Awash with appliances
Teradata have succeeded despite an apparently major obstacle: the highly proprietary nature of their offering. Large companies CIO departments generally loathe proprietary infrastructure, especially when they have just spent years trying to (just about) standardize on a particular database or hardware platform, so it is an uphill struggle for the appliance vendors. Red Brick briefly did well selling a database tuned for data warehouse applications, but eventually it could not shake off the idea that Oracle or IBM could just add a "star join" feature to their products and make it redundant. Hence it is to Teradata's credit that they have maintained clear blue water between themselves and Oracle/IBM/Microsoft at the high end of large data warehouses. This in turn has created a market large enough to attract new entrants such as Netezza and DATAllegro, who can offer an easy to understand "like Teradata, but cheaper" message to customers who have giant transaction datasets to analyze but balk at Teradata's high price tag and opaque pricing when it comes to maintenance payments. It will be very interesting to see whether IT departments will pass a blind eye over the proprietary nature of these offerings (after all, this objection was essentially what killed off object databases) in the way they have with Teradata, though rumor has it that Netezza at least is making good early progress.
Of course only a small subset of data warehouses have the kind of volumes and processing requirements that require such technology. A TDWI survey showed Teradata at just 3% market penetration of deployed data warehouse databases, but of course this is a very attractive 3%, with typical deals in the million dollar range. Teradata has managed to overcome the proprietary stigma that bedeviled object databases in the 1990s and carved out an attractive high end niche that Oracle etc seem unable to really compete with. Its challenge now is growth, with competitors like Netezza nibbling into its margins and general purpose databases that get more powerful with each release. However the boom in raw data e.g. RFID seems likely to mean there is plenty of demand yet for raw power.