Don Scansen published his speculations on Aptina's past and future in his SemiSerious blog. His main conclusion is that by creating Aptina Micron is thinking less about how to run an imaging business than how to sell it.
The problem with Aptina is that it has lost its time and technology lead. I was dissapointed to see that no EDoF products were announced yet, while many competitors are samlping them. 1.4um generation is to sample by this summer, simultanously with its major competitors. So Aptina is slowly tranfroming from the industry leader to "me too" vendor competing on price. In this condition its best move would be to partner with some very cheap foundry. I'm happy to hear that Aptina is free to choose such a partner, as this is the key to its future success.
Micron's success has always hinged on making wafers at a lower cost than (most) of their competitors. Seven or so years ago they were definitely less expensive than the cheapest cheap foundry. What makes you think that is not still true today?
I think as a spin off they will get more fab focus and fewer prioritization conflicts.
Again, lean and mean is what they need to focus on. Some of the upper managers know that already from PB days. Some come from big company environments with deep pockets. Hope they evolve in time...
I'm not sure what makes Aptina lose money these days. Probably this is a compbination of few things.
I believe Micron's wafer price is still very competitive, but process R&D might be too expensive. Some foundries manage to develop process in a more cost efficient manner, may be because labor cost on Far East is lower.
Most Aptina's imagers have larger dies than its competitors. This partially offsets the lower wafer cost.
For sensors with integrated ISPs, DRAM process is less efficient than CMOS, in logic portion anyway. So the die becomes even bigger.
I see no other explanation of why Aptina is not profitable, despite the low production cost. So, the natural solution is to seek for even lower wafer cost and remove process development expenses from the balance sheet.
If Aptina wanted to be a "me-too" sensor vendor, using the same foundry process as everyone else is a easy way to make that happen.
Reduction of R&D expenses on fab development is otherwise a good idea but that reduction might be easier accomplished by (1) applying more engineering to process development and running and testing fewer wafers, and (2) more careful selection of product opportunities. I also suspect there is a lot of engineering dead wood at Aptina being carried along out of loyalty.
There is no "the same foundry process as everyone else" is using. Everyone is modifying it in one way or another. Even the smallest players on the market do this.
And it's very hard to apply "more engineering to process development" to run less wafers. The problem is that most key pixel parameters can not be simulated, even though TCAD companies claim the contrary. So most designers try to build empirical models and they need wafer data to do this.
In fact I was surprised to read that Omnivision managed to complete its 2nd generation 1.75um pixel in only 100 pixel and 100 process iterations. This is a truly remarkable result.
To get back to Aptina, it needs to innovate to really succeed. I dont think that doing "more of the same" just leaner and meaner can keep it in leaders. Talking about innovation, it can come from technology or business side. For instance, 1.1um pixel sampling this summer could be great. Or partnership with ProMOS, Winbond, SMIC or Hynix (all having cheap production, CIS ambitions and zero pixel knowledge) could bring a fresh stream to the market.
I don't think it was a conclusion, I think it was just wondering out loud.
ReplyDeleteLean, mean, fighting machine and cut the apron strings with Mother Micron. That is what they need to do now.
-EF
The problem with Aptina is that it has lost its time and technology lead. I was dissapointed to see that no EDoF products were announced yet, while many competitors are samlping them. 1.4um generation is to sample by this summer, simultanously with its major competitors.
ReplyDeleteSo Aptina is slowly tranfroming from the industry leader to "me too" vendor competing on price. In this condition its best move would be to partner with some very cheap foundry. I'm happy to hear that Aptina is free to choose such a partner, as this is the key to its future success.
Micron's success has always hinged on making wafers at a lower cost than (most) of their competitors. Seven or so years ago they were definitely less expensive than the cheapest cheap foundry. What makes you think that is not still true today?
ReplyDeleteI think as a spin off they will get more fab focus and fewer prioritization conflicts.
Again, lean and mean is what they need to focus on. Some of the upper managers know that already from PB days. Some come from big company environments with deep pockets. Hope they evolve in time...
-EF
I'm not sure what makes Aptina lose money these days. Probably this is a compbination of few things.
ReplyDeleteI believe Micron's wafer price is still very competitive, but process R&D might be too expensive. Some foundries manage to develop process in a more cost efficient manner, may be because labor cost on Far East is lower.
Most Aptina's imagers have larger dies than its competitors. This partially offsets the lower wafer cost.
For sensors with integrated ISPs, DRAM process is less efficient than CMOS, in logic portion anyway. So the die becomes even bigger.
I see no other explanation of why Aptina is not profitable, despite the low production cost. So, the natural solution is to seek for even lower wafer cost and remove process development expenses from the balance sheet.
If Aptina wanted to be a "me-too" sensor vendor, using the same foundry process as everyone else is a easy way to make that happen.
ReplyDeleteReduction of R&D expenses on fab development is otherwise a good idea but that reduction might be easier accomplished by (1) applying more engineering to process development and running and testing fewer wafers, and (2) more careful selection of product opportunities. I also suspect there is a lot of engineering dead wood at Aptina being carried along out of loyalty.
Well, what do I know?
Lean and mean. Godspeed Aptina.
-EF
There is no "the same foundry process as everyone else" is using. Everyone is modifying it in one way or another. Even the smallest players on the market do this.
ReplyDeleteAnd it's very hard to apply "more engineering to process development" to run less wafers. The problem is that most key pixel parameters can not be simulated, even though TCAD companies claim the contrary. So most designers try to build empirical models and they need wafer data to do this.
In fact I was surprised to read that Omnivision managed to complete its 2nd generation 1.75um pixel in only 100 pixel and 100 process iterations. This is a truly remarkable result.
To get back to Aptina, it needs to innovate to really succeed. I dont think that doing "more of the same" just leaner and meaner can keep it in leaders. Talking about innovation, it can come from technology or business side. For instance, 1.1um pixel sampling this summer could be great. Or partnership with ProMOS, Winbond, SMIC or Hynix (all having cheap production, CIS ambitions and zero pixel knowledge) could bring a fresh stream to the market.