Data storage service recommendation (cost vs query latency) when using dataset with many time series

[Floris Calkoen]

I'm designing a small Flask Google App web-application/dashboard which will show some evolution patterns of climate data at global level. 

The observations were made at approx 1.5 million sites. Each site has approx 50 observations. Hence, the dataset consists around 1.5 million time series with in total 75 million rows.

Unfortunately we don't have a large budget for this app. However, we don't expect that many users (few per day). Nevertheless query latency should be kept as low as possible. 

What would be the optimal data storage service considering both query latency and costs? Should I just use Cloud Bigtable? Or maybe Cloud SQL?