This isn’t a question to answer in few sentences. It depend how much data you’ll load to Power BI, should it be basically transferred and visualized? How much data you load in every cycle how much data you’ll store in all? There are many questions to answer before giving a architectural recommendation.
Basic approach could be to load data directly to the Power BI service. Or use Azure Data Factory or Azure Synapse Studio (Currently in preview) to ELT/ETL the data to another Data store. If you need the DWH features of Synapse use them.
Sorry but with the given information I can’t give a better recommendation. May the first approach sufficient? Go for it!