At Intel's (INTC +2.20%) investor meeting in February, the company's chief technical officer, Murthy Renduchintala, explained why the company's in-house chip manufacturing is a competitive advantage.
One slide Renduchintala showed claimed that Intel "will have enjoyed a leadoff [roughly] 3 years when competitors launch 10nm process."
Image source: Intel.
Then, about a month and a half later, Intel hosted a Technology and Manufacturing day. During this event, the company went into even more detail about its chip manufacturing technologies and strategies.
Again, Intel showed a nearly identical chart to the one that Renduchintala showed and once again claimed to be three years ahead of the competition.
Image source: Intel.
However, I think Intel's claims of such leadership don't stand up to scrutiny. Not only does Intel appear to no longer have a lead (at least when it comes to the density characteristics of currently shipping products), it seems to be a little behind what its competitors appear to be doing.
Indeed, today one can go out and purchase products built using 10-nanometer foundry technologies (which, by Intel's own admission, are denser than its own 14-nanometer technology).
To that end, I think Intel's top brass needs to start seriously thinking about doing the unthinkable: going "fabless."
Intel's competitors executing like clockwork
The reality is that the third-party contract chip manufacturers have been executing rather well over the last several years, delivering chips in high quantities to very demanding customers on schedule.
Take, for example, the world's largest contract chip manufacturer, Taiwan Semiconductor Manufacturing Company (TSM 1.86%). This company has released new leading-edge manufacturing technologies and/or substantial refinements on an annual cadence for years now.
TSMC's 20-nanometer technology was ready to go to support the Apple A8 ramp-up in the second half of 2014, its 16-nanometer technology supported Apple's A9 ramp-up in the second half of 2015, an enhanced variant of its 16-nanometer technology supported the A10 Fusion ramp up, and now TSMC's is cranking out Apple A11 Fusion chips using its new 10-nanometer manufacturing technology.
TSMC has also committed to ramping up its 7-nanometer technology in 2018, an enhanced version of 7-nanometer in 2019, and its new 5-nanometer technology in 2020.
This is a road map that TSMC's customers (which are, not-so-coincidentally, Intel's competitors) can rely on and plan their products around.
Intel, by contrast, is merely committing to "introducing" its first 10-nanometer chips by the end of this year, with serious volumes ramping up across 2018 after substantial delays.
It has also committed to an enhanced version of its 10-nanometer technology (called 10-nanometer+) approximately a year after the first version ramps up, and it has indicated that a further-enhanced version (called 10-nanometer++) would come at some point afterward (no specifics, but Intel's chart seems to indicate that it'll take more than a year).
Image source: Intel.
As far as Intel's 7-nanometer technology goes? CEO Brian Krzanich recently gave a vague timeline, saying it could go into production two to three years after 10-nanometer goes into production.
Intel's execution in chip manufacturing seems inconsistent (though the derivative technologies that enhance performance have helped to bring some predictability), and it is hurting the competitiveness of Intel's product portfolio as manufacturing delays push out the introduction of new processors.
Maybe just go fabless, then?
If Intel can't get its act together with respect to chip manufacturing technology, it should seriously consider going fabless -- that is, Intel should still design its own chips, but it should outsource the manufacturing of those chips to potentially more consistent suppliers.
Such a transition wouldn't happen quickly, and I don't think it'd be wise for Intel to simply decide one day to end its internal chip manufacturing efforts.
Instead, Intel may be best served by giving its product groups the ability to evaluate competitive technologies as viable options for future products. Those groups should have the complete freedom to either choose Intel manufacturing technologies or third-party manufacturing technologies depending on product quality and time-to-market concerns.
If Intel finds that its product teams are overwhelmingly choosing third-party manufacturers for its high-volume products, then Intel could begin the process of winding down its investments in its chip manufacturing plants, technology recipes, and so on.
If Intel finds that its own manufacturing technologies can deliver the best products in a timely fashion, then it could continue its investments in chip manufacturing technology.
The bottom line is this, though: Intel's chip manufacturing group has, as of late, negatively impacted Intel's product competitiveness, mainly by showing up way too late to the party.
Either Intel's chip manufacturing group improves its execution and becomes an asset, not a liability, to Intel's business, or it should be shut down.
