Building a Super Computer with Amazon HPC Clusters

Please what are some disadvantages of wanting to build a Super computer using Amazon High performance computing clusters?