If money is no object
I'd go for an SMP AMD64 based system. The point of this is research to replace other systems... so get the system that looks like it will deliver most bang for the buck, and be the logical jumping off point for the future development you're talking about. The fact that the new kernels are going to start off a bit squiffy (malodorous, you might say;) is not the point... the point is that you are going to be trying to develop two main sets from the use of this system; a skill set and a tool set that your firm can use. The skill set is about being able to build the systems they want, and the tool set is the combination of hardware and software you're going to use to do it.
See if you can get them to spring for a two or four way, a boatload of RAM, and a nice RAID disk array, as well as the interfaces to go for massive cheap storage (eg- there's a new USB 2.0 box out that takes SATA and/or IDE drives, and interfaces with the PC via a USB cable). Then, you can get down and start messing around with the Right Way to tweak the bits... the kernel options, what's the good compiler, the good packages (postgres? Oracle?) for building the applications you need, and so on. Start making test applications, and running tests on them, for correctness, performance, and how long before it cracks under the load. Two years down the road... it may be time for a massive overhaul of your back room.
--\n-------------------------------------------------------------------\n* Jack Troughton jake at consultron.ca *\n* [link|http://consultron.ca|http://consultron.ca] [link|irc://irc.ecomstation.ca|irc://irc.ecomstation.ca] *\n* Kingston Ontario Canada [link|news://news.consultron.ca|news://news.consultron.ca] *\n-------------------------------------------------------------------