Google has - and continues - to do some amazing things. It's easy to be awed by their products and their infrastructure, but this one was even more interesting - they designed their own database to host all that information. I'm a product centric professional at this point in my life using SQL Server exclusively, and only tend to glance at other products (sometimes with admiration, contempt, and/or jealousy!) in the database space. To me it's impressive that they were willing to devote resources to writing a product in a space that is reasonably commoditized. Perhaps cost was part of it, not wanting to license 10k SQL instances (would Ballmer have cut them a deal to get the business?!), but it looks like they did it because they could see that nothing on the market really suited their very specialized needs. Read the doc to see some of the very gritty details.

Buy or build is still one of the hardest and riskiest decisions businesses make. Looks like they made a good decision, would we have had the nerve to do the same, or we would have stuck with the product we knew and just tried to make it fit (and would it fit?)? Good questions to ponder to make sure we don't fall into the all we have is a hammer all the data is a nail mentality!