It's hardly that surprising. Clouderaworks has always been a significant player in the NoSQL space - historically something like 20-30% of the respective companies' revenues were traced back to HBase and adjacent gubbins (Solr, Phoenix etc.), and usually those NoSQL-centric use cases sit adjacent to more traditional analytical applications driving further linked revenue.
It would have been a much bigger surprise if they didn't launch a NoSQL cloud product, just like it'll be a surprise if they don't launch some form of data engineering-focused product in the near future. Likewise don't let your jaw hit the floor if there's a search or a graph-centric offering. They already sell and support those things extensively on premises and if they don't sell those things on the cloud then their whole value proposition - end to end data lifecycle - falls down.
The question is whether the company can do them _well_ on the cloud. Product quality expectations are radically different when you're buying something billed by the minute and sold at the push of the button. So while being able to do the same stack on your boring old tin as you do your on prem k8s as you do your aws or your Azure or your GCP has a certain value to their customer base, the actual end users expect this stuff to work easily and with minimal fuss.
Engineering "minimal fuss" onto any tech stack that traces its heritage back to the good old days of unsexy on premises Java-based data management tooling is going to be a tricky task.