A new paper on the arXiv today. In this work we describe how to make a scalable quantum computer using atomic ensembles - small gas samples, which can be manipulated using laser light. These ensembles make very attractive qubits since you don't need to be able to address and manipulate single atoms. Instead you have a sample containing perhaps billions of atoms which can be collectively manipulated. They have very long decoherence times. Several experimental groups around the world have already demonstrated simple quantum information processing applications using atomic ensembles. However, none of the experiments to date are scalable, i.e. they cannot, even in principle, be scaled up to make a large quantum computer. In this paper we show to achieve scalability. All of the building blocks in our protocol are experimentally viable using present day technology, so hopefully at some stage in the near future someone will actually be able to demonstrate the elementary components of our protocol.