Yes and no.
What they're describing what I'd describe as an OLAP 2.0. They're taking similar capabilities (central data store, cubed data) and combining them with user generated content, sharing and the cloud.
The system looks extremely similar to an BI system.
I'd make an counter point to TFA: I actually think that this is probablly Business Objects / Microstrategy / Cognos's biggest dream: the system shows the power that effectively BI can provide an business with data which is effectively shared and public.
Google are making their business case: give vendor lots-of-money and they can gain the capability over your own data, but in an nicely managable manner (so your competitors won't be getting access to it).
It's not even OLAP 0.5.
Fusion Tables is to OLAP what Dreamweaver is to Typepad. It's a very elementary storage capability that demonstrates Google's ability to abstract what they do on the back end to 'tables'. It is so far from an OLAP or BI system as to be a joke. Oracle and Microsoft have nothing to fear just like Bloomberg has nothing to fear from Google Finance. There are three reasons.
1. It's not OLAP. As a very elementary and basic thing, you'd have to be able to do operations in an abstracted, dimensionally aware language across multiple entities. You should be able to say 'reduce all of my global warming statistics by 5%'. Fusion Tables doesn't come close to being able to do that, much less handle conditional logic.
2. There's no migration facility. Upload a spreadsheet? You couldn't even get a business that runs Quickbooks to upload their records sensibly into Fusion Tables, let alone an enterprise.
3. Everybody who actually does BI for a living is not impressed. There's no *reason* to move because this offers nothing *new*. There are fundamental reasons why good BI is hard. When open-source BI vendors like Pentaho start saying - hey we quit, then that's when it's time to pay attention.
Google would be wise to put in some facility to integrate these objects with the blogosphere, something nobody has yet done. When these little tables are containerized such that they can be embedded like YouTube videos, complete with visualization, that will be a success. Get it as good as a generalized Gapminder (pitifully slow at gapminder.com) and then we can talk; it will be more like OLAP but it still won't be real OLAP, much less enterprise OLAP.
--
I think there are some fundamental problems that Google has anyway, with regards to the size of data that works in parallel across their storage infrastructure that is going to screw up their ability to manage the nitpicky drips of data that matter in datasets of OLAPable interest.