It depends on what the improvement is and how much it costs to implement. The rumor (although likely false) suggests an improvement in yield, which is where there's a major difference between FF and APS-C. Sensors are cut from round silicon wafers, and according to Canon a single wafer can produce 20 FF sensors or ~200 APS-C sensors.
HMM! This is interesting. If Canon is producing ~200 APS-C sensors per wafer, and only 20 FF sensors per wafer, then that means they are already producing APS-C sensors on 300mm wafers, but are still producing FF sensors on 200mm wafers. If you run the numbers, the raw number of full APS-C sensors on a 200mm wafer is 94, on 300mm wafer is 212; the raw number for full Ff sensors on 200mm wafer is 36, on 300mm wafer is 81. Factor in losses, you get a bit less than 200 APS-C/300mm wafer, maybe 20 FF/200mm wafer. I suspect that the actual number of total FF sensors is less than 36, since every time I've seen a photo of large sensors on a wafer, there is usually plenty of blank space and an unetched border around the edge. So maybe Canon gets 190 APS-C out of a 300mm wafer, and indeed only about 20 FF out of a 200mm wafer. Assuming similar losses with larger wafers, Canon should get almost 70 FF sensors out of a 300mm wafer if they do indeed make the move.
How about a hypothetical example… Say it costs $20,000 for the raw silicon wafer and the stamping and cutting (I have no idea how wild-assed that guess is). That means a FF sensor costs $1000 and an APS-C sensor costs $100. Now, say there are on average two random local defects per wafer that result in the loss of the sensors where they occur. FF production takes a 10% hit on yield, whereas APS-C takes only a 1% hit on yield. Taking QC defects into account, the cost of a FF sensor is $1111 and an APS-C sensor is $101. Now, suppose this new process cuts the defect rate in half, to one per wafer, and increases production costs by 2% per wafer. That drops the cost of a FF sensor to $1074, a 3.3% savings. However, that 'improvement' results in an APS-C cost per sensor of $102.50, an increase of 1.5% per sensor for APS-C production.
Sell 5,000,000 FF cameras and save $37 each and that's a 185 million dollar profit….
Sell 100,000,000 APSC cameras and spend an extra $1.50 each and that's a 150 million dollar loss....
Now which pile of money do you think Canon would go for first?
Granted, this is only a hypothetical example. Hwever, it does demonstrate one scenario in which application of a process improvement for FF production would not be cost-effective when applied to APS-C production.
I totally agree. I think increasing yield on the FF sensor front is really where they can save the most money, especially if they are still using 200mm wafers. They have to waste a proportionally much larger area of a 200mm wafer than a 300mm wafer when fabbing FF sensors.