The semiconductor foundry business has not seen the level of cutthroat competition that we are seeing today in a very long time. Taiwan Semiconductor (NYSE:TSM) has had a virtual monopoly on leading edge foundry capacity at the 28 nanometer node until very recently. This is primarily due to the fact that TSMC's management went with a "gate last" implementation of high-K metal gate transistors while the IBM Fab Club consisting of Samsung (NASDAQOTH:SSNLF), Global Foundries, IBM (NYSE:IBM) mistakenly pursued the much more difficult to manufacture "gate first" implementation of high-K metal gate at 28 nanometer.
While Samsung races to get its 20 nanometer node in production in order to counter TSMC (which has made it clear that it sees "very little competition" at this node), the company is already talking up its 14 nanometer node (which is the first node at which the company implements FinFETs) and the press and rumor mill alike are going wild with speculation. There are now claims floating around that Samsung will release smartphones with this FinFET technology by the end of 2014 – ahead of TSMC and apparently in-line with Intel's (NASDAQ:INTC) own 14 nanometer technology. This does not pass the smell test.
Samsung, and the rest of the foundry world, was caught off guard
When Intel announced back in 2011 that its lead PC processor codenamed Ivy Bridge would be built on the company's 22 nanometer FinFET process, this sent the rest of the semiconductor world into a race to try to catch up. This led to a dramatic revision in PowerPoint slides from Samsung, Global Foundries, and TSMC claiming that their FinFETs would be available by the end of 2014/early 2015. Investors were even treated to a flurry of announcements that test chips had taped out, but while this may sound impressive, these all suggest that the foundries are quite a ways off from production-level yields.
While Taiwan Semiconductor claims that its first FinFET process will be in high volume production by the first half of 2015, Samsung has been boldly claiming that it would be ready to go into high volume production during the back half of 2014. While only Samsung knows exactly when designs based on its 14 nanometer process will actually be available in products/high volumes, there are a number of clues that suggest that "late 2014" or even "early 2015" for products shipping in consumer devices is highly optimistic.
Intel clues the rest of the world in
At the ARM TechCon, Samsung – in a partnership with Cadence Design Systems and ARM (NASDAQ:ARMH) – demonstrated the first working silicon of its 14 nanometer process. This was a test chip that was based on an ARM Cortex A7 processor core (this is a small, low-end core) coupled with a 128Mbit SRAM cell. The big achievement here was that the test chip worked.
Unfortunately, for a company claiming that it will be hot on the heels of Intel, this doesn't quite add up. At the Intel Developer forum, Intel demonstrated a working sample of its next generation "Broadwell" system-on-chip. This wasn't a test chip, but rather an engineering sample of a highly complex CPU combined with graphics and a large L3 cache. It wasn't just up and running, but it had successfully booted Windows and even successfully completed a very sophisticated CPU benchmark. Intel claims that it will be going into high volume manufacturing of these products (which took years to design) in the first quarter of 2014 for a launch in the second half of 2014.
So, if Intel – which is demonstrating fully completed, complex designs on its 14 nanometer process – is going into production in the early part of 2014 and not launching products until the back half of 2014, what makes the investment community think that Samsung will be able to ship smartphones with fully completed 14 nanometer designs by the end of 2014 or even early 2015? The timelines just don't appear to work.
Foolish bottom line
Right now, the foundries are in a race to "catch up" to Intel, but investors who buy into this seem to make the mistake that Intel is simply standing still. TSMC will go into production on its 20 nanometer process (probably high-end FPGAs will be first) during Q1 2014 and it is likely that mobile SoCs based on this process will be in devices during the second half of 2014 (think iPhone 6). Mobile chips from non-Intel players featuring FinFETs likely won't find their way into the hands of consumers until mid to late 2015, which means that Intel – which will launch its 14 nanometer tablet chips in the back half of 2014 – still has a sizable process technology lead over Samsung and the rest of the industry.