An application’s software carbon intensity (SCI) is the sum of its operational emissions and embodied hardware emissions. Serverless, or functions as a service (FaaS), provide a path towards reducing operational emissions by running event-driven applications only as needed. However, many serverless applications are limited by the underlying unit of compute. For example, containers provide isolation for applications in a multitenant cloud, but they take seconds to start. Similarly, MicroVMs, while more lightweight than traditional virtual machines, can still have slower startup times compared to containers. WebAssembly (Wasm) may solve this cold-start issue while reducing an app’s SCI. Sohan Maheshwar and Kate Goldenring discuss why Wasm is the greenest unit of compute for serverless applications and how you can get started using serverless Wasm with Spin, an open source developer tool. They’ll also demo running AI-inferencing Spin applications and discuss how serverless Wasm also enables better GPU resource sharing. By the end of the talk, you’ll have an understanding of the characteristics that make Wasm a cost-effective and sustainable unit for a greener cloud.