Economics of Wasm serverless hosting for teams

Comparison of Fermyon and Cloudflare Pricing

Fermyon Cloud provides a serverless environment for developers using the Spin framework. The Starter plan costs $0 per month and includes 100,000 requests, 5 custom domains, 5 subdomains, 5 KV data stores, 5 KV storage, 1GB of bandwidth, and 1 region deployment. Teams needing more capacity can use the Growth plan for $19.38 per month, which includes 1,000,000 requests, 100 custom domains, 100 subdomains, 100 KV data stores, 100 KV storage, 2GB of KV storage, 50GB of bandwidth, and 1 region deployment. Fermyon Wasm Functions remains in early access and requires an onboarding request. This service provides no service-level agreements because it is an open beta.

Cloudflare Workers provides a different scaling model for edge workloads. The Paid plan requires a minimum monthly charge of $5 per account and includes Workers, Pages Functions, Workers KV, Hyperdrive, and Durable Objects. Users in the Standard model pay $0.30 per million requests and $0.02 per million CPU milliseconds. Cloudflare also provides specialized storage. For Workers KV, the Paid plan costs $0.15 per million requests and $0.20 per GB of storage per month. D1 database usage costs $0.15 per million requests, $12.50 per million GB-s, $0.001 per million rows read, $1.00 per million rows written, and $0.20 per GB-month. R2 provides S3-compatible object storage at a cost of $0.05 per GB.

Latency and Resource Efficiency

Cloudflare Workers wins for high-volume, latency-sensitive Wasm workloads because the pricing avoids a separate duration meter for request-heavy tasks. The platform uses V8 isolates to multiplex thousands of tenants on a single machine. Because the Wasm runtime stays loaded in server memory waiting for requests, a new request arrives and the runtime instantiates the module instantly with a speed advantage that exceeds container-based solutions that require full boot sequences. Wasm modules initialize in under 1ms. This speed is a significant improvement over Lambda@Edge, which sees cold starts between 250 and 800ms.

Wasm modules consume significantly less memory than containers. A typical Rust-compiled Wasm microservice uses between 1 and 5MB of memory. A Docker image for the same functionality starts at 50MB and often exceeds 200MB. This allows a single 1GB node to host hundreds of Wasm pods, whereas the same node only hosts around twenty Alpine container pods. The stability of WASI Preview 2 in 2025 turned WebAssembly into a genuine server-side runtime. Developers use the Component Model to package Wasm code into modular, portable libraries. These libraries act like building blocks that connect across different languages.

The CNCF 2026 cloud-native survey shows that 31% of organizations are now evaluating or using WebAssembly for serverless workloads, up from 15% two years ago. While Docker remains the default for 85% of containerized workloads, Wasm fills the gap for sub-millisecond cold starts. Common runtimes include WasmTime, Wasmer, WAMR, and Lucet. WasmTime is the reference implementation from the Bytecode Alliance, which includes members like Fastly and Intel. WasmEdge serves AI and edge workloads and acts as the default runtime for Docker Desktop’s Wasm integration.

Cloudflare operates in 320+ cities. Fermyon uses Akamai’s 4,000+ global edge locations following its acquisition in December 2025.

Selection Logic for Edge Workloads

The verdict: Cloudflare Workers wins for high-volume, latency-sensitive Wasm workloads because the pricing avoids a separate duration meter for request-heavy tasks.

Cloudflare pairs Workers with specialized storage. Workers KV provides eventually consistent reads, while Durable Objects provides strong consistency. D1 provides a serverless SQL database. R2 provides S3-compatible object storage with no egress fees. You already know that cold starts kill user experience.

Fermyon works with the Spin framework. The Growth plan provides 1,000,000 requests and 50GB of bandwidth. Cloudflare handles high-scale requests through its Standard usage model. Cloudflare also supports native Wasm modules and a beta Python runtime. Developers use various languages including Rust, C/C++, Go, and AssemblyScript to build these modules.

Which team will find the most value in the early access onboarding process of Fermyon?