Eric Jhonsa, writing for The Street, recently argued that microprocessor giant Intel's (NASDAQ:INTC) recently announced seventh-generation Core processor family (which I'll refer to from here on out by the group's code name, Kaby Lake) is "the product of a cautious strategy that could backfire."
What he's referring to specifically is the fact that Kaby Lake is the third generation of product built on Intel's 14-nanometer chip manufacturing. Jhonsa argues that Intel's new strategy "allows it to spread out capital investments for a given process over a longer time frame."
He further goes on to say that Intel is "betting the cost savings that come from paring back its annual PC CPU and manufacturing investments will more than offset the narrowing (if not the narrowing) of its traditional technology leads" against key fabless chipmakers and their respective manufacturing partners.
"It's a calculated gamble," he writes. "one that could easily start to backfire next year."
I have a great deal of respect for Jhonsa and his work, but I disagree with him on this. Here's why.
Why Intel really shifted to this new strategy
It's true that moving to new manufacturing technologies has become more difficult over time. Shrinking transistors and the associated metal stacks while keeping yields under control and moving performance up is hard and only getting harder. This is especially true as the industry as advancements in lithography tools have forced chip manufacturers to adopt increasingly exotic and expensive multipatterning schemes to build smaller chip features.
This isn't a case of Intel paring back investments in chip technology research and development, nor is it a case of Intel wanting to wring more out of its capital investments. Rather, it's simply a case of Intel running into very difficult research and development challenges that force it to take longer to be able to get new manufacturing technologies into a production-worthy state.
So, what's Intel doing about it?
If Intel were simply just paring back investments in manufacturing technology as well as in chip technology, then the company would not have released Kaby Lake; it would have simply allowed the prior-generation Skylake parts to stay in the market for an additional year.
Instead, the company did a couple of really interesting things. First of all, Intel went ahead and improved the 14-nanometer manufacturing technology that's used to build Kaby Lake. The company actually made real changes to the transistors as well as to the metal stack to wring out, according to the company, a full 12% worth of process-related performance improvement.
Now, the changes done to the 14-nanometer process to create what Intel calls 14-nanometer+ don't just give a "free" performance boost. That is, Intel can't just take last year's designs, push a button, and have those designs roll off the line on the enhanced process. Instead, Intel had to go back and do quite a bit of work to the design to allow it to actually reap the benefits of this new manufacturing technology.
All of this work ultimately led to Intel getting a product family out that delivers a good deal more performance in a given power envelope than its predecessor.
Going forward, Intel isn't planning on letting up, either. The company says it will do three generations of products on its 10-nanometer manufacturing technology: 10-nanometer, 10-nanometer+, and 10-nanometer++. I expect this new cadence to allow Intel to put out newer, more efficient, and better-performing products each year without the inherent unpredictability that trying to move to a whole new manufacturing technology generation (and failing) leads to.
If anything, this new strategy put Intel in a better position to maintain performance/power leadership in key product areas (PC processors, server processors, and so on) going forward.