Image source: Intel.
Microprocessor giant Intel (INTC 0.07%) went into some detail about its plans to build up its contract chip manufacturing business, known as Intel Custom Foundry. According to the company, it offers three main manufacturing technologies to prospective clients: 22-nanometer, 14-nanometer, and 10-nanometer. The smaller the number, the more advanced the technology is when comparing two technologies from the same manufacturer.
Interestingly, in an interview with Semiconductor Engineering, Intel's Mark Bohr had the following to say with respect to the company's 22-nanometer offerings:
Over time, our 22 [nanometer] platform may become more attractive to customers that are more at the trailing edge. As people look to move forward from 28 [nanometer] today, they may look at Intel 22 [nanometer] and 14 [nanometer] as options. There are also people on 40 [nanometer] today that are still thinking about going to 28 [nanometer] or 22 [nanometer].
Although on the surface this argument seems pretty reasonable, I don't think Intel will see many, if any, takers going forward on this technology. Here's why.
The competition is quite fierce
Intel faces some very stiff competition from Taiwan Semiconductor Manufacturing Company (TSM +1.97%) for the customers that the former would like to court to use its 22-nanometer manufacturing technology. On one hand, TSMC has continued to offer cost/performance enhanced variants of its 28-nanometer technology over time, such as 28 HPC and 28 HPC+.
On the other, TSMC is also offering a cost-reduced variant of its 16-nanometer technology known as 16FFC.
So, in terms of cost-optimized relatively high-performance processes, Intel's 22-nanometer technology gets squeezed in between TSMC's 28 nanometer HPC+, which I suspect offers worse performance than Intel's 22-nanometer tech but is cheaper, and TSMC's 16-nanometer FFC, which should offer more performance than Intel's 22-nanometer technology but could be more expensive.
I just can't see 22-nanometer doing particularly well with fabless customers -- customers who outsource fabrication, particularly given the breadth and depth of TSMC's offerings targeted at such customers.
Where Intel's foundry biz can create value
In general, I don't think Intel Custom Foundry will be all that successful in winning business from customers that are looking for "yesterday's technology on the cheap." In fact, the only reason I can see Intel Custom Foundry potentially succeeding is in offering differentiated solutions for customers looking for performance that they can't get from other manufacturing partners.
To that end, I can see Intel's 14-nanometer technology potentially winning away business that might have otherwise gone to TSMC's 16-nanometer technology, and Intel's 10-nanometer technology being used to lure customers who might have chosen TSMC's 10-nanometer or 7-nanometer technologies instead.
That being said, aside from the recently announced deals with LG, Spreadtrum, a company in which Intel owns a partial stake, and Altera, now owned by Intel, Intel's foundry business doesn't seem to have gained a ton of traction.
Before investors declare victory for Intel or doom and gloom for TSMC, I recommend waiting two to three years to see if Intel can land significant wins with its leading-edge technology. If Intel can actually drive significant volumes of third-party chips on its older 22-nanometer technology, I would be beyond surprised.
