One of the primary drawbacks to using virtualization for server consolidation is that it dramatically reduces the number of points of failure in the datacenter, so that security breach or a hardware failure on a single machine can expose multiple (virtual) servers to compromise or malfunction. The industry is acutely aware of this issue, as evidenced by the growing number products aim at improving fault tolerance for virtual servers.
IBM, the company that pioneered the concept of virtualization with its mainframe systems, is tackling the security issue with Project PHANTOM, an initiative that's so secret that IBM won't even say what the name means. Still, that hasn't stopped them from putting out a press release on it, touting it as a "research breakthrough."
Update
I had a quick call with IBM yesterday and got some detail on PHANTOM. The company was still not willing to talk in any detail about it, but I did learn some important information that answers the questions I raised in my original post, which I've included below in its own section.
For starters, PHANTOM is not one particular technology, but rather a widespread research initiative within IBM that will eventually result in a range of products, services, best practices whitepapers, etc.. The initiative was started two years ago as a collaboration among various hardware and software groups within IBM, and has since expanded to embrace some third parties whose identities IBM isn't revealing just yet. The internal groups involved in the initiative include IBM's X-Force Threat Analysis Service (a division of IBM's Internet Security Systems), IBM Watson research center, and the server platform groups behind the z- and p-series servers, among others.
IBM stressed to me that the initiative will produce results for a wide variety of hardware/software combinations, including x86 systems, Windows, Linux, POWER, and others. So the scope of PHANTOM, broadly defined, includes all virtualization platforms, products, and services.