Google Cloud Pub/Sub

Kafka looks like and ordered queue, there no deliver backoff, so if a message has a problem, it doesn't advance to the next one. Google Cloud Pub/Sub looks like more a SET of messages, and kafka like a LIST. In kafka a same message will repeat instantaneously while it is being NACKED, on the other hand Google Cloud Pub/Sub will just deliver another and apply a backoff to the previous NACKED one, never stopping on a single message forever. Dead Letter Queues are innate to Google Cloud Pub/Sub while in Kafka is not a feature at all. One can configure the maximum amount of NACKs before sending a message to the DLQ, this is very powerful if combined with exponential backoff. Google Cloud Pub/Sub lets users scale consumers at will with no constraint, Kafka has a rather convoluted concept of partitions and consumer groups that requires one to plan ahead how many consumers they would like to plug to the queue because these constraints prevent a free scaling of consumers. If you plan for N consumers, you have to probably stick to that N forever or recreate your Kafka topic, in Confluent Cloud this is somewhat solved, but looks like a solution to a problem that should even exist in the first place.

Having used Amazon Web Services SNS & SQS I can say that even if the latter may offer more features, Google Cloud Pub/Sub is easier to use. On the other hand, usage of SNS & SQS as well as documentation and troubleshooting is easier with the AWS solution.
Since we are not using GCP only for Pub/Sub the choice depends on other variables.

Google Cloud Pub/Sub as a Managed service is significantly more easy to use than a self Managed Kafka cluster. As our software was already on GCP it was a no-brainer to use Pub/Sub due to the high level of integration and ease of use with other Google Cloud Platform services.

Compute Engine is not a direct competitor, but in fact, works well coupled with Pub/Sub.

Amazon SQS has no integration with google play billing whereas Google Cloud Pub/Sub had native integration with google cloud billing, Play billing.

We considered several messaging platforms including Kafka and Kinesis but both would have required more developer work and didn't integrate as nicely with our ecosystem. RabbitMQ is another messaging platform I've researched and prototyped on; it also would have required more developer time to implement as well but might be a good alternative for a hobbyist or a team looking for more customization.