Public data availability in cloud-hosted databases raises interests in systems providing multi-cloud querying capabilities. Since data access in this context induces monetary costs, we suggest a method to compute SLAs for multi-cloud queries. It consists in decomposing queries into a directed graph of maximal sub-queries per provider, and finding a financially cheapest execution plan. This method yields SLAs with monetary costs lower than a download-all approach, granted that inter-provider data transfers are minimised.