Intel has been buffeted in recent weeks by a significant layoffs and the need for a new approach to long-term growth and stability. On Tuesday, CEO Brian Krzanich laid out his vision for the future of the CPU giant in a lengthy public memo.
Krzanich’s memo highlights five areas he believes will be critical to Intel’s growth and future: The cloud, the Internet of Things, memory and FPGA technology, 5G wireless, and Moore’s Law. This last is more of a pro forma bullet point than an argued position, since the broad consensus in the semiconductor industry is that Moore’s Law as traditionally defined is more-or-less dead.
The data center points are fairly standard, too, though it’s not clear how serious Intel is about trying to expand this segment. The advent of the cloud will probably continue to boost data center sales, but while this market is extremely profitable per unit, it’s a mature space that Intel already dominates.
The Internet of Things needs some sharpening, which Krzanich offers. Intel isn’t interested in chasing the entire IoT space, just specific segments of it. Krzanich writes: “At Intel, we will focus on autonomous vehicles, industrial and retail as our primary growth drivers of the Internet of Things.” Autonomous vehicles, industrial applications, and retail POS products are going to be higher margin and higher visibility as compared with legions of FitBit trackers or even smartphones. Intel’s move towards the higher-margin areas of IoT are meant to protect its margins and cost structures as much as anything.
FPGAs (via the Altera purchase) and Intel’s 3D XPoint will be part of the company’s cloud and data center offerings and are again intended to help boost the bottom line. 3D XPoint is an interesting technology, to be sure, but one with as-yet unproven potential.
Intel’s decision to list connectivity as a major focus is a little surprising given how little success the firm has had in modems. Its current designs are still built at TSMC and all of Intel’s modems, including its as-yet unreleased XMM 7480, are still based on 28nm hardware. While it’s absolutely true that baseband radios don’t gain from process nodes the way semiconductor logic does, it’s also strange to see the company banging the 28nm drum when competitors like Samsung and Qualcomm are already pushing 14nm radios in products. Intel is promising to deliver end-to-end 5G compatibility, which may mean that the company is planning an aggressive leadership position it never struck in the 4G market.
We’re not going to tear into the Moore’s Law arguments because we’ve covered talking points like this before. In the best-case scenario, Moore’s law is reinvented and remains a useful way of talking about progression in systems and engineering but stops describing anything specific about transistor densities over time. In the worst case scenario, scaling and integrating slam into various roadblocks over time — never stopping, necessarily, but advancing less and less in any given period.
Krzanich’s memo makes it extremely clear how the PC fits into Intel’s new market paradigm. Remember the quote above about the Internet of Things? Here’s the very next sentence. “[W]e view our core Client business of PCs and mobile as among the many variations of connected things, which is driving our strategy of differentiation and segmentation in the Internet of Things business.”
In other words, yes, PCs are still important, but not because their PCs. In the future, your PC is just one more device you happen to access the cloud with. The market will continue to be important to Intel, but it’s no longer a growth platform or even a major pillar — at least not rhetorically.
Intel’s newfound and hardcore data center focus could spell bad news for AMD when the smaller chip company attempts to re-enter the market for data center parts with its Zen CPU. You can bet that Intel, who increasingly leans on the data center for profits and the kinds of earnings that make Wall Street happy, will not give up market share easily or quickly. While I expect Zen to be a reasonably strong core, there’s very little chance that AMD’s performance will leap from current levels to matching the latest Skylake processors. The gap between the two companies is now much wider than it was at the end of the K7 era, when AMD was pushing its last gen forward on fumes and frantically ramping Opteron towards commercial availability.
If AMD’s products are good it’ll undoubtedly gain some market share for itself, but it has the misfortune of facing down an entrenched competitor with extremely good reason to fight it tooth and nail. Getting Zen out the door on-time is a major goal for AMD, but that’s just the start of the fight, not the end.