Performance will win in the end!
There are inherent limitations to the terminal's performance. Whether laptop, tablet, or smart phone, or even PC, there is a natural limit to performance in terms of processing speed and storage capacity. There are also functional limitations such as the ability to guarantee the integrity and safety of your data (i.e. backups).
The cloud, on the other hand, has virtually no limit (except that of the host's commitment to bandwidth, response time, and storage capacity). In particular, response time, which includes not merely bandwidth capacity but processing capacity (i.e. multiple processor arrays). The local hardware (tablet, smartphone, etc.) will be more concerned with the presentation level, behind that the processing capacity of the cloud hardware will dominate as applications increase in virtual intelligence (complexity). This will ultimately tilt the balance towards the data-centric and process-centric aspects of the cloud-centric model as opposed to the terminal-centric (i.e. tablet-centric, etc.) model.
The economics of increased bandwidth will be driven by the necessity of performance and safety that only the cloud-centric model will bring. In this regard Google is on the right track. Steve Jobs statement that: "We are going to demote the PC to just be a device." seems to indicate that Apple will move slowly in this direction as bandwidth permits.
Google's scramble towards local apps (i.e. tablet-centric, etc.) is a recognition that it will take time for the bandwidth capacity and cloud processing capacity to reach the level required for what will ultimately be a cloud-centric model.
That said, there will always be a need for client software, but the dominance of what might be viewed as a server-centric (cloud-centric) model will ultimately prevail.
