For comparison the PS3 uses a cell based processor with 1 PPE (Power processing element) @ 3.2 GHz which is dual thread i.e. capable of processing 2 streams of instructions at a time and 8 SPEs (Synergistic Processing Elements) @ 3.2GHz of which 1 is disabled for yield purposes and 1 is reserved for security purposes such as encryption and decryption of the hard drive in real time. The cell in the PS3 has an overall performance of 204 GFLOPs.
Sony had invested hundreds of hundreds of millions of dollars in to the development of the Cell Processor and on building fabrication factories (of which cost over $500,000,000 each!), all that money would go to waste if Sony chooses not to include a Cell processor in their next console.
I believe that Sony will add more SPEs to have an overall amount of 16 SPEs as this has an optimal power to heat ratio if manufactured at 32nm. It should consume the same amount of energy as the 90nm Cell processor found in the 60/20Gb PS3s. 1 will be disabled for yield purposes and 1 will be reserved for security purposes. If this cell processor is made then it should start at 32nm as Sony are already at work shrinking the PS3′s cell to 32nm. At 32nm they could double the local store for each SPE to 512KiB which will add much more performance for games.
Clock speed could increase to up to 3.8GHz, however with the improved performance of the new cell and the amount of SPEs this is unneeded and will create too much heat. I can only see an increase in clock speed if fabrication starts at 22nm as this will help reduce the heat created. 22nm processors are expected to be released in 2011 so by the time the PS4 comes out this would be the norm, so 22nm is a feasible starting fabrication. This will make it have an overall performance of around 500 GFLOPs for the whole Cell Processor at 3.2GHz.