Chip giant Intel (INTC -1.87%) spent a significant amount of time last year trying to make the case to investors that it had a significant leadership position in chip manufacturing technology. The company claimed that its upcoming 10-nanometer technology can deliver twice the transistor density that other chip manufacturers' 10-nanometer technologies can.
While Intel's claim is probably true, the reality is that competing 10-nanometer technologies went into production late in 2016 or early in 2017, while Intel's own 10-nanometer technology is not yet being used to manufacture salable products. The huge delay in the mass production of Intel's 10-nanometer technology meant that Intel's comparisons were meaningless -- Intel's 10-nanometer technology won't be competing with rivals' 10-nanometer technologies, but instead with their upcoming 7-nanometer technologies.
Image source: Intel.
At the recent International Solid State Circuits Conference (ISSCC), an Intel engineer seemingly admitted to what Intel's marketing department won't: Intel's 10-nanometer technology is behind competing 7-nanometer technologies in a crucial area.
Density lead lost
Most computer processors incorporate a type of extremely fast memory, known as SRAM. Since SRAMs are such common structures, one simple way to compare the relative densities of chip manufacturing technologies is to look at the area that an SRAM bit cell occupies. Now, SRAM bit cells can be made up of different numbers of transistors (transistors are the basic building blocks of a chip), so for such a comparison to be valid, the SRAM bit cells need to be made up of the same number of transistors.
According to Intel, a six-transistor SRAM bit cell implemented in its 10-nanometer technology measures 0.0312 square micrometers. Competing six-transistor SRAM bit cells implemented in Samsung's (NASDAQOTH: SSNLF), Taiwan Semiconductor Manufacturing Company's (TSM -4.07%), and GlobalFoundries' 7-nanometer technologies measure in at 0.026 square micrometers, 0.0272 square micrometers, and 0.0269 square micrometers, respectively.
Image source: Intel.
Indeed, while the other three companies' 7-nanometer technologies are all within spitting distance of one another in terms of six-transistor SRAM bit cell size, Intel's 10-nanometer stands out as being woefully behind.
According to an EE Times article, the Intel engineer who presented the details of the company's 10-nanometer SRAM bit cells at the ISSCC admitted that Intel's technology was "within 15 percent of the smallest reported 7-nm cell."
Considering that Intel has routinely claimed to be substantially ahead of the competition in terms of transistor density, this is quite an admission.
What does this mean for investors?
Ultimately, Intel's loss of density leadership in chip manufacturing technology (at least as far as SRAM bit cell density is concerned) is just another sign that the company's chip manufacturing group simply isn't executing well.
Minor differences in SRAM bit cell sizes aren't going to fundamentally shake up the competitive landscape, but the trend that we're seeing very well could.
Intel's manufacturing delays have led to delays (and, in some cases, cancellations) of important products that have hurt the company's competitive position. Moreover, these manufacturing technology delays seem to affect Intel alone -- TSMC and Samsung have consistently met their manufacturing technology release schedules over the years, and it's even starting to look as if GlobalFoundries (which has historically been unreliable) is cleaning up its act.
In addition to delays, it's starting to seem like Intel's manufacturing technologies are less competitive than they historically were, no doubt due to the substantial delays. Intel seems to have lost the lead in density (which is important for keeping chip sizes competitive), and, if it's not careful, it could lose whatever lead it might have in manufacturing technology performance and power efficiency.
At this point, I think Intel has a decision to make: Does it double-down on its own chip manufacturing group and try to dramatically improve its execution, or does it simply shutter its chip manufacturing operations and outsource chip production to seemingly more-capable chip manufacturers?
The problem with Intel doubling down on its own chip manufacturing technology is that if it's unsuccessful, it risks seeing a further erosion in its competitive position (since Intel's competitors rely on TSMC, Samsung, and GlobalFoundries for chip manufacturing), which could negatively impact its business performance for many years to come.
