Oracle must hold on to the hardware division at all costs. The financial future of the computer industry is in hardware, not software. Software will be extremely cheap because necessity is about to unleash a revolution in software construction methodology that will turn every computer user into a programmer whether they know it or not. The future of profits in this industry is going to be strictly about who has the baddest, fastest and most energy-efficient parallel processors. The software will just sprout like mushrooms.
The painful (and scary to many) transition to parallel computing and the crisis that has ensued does not bode well for the status quo. Who would want to spend millions or billions converting legacy software into multi-threaded code only to find out afterwards that multithreading is not the part of the future of parallel computing. The baby boomer generation (the Turing Machine worshippers) whose bankrupt ideas on computing led to this cisis must be forcibly retired even if it creates an uproar. This will allow new minds and new ideas to flourish so that the industry can leap beyond last century's flawed paradigms and forge a new future.
Oracle has an unprecedented opportunity to make a killing by doing the right move. Sun's hardware engineers are a talented bunch and it would be a dumb idea to let them go. But if the sale goes through, I hope HP realizes the importance of hardware and immediately start dumping loads of cash into another big-chip parallel processing project (and please do not resurrect the Rock project).
Having said that, the solution to the parallel programming crisis that will revolutionize computer programming means building a new type of computer to support a radically different programming model. There is no escaping this. Read How to Solve the Parallel Programming Crisis for more on this topic.