Intel announced its Q4 2017 and full year 2017 earnings, in what has to be one of the higher-profile earnings calls the company has ever done. For the past three weeks, Intel has taken unrelenting heat for the Spectre and Meltdown flaws. On the one hand, it’s true Intel is easily the most exposed of any CPU vendor, even ARM (which has far more CPUs in the field), and Intel’s initial fixes had to be yanked for being buggy and causing frequent reboots. On the other, some of the more grandiose predictions we’ve seen on the potential ramifications of the two flaws are misguided or generally incorrect. Still, it’s a serious problem, and investors have been waiting to hear what Intel is going to do about it.
Almost secondary to this is the question of how Intel performed throughout 2017 and how AMD’s mobile and desktop Ryzen launches might have impacted the company’s full year results. Luckily, we’ve got info on all three topics.
Intel’s full year revenue was $62.8B, a record for the company, and driven by the growth of its IoT, data center, and cloud computing business segments. Full-year revenue grew 9 percent year over year, driven by 11 percent growth in data center income, 20 percent in IoT (still a comparatively small chunk of Intel’s earnings, at $3.2B for the year), 37 percent growth in non-volatile memory sales (Optane, NAND, $3.5B for 2017), and 14 percent growth in programmable solutions (FPGAs, $1.9B for 2017).
The only less-than-bright spot in Intel’s 2017 was weakness in the consumer market, but we’ll speak to that later.
10nm, Meltdown, and Spectre
Intel confirmed on its conference call that it began shipping some 10nm products very late in Q4 2017 and will continue ramping towards volume production in the latter half of this year.
One interesting point worth highlighting here is how much the dialog around process node ramps has changed within Intel. Five to six years ago, Intel would quickly point out that one major difference between itself and rival foundries like TSMC was the way it talked about new process nodes. It was not (and is not) unusual for TSMC to declare it has hit volume manufacturing milestones 6-9 months before products actually hit store shelves. This is a consequence of the foundry-customer relationship, in which products go through additional lengthy verification steps before ever going on sale.
TSMC’s messaging practices haven’t changed, but Intel’s have. Once upon a time, “volume manufacturing” was something Intel announced a month or two before products were available for purchase. Now, the company is talking about a Q4 2017 early ramp, with actual commercial availability in Q3 or Q4. New process nodes are difficult, and Intel’s 10nm is essentially equivalent to the 7nm nodes at GlobalFoundries, Samsung, and TSMC, but there’s no denying that the company’s introduction times have lengthened.
As for the in-silicon patches for Spectre and Meltdown, here’s what CEO Brian Krzanich had to say:
We’re working to incorporate silicon-based changed to future products that will directly address the Spectre and Meltdown threats in hardware. And those products will begin appearing later this year.
This indirectly confirms Intel knew about these flaws long before they became common knowledge. Typically, a new CPU design is frozen at least a year before products ship for revenue. For Intel to have hardware mitigation strategies ready to go this quickly, it must have had time to bake those improvements into Ice Lake before that CPU finished tapeout. Intel also plans to launch some new 14nm products this year, but there was no word on what those might be.
What About AMD?
All year, AMD fans and enthusiasts have wanted to know what impact Ryzen and Ryzen Mobile would have on Intel’s market share. AMD’s earnings have made it clear Ryzen delivered substantial revenue improvements, but revenue gains and market share gains don’t always cleanly track each other, and Ryzen commands far better prices than Piledriver ever did.
For full-year 2017, Intel’s notebook platform shipments rose by 5 percent, ASPs (average selling prices) and ASPs rose 2 percent. Given that Ryzen Mobile only barely shipped in 2017, this is not particularly surprising.
Intel never mentioned AMD by name in its conference call, but it’s hard not to see the company’s hand in Intel’s desktop performance. Desktop sales fell 5 percent in 2017, a much steeper drop in that segment than forecast for the year, and the decline in desktop ASPs from Q4 2016 to Q4 2017 could also be indicative of increased competition from Team Green. Intel made generic reference to increased competition, but otherwise blamed these declines in continuing weakness in the PC market. It’s not at all clear that’s true, and we should know more next week when AMD reports its own 2017 results.
Will Meltdown, Spectre, Prove a Benefit or Liability?
The last point we want to touch on is whether Meltdown and Spectre will actually harm Intel at all, long-term. Data centers and servers are the hardest hit by the software patches that mitigate the bugs. But some analysts think this could work to Intel’s advantage, increasing the chances that customers will jump for early refresh cycles once new 10nm chips are available that aren’t susceptible to this vulnerability.
Whether this works out in Intel’s favor may depend on how much momentum AMD is able to build for its Epyc server family, and how much of a performance whack its own chips take (AMD’s exposure on some of these issues is different than Intel’s). It’s also not hard to see how customers might feel burned by Intel and look to AMD as a result, but Lisa Su has previously stressed that the Epyc ramp would be a slow, gradual one — and AMD had prior knowledge of Meltdown and Spectre as well.
Of course, it’s possible both outcomes could happen. Meltdown and Spectre fixes in-silicon could spur many companies to upgrade early, while some customers might opt to consider non-Intel solutions after getting burned by Chipzilla. Whether those sales accrue to Qualcomm’s Centriq server division, AMD’s Epyc, or some other vendor is something we’ll watch in 2018.
Should Spectre, Meltdown Be the Death Knell for the x86 Standard?
Spectre and Meltdown are serious CPU flaws, but do they warrant throwing out the entire closed-source CPU model?
Microsoft Pulls Spectre, Meltdown Patches for Older AMD Systems
Microsoft has halted the distribution of patches for Spectre, Meltdown on some AMD systems due to problems with BSODs and bricked systems.
AMD Releases Updated Risk Guidance on Meltdown, Spectre
AMD has clarified its own weaknesses against Spectre and Meltdown. One week later, things aren't quite as rosy as AMD hoped they'd be.