The anger of the major OEM's would far outweigh any cost benefits.
If I were Dell, HP, etc, I would see the purchase as Microsoft becoming a serious hardware competitor and I'd start to rethink my Linux positioning.
The existing OEM partner agreements have provided a great lock-in for MS on the desktop. The millions MS pays to existing hardware partners is really just grease on the OEM skids for keeping their OS position in those accounts. They would be much better served to spend a 1 billion on R&D with those OEM's to develop Windows-centric server, blade and datacenter cooling and power saving reference designs and technologies than to try and out-cheap Google.
Being cheap is just not part of Microsoft's DNA. They prefer to over-resource (see the waste that became Vista), and spend wildly because that is their biggest advantage, cash. There were thousands of IT's best and brightest programmers responsible for what is Vista. For all of it's faults, overabundance is what is killing Microsoft's ability to compete in the market today. With the exception of OneNote, what software has MS released that hasn't been an huge, bloated, mess?
It's the software version of an all you can eat code buffet. A dash of this, a pound of that, want some silverlight for dessert? It's all on Daddy Gates!
Google was forced to bootstrap it's way to market and that's still an inherent quality in Google. The backend systems were designed to be cheap and scalable because that was the only way they could keep up with growth and keep the lights on. Forget the free soda in the breakroom and free dry-cleaning, it's nothing compared to what they saved by establishing their own datacenter hardware processes.
Building reference designs and seeding their hardware partners would provide a much better return than any initial cost savings from the purchase. Rackable has a great focus today, but imagine what damage Microsoft's huge resources would inflict on Rackable's DNA and focus over the long term. In 5 years they would be forced to buy inefficent, out of date technology from an in-house subsidy rather than select the best tool for the job at hand.