You don't need Kafka in order to run Debezium. While that certainly is the most common deployment, you can use Debezium also with things like Kinesis, Google Cloud Pub/Sub, or even plain HTTP sinks, all via Debezium Server (https://debezium.io/documentation/reference/stable/operation...). You even can embed it as a library into your application for the highest degree of flexibility.

That said, streaming events into Kafka (or something like it), rather than pushing change events directly to destinations like Snowflake has big advantages when it comes to reprocessing data, setting up additional consumers, etc.