If you have the need to store 200 MB in Redis, you're charged $0.125 / GB-hour for that data, and the minimum metered data storage is 1 GB. So that's $3 per day or about $90 for any amount of data stored between 0 GB and 1 GB. And then you need to pay for compute on top of that. Alternative, you could get a whole cache.t4g.micro instance with 0.5 GB of memory for less than $12 per month.
Same for larger amounts of data. 10 GB of data storage (without any compute) would be $900 vs. a whole cache.m7g.2xlarge with 26.04 GB of memory is around $450 per month. Even accounting for replica instance, the only benefit you seem to be getting from the price premium is the scaling. And those 2xlarge instances probably still have plenty of headroom.
And for a "serverless" product, you can't really use it for infrequent usage either because of that crazy minimum storage metering. Imagine if your Lambda functions charged you $90 per month for just existing. "Pay-for-use billing model" where "you only pay for the data you store", and then some if you don't store enough.
> Minimum metered Data Storage: 1 GB per cache
That means each cache costs least $90/month.