Measuring costs with actual dollars rather than percentage signs is often the best way to judge the bottom line. Yes, relative infrastructure cost tend to go down at a larger scale - assuming that revenues increase in proportion to the infrastructure added, which may not necessarily be the case with Bing and advertising.

It is also important to acknowledge the opportunity costs for Microsoft that scaling Bing might entail, e.g. is infrastructure more profitably used for cloud services or search? Does an advertising revenue model interfere with the necessary trust which underpins Microsoft's business relationships with enterprise customers?

I am not saying it is unreasonable to compare Microsoft with Google at the corporate level - but using Bing and Google Search as the primary criterion for such a comparison is a category mistake.