areas of waste and excess in cloud spend due to inefficient resource usage (e.g., leaving lower environment servers running 24/7) and work with engineering/development teams to optimize resource usage to reduce cost and carbon emissions. Developer and Engineer. Design carbon-aware systems with published thresholds, metrics, and dashboards; write energy-efficient code using software carbon intensity measuring tools; and optimize cloud and AI compute resources to reduce energy consumption.

Ultimately, we all have a role in achieving cloud and AI sustainability in the actions we take, in big and small ways. For example, individuals can reduce the amount of data stored in the cloud by deleting emails, pictures, and applications and avoiding the increase of additional cloud storage space and costs. The difference between achieving the promise and realizing the downstream impacts of cloud computing and AI lies in our action or inaction for advancing green IT. Awareness of cloud carbon emissions and AI’s potential to be both the greatest cause and most efficient solution for the climate crisis offers a starting place for embedding sustainability practices within every organization. As the effects of climate change intensify, so too does our individual and collective responsibility to curb cloud carbon emissions in order to build a greener future at home, in cyberspace, and at the tactical edge.

energy industries and jobs through Federal sustainability” and partnering with carbon-conscious organizations to accelerate progress. While major cloud service providers (CSPs) offer some form of carbon footprint tracking tools for cloud-hosted workloads, the reporting does not currently include near real-time data with granularity at the project or workload level that could inform the C-suite, project manager, engineers, and regulators of how best to use and report this data. Some CSPs have roadmaps that include new features to approach more granular reporting of carbon emissions. However, this only includes the compute resources that power AI and not the software and algorithms themselves. In research and academia, most scholarly publications include information about an AI model’s accuracy, number of parameters, and duration of time to run, but few include efficiency metrics such as carbon emissions, training costs, or model accelerators. We have reached the frontier of carbon emissions tracking, and further progress requires a coalition of the willing and a commitment to environmental, social, and governance (ESG) causes to create the measures by which regulators and the public should assess our carbon footprint. Technology titans, technologists, and trainees have access to tools that can begin to close the chasm between what metrics we see and what metrics we should be tracking for cloud and AI carbon emissions reporting. The Green Software Foundation, an organization committed to carbon reduction in software development, created the Software Carbon Intensity (SCI) Specification, which describes how to calculate

the carbon intensity of a software application. The SCI GitHub page provides a methodology for “calculating the SCI score for any software application, from a large, distributed cloud system to a small, monolithic open-source library, any on-premises application, or even a serverless function.” Climatiq’s REST API supports organizations’ GHG data in technology products with real-time emissions calculations. In addition to climate-focused tools, most CSPs offer cloud optimization services that use automation to identify ways to improve cloud cost and usage efficiency and well-architected frameworks that include sustainability and/or performance optimization in their design principles. While no one tool singlehandedly resolves the transparency problem of measuring AI carbon emissions, education about these tools and methodologies can empower people to reduce carbon emissions during the design, development, and deployment of AI algorithms. Imminent changes to federal regulations for carbon emissions reporting have also upped the ante for action. In November 2022, the Department of Defense, General Services Administration, and NASA submitted a proposal to amend the Federal Acquisition Regulation to ensure major federal contractors disclose their GHG emissions and climate- related financial risks and set science-based targets to reduce GHG emissions. Earlier the same year, the Securities and Exchange Commission released a draft rule requiring public organizations to disclose Scope 1, Scope 2, and Scope 3 GHG emissions. While disclosure requirements are still evolving, the message is clear: Upcoming changes to policy aim to ensure federal suppliers make required disclosures and set targets to reduce GHG emissions.

Brianna Hogan focuses on emerging cloud capabilities and development in Booz Allen’s BrightLabs incubator, an experimentation organization designed to develop, test, and incubate mission-centric solutions rooted in emerging technology.

organization’s strategic roadmap, and design science-based targets looking toward federal regulation. Procurement Official. Include strategies, methods, and targets for carbon emissions reductions in requests for information (RFIs) and requirements for


While AI fuels demand for cloud computing to train and execute the models, AI offers solutions for curbing carbon emissions through standard machine learning (ML) techniques that improve energy efficiency.

Right-sizing ML models for a given purpose saves considerable amounts of energy by reserving computationally intense models for consequential efforts and less computationally intense models for mundane tasks.

Navigating enterprise sustainability begins with awareness of the connection between the cloud and carbon emissions and understanding every person’s role in the solution—from enterprise leadership to end user.




Powered by