Note: A reticle is a fancy name for the photomask used in the advanced form of photolithography found in deep-submicron chip plants. There is no conceivable reason you would need to know this, but “rockin’ the reticle” leads to better headlines than most alternatives.
In these lean times of $10 billion semiconductor manufacturing plants and bankrupt chip designers, you have to go to great lengths to find good friends. And now that the smallest feature size on a processor chip has plummeted well below 100 nanometers (or a tenth of a micron, hence the term “deep submicron”), there’s just not the standardization among chip wafer fabs there once was. Used to be, Intel and AMD and Chartered Semi and Samsung might all move from 100 nanometers to 60 nanometers at the same time. Now, the few Asian giants that dominate production have separate paths through 45 nm, 40 nm, moving down to the magic number of 28 nanometers, which represents a temporary brick wall for the densest chips made.
I mention all this to say that interesting coalitions are budding around that magical 28 number, and the coalitions met the smartbook players at the end of February, when ARM announced that it would team with Global Foundries to produce systems on a chip, or SoCs, that combined ARM Cortex A9 with useful peripherals that might be found in smartbook designs. Global Foundries is a new joint venture formed by AMD with financing from an Abi Dhabi investment firm.
But wait, pundits were saying – doesn’t Qualcomm already have a deal with Global Foundries for access to that same process technology? ARM may not see its primary business as being the volume production of packaged chips, but don’t the two Global Foundries pacts set up competing interests? Given ARM’s business structure, the short answer would be “Not really,” but in any event, Qualcomm and ARM would not be entering new territory. In fact, during the month of February, two aggressive direct competitors – Xilinx and Altera, producers of a programmable chip called a Field Programmable Gate Array or FPGA – found themselves at the mutual mercy of Taiwan Semiconductor Manufacturing Company. Both of them were dependent on getting their chips from a common 28-nm process technology, and both companies had to figure out ways to differentiate their daily bread when it came from the same baker.
In reality, such differentiation is not hard at all, since much of the value-add for complex processor chips comes from the basic logic design, and from the instruction set. Oops, cancel that last part – Qualcomm, ARM, Xilinx and Altera all use the ARM core in at least some of their designs. But the intellectual property of associated blocks, and the software residing on top of that ARM instruction set, allows many chip players to offer unique products, even as they all are forced to turn to a pair of sole sources of 28-nm silicon.
The two companies in a unique position to control destinies of dozens of players, though, are TSMC and Global Foundries. Everyone in the semiconductor industry complains about the outrageous capital costs involved in building 21st century chip plants. But once you’ve got a plant up and running, you’ve got the whole world in your hands.
Loring
