Vendor lock-in in serverless isn't about a physical contract. It's the *functional* and *architectural* cost of switching providers after you've built something substantial.
Think of it like building a complex marketing automation workflow in HubSpot. You've used their specific visual workflow builder, their proprietary "HubDB," their unique email templating language, and their built-in lead scoring logic. The *cost* isn't just the subscription fee; it's the immense effort to rebuild that exact business logic elsewhere. If you wanted to move to another platform, you couldn't just "lift and shift" your workflows—you'd have to manually reconstruct everything, likely with different tools and APIs, and retrain your team.
With serverless, the lock-in comes from three main areas:
* **Proprietary Event Sources & Triggers:** AWS Lambda is deeply wired to S3 events, SQS queues, and DynamoDB streams. Azure Functions are optimized for Cosmos DB changes. Replicating that event mesh on another cloud means re-engineering the entire plumbing.
* **Managed Services with Unique APIs:** Using AWS's Aurora Serverless, Cognito, or Step Functions means your application logic is written to their specific APIs and behavioral quirks. These aren't portable.
* **Tooling & Observability:** Your monitoring, deployment scripts (SAM, Serverless Framework provider plugins), and debugging practices become tailored to one ecosystem.
The trade-off is clear. You gain incredible development velocity and managed infrastructure by committing to a vendor's ecosystem. The "lock-in" is the price of that premium. Whether it's "worth it" depends on your tolerance for future migration cost versus your need for present-day speed. In martech terms, it's akin to choosing a monolithic, all-in-one suite (like HubSpot) versus a best-of-breed stack you integrate yourself. The suite has inherent lock-in but can be faster to market; the integrated stack offers flexibility but demands more ongoing integration work.
Show me the data
Exactly. It's that second point about managed services that hits hardest on the spreadsheet. The pricing and scaling behavior of, say, Step Functions vs. Azure Durable Functions vs. a self-managed Temporal cluster are so wildly different.
Your migration isn't just a port. You're doing a full cost re-model and performance re-benchmark. The business logic might translate, but the monthly bill and scaling limits become entirely new variables you have to solve for.