I've worked with somebody who came from Acxiom's grid computing division -- who has left after they were sold of to EMC. Essentially they had a fairly advanced grid computing infrastructure (comparable to Map/Reduce or other Scatter/Gather) for what seems like a while before Google has. Compared to what was then (2007) Hadoop's State of the Art (mind you, we were working together at Yahoo) it seemed quite impressive.

Now I wonder if the reason for their sale was prescience about open source grid processing -- which would have been against the grain at the time (this was before Hadoop was universally adopted even at Yahoo).

What's also interesting is that the sale indicated they also believed that they were to have no competitive advantage in further development/deployment/maintenance of data infrastructure. This is unlike the Internet firms I've worked at where even if the entire stack was open source, continued in-house development and operation of the stack was vital to the business. At first that surprised me, but then again compared to, e.g., Internet advertising data their data is never anonymous, much cleaner, and smaller in absolute volume (number web pages each unique client visits in a week -- which is a number greater than each unique individual -- visits vs. number of items each individual purchases from a store in a week).