This past week, Apple (NASDAQ:AAPL) dealt a terrible blow to Imagination technologies. Imagination technologies' investors saw their stock drop by nearly two-thirds in a single session. The reason? Apple warned Imagination Technologies that it would no longer be licensing its GPU IP in the future.
Yesterday, Dialog Semiconductor (OTCPK:DLGNF) got some of the same treatment and at one point lost more than one-third of its value. This happened just because one analyst put out his opinion that Apple might also do without its wares (power management integrated circuits).
So, first Apple came for the GPU suppliers and you did not care, because you had no shares in Imagination technologies.
Then Apple for the PMIC suppliers and you did not care, because you had no shares in Dialog Semiconductor.
But then, who will Apple come for next?
Adapted from Martin Niemöller's original
There is a method to what Apple is doing. And it isn't madness. Apple didn't start with GPUs, either. Instead, Apple started this process with CPUs, back in the transition from the Apple A5 used in the iPhone 4, to the Apple A6 used in the iPhone 5. With the Apple A5, Apple used standard ARM A9 cores. With the Apple A6, Apple started using its instruction set compatible, but custom designed, Swift cores.
Why did Apple do this? Quite obviously, because it saw an advantage in doing so. In this case, the advantage was in furthering characteristics which it saw as large advantages for the mobile use case. Those advantages were materialized through using a wider, higher single-threaded performance CPU, instead of the much lower single-threaded performance of the standard ARM designs. The advantages were:
- Lower power consumption per unit of performance (due to higher performance per clock cycle).
- Much greater performance in lightly threaded applications. At the time, with users basically using one app at a time in a smartphone, this translated into much higher performance for the task the user was looking at (be it the user interface itself, or the app the user was using).
The jump in single-threaded performance was tremendous, more than 150% if we take Geekbench 4 as a proxy:
Source: PrimateLabs.com; Geekbench browser
The jump in performance wasn't just tremendous relative to its own past performance. It also gave Apple an instant competitive advantage: at the time, only Qualcomm (NASDAQ:QCOM) didn't rely on standard ARM cores. What happened was that competitors, Samsung chief amongst them, had to rely on a route they're still following today:
- They had to compete using more cores at a higher clock speed, because each core delivered much less performance than Apple's custom approach.
The result was that while the Samsung SII could go head to head with the iPhone 4s when it came to performance. Indeed, it could surpass the iPhone 4s handily due to a higher clocked CPU. The Samsung SIII could no longer make such claim, as both its standard A9 cores and Qualcomm's Krait cores at 1.3GHz were all incredibly far from Apple A6's performance (their single-core improvement over the iPhone 4s generation was around 50%, versus Apple's 150% advance).
So again, why did Apple go this route of using its own IP? Because it gave it an advantage. In this case it was both a performance (and thus usability) and a power consumption advantage (faster race to idle, lower core count, lower frequencies).
As we saw, Imagination Technologies has now been the victim of Apple internalizing IP. Why would Apple do so? Certainly, to seek an advantage. What could that advantage be? Drawing from the past and speculation we could argue what it's not:
- A cost advantage. Not likely. At $81 million in royalties paid to Imagination Technologies over FY2016, this comes to less than $0.40 per iPhone and is not a major cost driver. Any effort by Apple to replace Imagination Technologies IP is likely to be as expensive or more expensive than that, between own resources and third-party royalties.
Two other reasons remain:
- Primarily a performance advantage. However, a performance advantage can then be turned into a power consumption advantage simply by running the GPU at lower frequencies (the same way Apple proceeded with its CPUs).
In my view, performance was the problem which ultimately drove Apple away. Consider the following:
- Apple has already been building its own custom GPUs (since the Apple A8 SoC), based on Imagination Technologies' IP.
- Since the Apple 6, Apple thus has had a large advantage in GPU performance, especially considering that it also generally uses lower resolutions than its competitors (750p on the smaller iPhone, 1080p on the Plus models). However, in terms of raw performance, we can point to benchmarks showing a closing of the performance gap:
Apple might thus have seen that the current trend would put it at parity or at a disadvantage within the next couple of years. In fact, the Snapdragon 835 seems to have marked a letting up of competitive pressure, but Apple couldn't have guessed that. At this point, to rekindle its competitive advantage, Apple thus had two possible choices:
- The harder way, which Imagination Technologies doesn't believe Apple is able to (and thus threatens Apple with a lawsuit if it tries). This would be for Apple to evolve its own IP enough for a ground-up GPU design.
- And the easier way, which is for Apple to continue producing custom GPUs, but having their basic design be based on someone else's IP. For instance, this could be ARM's Mali (since the recent G71 design seem competitive), AMD (NASDAQ:AMD), or Nvidia (NASDAQ:NVDA) - with Nvidia having already proved able to provide class-leading mobile GPU performance. For instance, the Tegra X1, though admittedly is only used in tablet-class devices, pushes 118fps on the same T-Rex benchmark as above, and 63fps on the Manhattan benchmark. The Tegra X1 is Nvidia's previous Maxwell-based generation (and is also the SoC used in Nintendo's Switch). Nvidia has now moved to Pascal-based products (and soon enough, Volta).
What Next, Then?
So we've seen why Apple went into making custom CPUs. We've also seen why it first went into making custom GPUs, and likely why it's now ditching Imagination Technologies.
In these two central pieces of its SoC, Apple seems to want to leave nothing to chance. It wants a leadership position in performance and power consumption, while quite likely also lending an eye to cost/margin.
What other areas are prone to this kind of substitution? Let's see:
I've already talked about how Mac CPUs might be up for grabs, in my article titled "Intel Should Worry A Lot If Apple Excludes It From Future Macs."
This is a result of Apple's ongoing improvement of its mobile SoCs. Apple is getting to a point where they're powerful enough to drive MacBooks. If the evolution continues - and there's a lot of room for it to continue, for instance by using more higher performance cores, then at some point it's a near-certainty that Apple will ditch Intel (NASDAQ:INTC) and use its own SoCs.
When will this happen? This is now a risk which will be present every year going forward, with each Mac iteration. The CPUs have theoretically attained the necessary performance already. We should keep in mind that a version of these CPUs for laptops would run to a higher power envelope, and thus deliver even higher performance than the "good enough" performance Apple is already showing in its mobile CPUs.
The impact for Intel is large, as explained in the above-linked article. This move would also deliver a large margin advantage to Apple, hence Apple has a very large incentive in going forward with it.
Presently Apple uses separate modem solutions (chips) from either Qualcomm or Intel.
However, integrating WiFi and especially a LTE modem is a no-brainer in terms of power consumption and logic board simplification. This is what Qualcomm and even Huawei already do in their respective Snapdragon and Kirin high-end SoCs. Samsung, for its part, is expected to do the same in the Exynos line later in 2017, by integrating a Shannon modem.
If and when Apple integrates a modem in its SoC, Intel and Qualcomm will lose some hardware business. However Qualcomm keeps the most important thing - royalties on necessary wireless patents (but that's a whole other ball of wax).
Presently, Apple uses 4 audio ICs in its iPhone 7 line. There are:
- Three in the logic board.
- One inside the lightning to headphone jack adapter bundled with the phone.
Of the three within the logic board, two are expected to drive both top and bottom speakers, and one is speculated to be an amplifier working through the lightning port. This latest explanation makes no sense to me as the lightning port is fully digital. I'll wildly speculate, here that this third DAC is there because the decision to include or not include the headphone jack in the iPhone 7 was made late in the process.
Most importantly, the audio IC within the headphone adapter will only be present for as long as Apple bundles said adapter. Arguably, bundling the adapter was a short-term decision to appease the masses regarding the lack of a headphone jack. It's likely that sooner or later, perhaps as soon as the next iPhone iteration, the headphone adapter will be gone, and the associated audio IC will be gone with it.
This move by Apple would simplify the logic board and reduce costs / increase margins. Indeed, the savings obtained by Apple if it did this would exceed any savings from not paying less than $0.40 per device to Imagination Technologies.
So this is the risk: Cirrus Logic (NASDAQ:CRUS) might see ¼ to ½ of its iPhone content be gone as soon as those two "at risk" audio ICs are gone. One of them being gone can be readily established just by seeing that the headphone adapter is no longer present. The other will require teardowns.
In its latest quarter Cirrus Logic derived 85% of its revenues from Apple, so this is a very relevant risk for Cirrus.
There are many Apple suppliers which can fall victim to Apple integrating or otherwise reducing their participation in Apple products.
Apple clearly does this when it sees a good enough advantage, be it on performance, power consumption or cost. There are several identified areas which are likely to see moves by Apple in the next few months and years. These include the Mac PC and laptop lines, audio ICs and modems.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.