Chipmaker Intel (INTC 4.13%) is a semiconductor industry behemoth. The company generates on the order of $60 billion in annual revenue, throws off a lot of free cash flow, and dominates many of the markets in which it participates, including that of processors aimed at both the data center and personal computer markets.
However, while Intel's business is financially healthy, the company's current financial health seems largely a product of its past successes.
Image source: Intel.
Although I do think that Intel now has some strong executives running critical businesses (Dr. Venkata "Murthy" Renduchintala -- Intel's chief technical officer and a relatively recent addition to the team -- immediately comes to mind, for example), I think that since Intel CEO Brian Krzanich took over in May 2013, the company's execution has degraded on a number of fronts.
Here's one mistake by Intel management that I think is, frankly, staggering.
Stagnant architectural innovation
Historically, Intel developed products on a "tick-tock" cadence. A "tick" would involve transitioning to a new manufacturing technology that would deliver the following: chip area reduction and power efficiency improvement. It would also often include minor enhancements to the architecture of the chip (e.g., slightly enhanced CPU and graphics blocks, potentially updated chip packaging technology, and so on).
A "tock" would involve the company reusing the same manufacturing technology that was introduced in the "tick" step (perhaps with some mild enhancements but nothing major), but delivering a new chip architecture.
This would usually mean new features added to the chip (e.g., integrated image signal processor), as well as substantially improved CPU, graphics, memory controller, and so on.
However, over the last several years, Intel's architectural innovation has slowed to a crawl. Intel last introduced a new architecture in the second half of 2015, known as Skylake. The company's Kaby Lake -- first introduced in the second half of 2016 -- is nothing more than Skylake on a higher-performing variant of the 14-nanometer technology used to build Skylake and its predecessor, Broadwell, called 14-nanometer+.
Later this year, Intel is expected to begin rolling out its eighth-generation Core processor family (Coffee Lake/Kaby Lake-Refresh), and these will, once again, follow the "Kaby Lake" playbook -- same basic processor architectures but built using a newer, performance-enhanced variant of its 14-nanometer technology called 14-nanometer++.
The main "trick" for the eighth-generation Core products, though, is expected to be an increase in core counts for both notebook and desktop parts.
This kind of architectural stagnation -- particularly as industrywide competitive pressures mount -- has been hugely disappointing.
Now, to be fair to Intel, the company is planning to introduce a family of processors known as Cannon Lake in the first half of 2018, and these chips will likely use an updated version of the Skylake architecture, but here's the catch: Cannon Lake will only be coming to some of Intel's personal computer product lines (the ultra-low power "Y" series chips and "U" series chips for thin-and-light notebooks).
How Intel can fix this
There's nothing that Intel can do about the past -- what's done is done, and I sincerely hope that the company learns from its errors and substantially improves its execution here going forward.
At the very highest level -- to make up for lost time -- I would like to see Intel deliver consistent, annual architectural improvements with each of its products going forward.
Such improvements, coupled with annual improvements in chip manufacturing technology performance, could make for a compelling product road map.
Will Intel management invest the necessary financial and human resources to make this happen? I hope so, but only time will tell.
