It shouldn't really be does as a black-ops project but it might have to start as such (cf. how YouTube convinced the internet to stop using Internet Explorer 6) with a proof of concept. I suspect that the vast majority of any company's databases does nothing special and could migrated to another system with a dump/load. The rest will, of course, take research, planning and costing but the savings can quickly become substantial once you stop having to buy licences for everyone in the company on all their devices.
If Oracle is required for "mission critical" stuff then work around it. Once you can demonstrate functional equivalance without the licensing costs for some projects, it should be possible to draw up a high-level analysis that quotes Amazon as having done the same thing already. Going to the "cloud" will probably be mentioned and any references to this new form of lock-in are likely to fall on deaf ears, because "cloud" sound a lot like "lower headcount" to a PHB.
Note, as soon as Oracle finds out that something like this is in the offing they'll launch the sales droids with FUD bombs to try and frighten people off, so you must have functional equivalence for backup/restore, etc. But, also, once you've moved one proper application you can also start asking for rebates.
At the end of the day, there should be nothing wrong in keeping some stuff in Oracle if it really is the best tool for the job. The problem is that Oracle needs to learn that that is what they have to provide and not Faustian contracts.
Biting the hand that feeds IT
