Thoughts on AR and Smartphones

Apple’s recent announcement at WWDC 2017 of ARKit has suddenly sparked a new interest in AR, and just a few days ago, Google responded with their ARCore API which does basically the same things. These APIs promise to bring sophisticated AR to all owners of compatible smartphone, dramatically enlarging the addressable market. This has excited pundits and developers alike, and it is now very likely that we will see AR breaking out of the early adopter market and into the early majority.

I would like to jot down my thoughts on this, regarding how innovation tends to happen, and also regarding how this affects the smartphone market as a whole.

The innovation trajectory

Innovation usually starts out as basic research or a new invention somewhere in a lab setting, with specialised and extremely expensive equipment. Then it becomes a specialised instrument that is somewhat cheaper, but still  very hard to use, catering only to the enthusiasts. Finally, it becomes cheap and easy enough to use for the majority of the market, leading to an explosion of usage.

If you look at the PC market for example, it started out when computers like the ENIAC were first invented. Then after revolutions in semiconductors, the computer became cheap enough for enthusiasts to own and tinker with, but still the people who used them were tinkerers. This was the era of the Apple II, Commodore and Amiga. Then as the IBM PC and clones came out, the price came down whilst performance improved dramatically and application software became available, making the PC both useful and affordable thus bringing it into the mainstream.

What we are seeing with ARKit very much mirrors this. We started out with work that was done in the labs, and we recently started to see implementations that require special headsets and powerful PCs connected to them. It was clear at this point that these devices would not go to the mainstream, and the developer community in general did not yet think that it was a market worth pursuing. With ARKit, we are likely to see an IBM PC moment where AR goes mainstream on devices that are affordable and where the developers are also excited to reach large audiences.

One interesting thing to observe is, although the innovation in the lab tends to follow a steady path, the introduction into the mainstream can feel very abrupt, caused by new products being introduced into the market. ARKit is one of these examples. Windows 95 is another. So is the iPhone of course. Sometimes the company introducing the product if far ahead in technology, but this is not necessarily the case. Looking at Windows 95, one could argue that you could be technically behind, but still make a huge and abrupt impact to the market. Judging by how quick Google was to introduce ARCore, it would be fair to say that Apple’s ARKit team was not also necessarily ahead of Google’s, but simply had the right marketing priorities in place.

Implication for smartphone sales

Until very recently, the general narrative was that innovation in smartphones had winded down, and that there was increasingly little to differentiate the new high-end flagships from the models from previous years, or those from mid-tier vendors. However if ARKit and ARCore become mainstream and are adopted by major developers, this changes the whole game. Apple’s ARKit reportedly only works with A9 chips and above, which means that any models prior to the iPhone 6s are not eligible. The iPhone 6 which was introduced in 2014 will not be able to run ARKit apps and neither will the iPad Air 2 (also introduced in 2014). The situation on the Android side is even more dire (as usual) with only the Pixel (introduced in late 2016) and the Samsung Galaxy S8 (introduced in 2017) being eligible to date. Therefore these two platforms may become a strong differentiator for the newest and greatest models, and drive customers who would have otherwise have been content with an old or midrange model to look upmarket.

This has quite a few implications.

  1. Since we know that the high-end smartphone market is generally dominated by Apple and Samsung, this trend will strongly favour these two companies.
  2. Regarding the iOS vs. Android balance, in the high-end (dependent on country) this often is tipped towards iOS. Therefore, we might see a rise of iOS market share.
  3. The interesting thing is that since ARCore currently only supports the Pixel and the Galaxy S8, and since sales of the Pixel are minuscule to date, in reality the vast bulk of ARCore capable smartphones will be Samsung models. As a result, depending on how Samsung plays its cards, it may be possible for Samsung to exert a huge amount of power over Google. Samsung might customise or add proprietary features and APIs to ARCore that would take advantage of the specific hardware on their devices (which might be Samsung silicon), and since the bulk of the market will be Samsung devices anyway, the developers might bite this time. Samsung has been flexing its muscles looking for a chance to break away from Google’s control for quite some time (Tizen, Samsung Pay, security features, etc.), so it would be interesting to see how they plan to take advantage of this situation.

According to Clayton Christensen’s theories on integration and modularity, markets where customers are still looking for innovation should generally favour integration due to the capabilities that this makes possible, and in the smartphone market, the integrated players are Apple and Samsung (Sony is also interesting in that it is integrated around the hardware that matters most – the imaging sensors). It will be interesting to see how this all plays out.

  • obarthelemy

    You’re reading way too much into ARcore’s initial launch devices.

    1- Google always introduces non-core features on a very limited list of phones, think of it as a soak test. Google Assistant and Daydream started Pixel-only; nowadays Google Assistant is on all Android phones above 6.0 (50% of Androids then, you can sideload it onto 5.0+ IIRC) and Daydream on phones from $400 list in the US ( https://vr.google.com/daydream/smartphonevr/phones/ ) this one has a laundry list of requirements: https://source.android.com/compatibility/android-cdd#7_9_virtual_reality , expect ARcore to be mostly same + some requirements on camera -though not dual-lens-.
    This time around, they’re introducing ARcore on an OEM device instead of just a Pixel, that’s actually a broader rollout ^^

    1b- remember, Android isn’t iOS, ARcore will be supported not via an OS update, but via independent modular libraries, so new (and by new I mean old) devices can easily be retrofitted to support it assuming the hardware elements are present (not certain, they’re a bit… esoteric).

    2- Assuming AR/VR gain traction, which they haven’t yet, I’d assume phones above $150 to start supporting either or both. It seems Qualcomm’s entry-level 425 ( https://www.qualcomm.com/products/snapdragon/processors/425 ) found in Xiaomi’s $100 Redmi 4A ( https://www.geekbuying.com/item/Xiaomi-Redmi-4A-2GB-16GB-Smartphone—Gold-373914.html ) is enough for Daydream in terms of CPU/GPU/VPU/ISP/… power and features. That phone misses other requirements: 1080p screen; maybe also screen reactivity, some sensors, and some low-level drivers/interfaces, but not $50 worth in all; we’re still waiting for the precise ARcore requirements, but if it’s Daydream+camera stuff, it’s not $50 either.

    Look at the $125 Redmi 4X or the bigger $150 5.5″ Redmi Note 4x: they have a metal body, touchID, SD slot, FM radio, 4000mAh battery, IR blaster, 3GB/32GB, even honest if not exceptional 13MP & 5MP cameras… most everything a flagship offers; actually, some stuff some so-called flagships don’t even offer ^^ And that’s today’s features and prices, I’d expect a motivated OEM to be able to make a $150 Daydream+ARcore phone next year. But there will probably be no reason to so soon, let’s save $20 instead.

    3- As for Samsung “embrace and extend”-ing ARcore… It’s a weird remark because a) it’s obvious they won’t be exclusive very long and b) they’ve tried with various features for years (Windowing in 2011, Pen in 2011, GearVR in 2015, DeX in 2017, even Tizen). They have yet to get any 3rd-party/dev traction, ever, on any of those. I wouldn’t hold my breath. That angle sounds a bit FUDish actually ? You come across as partisan. Edit: actually, Android is designed to foster OEM innovation. And if something sticks, Google swoops up the feature/market… it’s a dupes’ bargain, but unimpeachable by now.

    Edits: clarity, grammar.

    • Lots of stuff here that we don’t know yet. Apple has restricted ARKit to a limited set of devices which is a subset of which will get iOS 11, and so it would seem that quite a high level of hardware is certainly needed. If ARCode is similar, then maybe it will be harder to support on OEM devices than Google Assistant. This is what I’m assuming. Yes other OEMs will want to support it, but there will be costs associated and there might be a certain level of hesitation.

      And regarding Samsung, Tizen SmartWatches today command a higher market share than all Android Wear combined, I believe. Sure they have failed many times trying to distance themselves from Google, but that hasn’t stopped them from trying, and I do expect them to succeed in a big way eventually. For stuff that requires 3rd party developer support, it will naturally be harder, but if AR turns out to be a high-end play for the next few years, there may be a slight window that Samsung might be able to exploit. I at least expect Samsung to give it a try.

      • obarthelemy

        ARcore certainly requires more than gAssistant, I was just citing that as an example of a limited initial rollout that quickly got wider, so let’s wait a quarter or two. Daydream is more comparable, and in spite of utter lack of consumer interest, there are 8 phones from 6 OEMs, $400 and up, after 10 months, from an initial rollout on 2 Google devices. $400 is hardly premium, and we can expect that cost to tumble down quickly if the feature ever becomes mainstream.

        Look at the actual requirements for Daydream, they’re not onerous: basically, a post-2015 non rock-bottom (Snapdragon 200/300) SoC, a good screen, and no shortcuts on low-level software/APIs implementations (Vulkan, OpenGL…). It’s not the hardware but the consumer demand, hence the OEM will, that is lacking. Maybe AR/VR is not so much high-end as niche, ie comparable to “rugged”, “huge”, “dual-SIM”, … that comes with its own price premium, as opposed to the laundry list of mainstream features I gave you, which are about as hard to implement but way cheaper.

        Tizen on smartwatches is peripheral(s). Samsung never managed any e+e-ing on actual smartphones. I also do expect them to keep trying. As for succeeding… they have in a way, but by brute-forcing better hardware and specific software (camera…), not by out-maneuvering Google on big picture strategic stuff or even core OS features. We can’t extrapolate ad infinitum from that, but we can’t randomly assume they’ll succeed this time either.

        Edit: as for Apple putting a floor on ARKIt compatibility, that’s demand management, and fragmentation. Fat chance we’ll be treated to much analysis of that ^^

        • Samsung’s strength is “brute force” or the ability to push products into the market despite no clear advantages (or even if they are inferior in software). This is what Google generally lacks (unless they are giving it out for free), and it was this combination of complementary skills that made Android a success in the early 2010s, during the period that Android was clearly inferior to iOS (The early Galaxies during the Android 2.x days).

          I would not discount Samsung’s “brute force”.

          • obarthelemy

            The skill sets needed to release a nice phone and to create an ecosystem are utterly unrelated.

            Not only that, but being a hardware player poisons the relations with the content subsidiaries of other OEMs. Sony Pictures, LeEco (streaming service), Nintendo, … have Android apps because it’s a huge market. A proprietary Samsung platform with a fraction (10 to 30% ?) of that size can probably be safely blackballed, all the more since Samsung will support basic Google content anyway. Samsung would have to create a customer stampede to get traction, needs content to do that, which will come when there are customers…

            The “Samsung will splinter Android” FUD has been making the rounds for years. I’m sure Samsung and all other OEMs would love to do it. I’m fairly sure the reasons they’ve failed up to now hold stronger than ever: ever more numerous interconnected layers/modules/services (OS, security, location/mapping, assistant, AR/VR,… ) increase complexity exponentially. What’s Tizen’s roadmap re: AR/VR ? How successful is Bixby ?

            I’m not especially happy about it, a handful of tech companies are building new-age monopolies based on their stranglehold on data, ability to monetize, and lock-in+network effects. But I think it’s beyond one OEM’s power to change that situation, same as nobody could unseat ATT… Maybe in a few decades, once rot sets in à la Microsoft. We’re clearly not there yet.

          • I hope you understand that I am making some assumptions that you might not agree with in writing this post. I am assuming that AR currently demands a high-end smartphone. I am also assuming that AR is still in its early days, and although low-end Android phones may be able to do the AR of today in a couple year’s time, by then the state of the art will have progressed much farther. That is to say, I am assuming that AR capability will continue to be a differentiator for the mid-term. One way to check whether this assumption holds true or not, is by looking at how quickly Google can expand the range of smartphones that it will support.

            If this assumption does hold true, then we should gauge Samsung’s market power, not by its share of total smartphone sales, but its share of high-end Android sales. According to at least one report, Samsung has more than half of this market, and so it will be much stronger than it has previously been. ( http://www.businessinsider.com/samsung-galaxy-apple-iphone-premium-market-share-chart-2016-12 )

            If AR does not demand a high-end smartphone, and if AR does not evolve to do things that are even more demanding, then AR will come done to mid-range phones and the market will be commoditised. I agree with you on this point.

            Therefore, the important thing is to get some level of understanding of the future trajectory of AR technology. This will help me predict whether my assumptions will hold true or not, at which point I will be able to make predictions based on the arguments that I presented on this post.

          • obarthelemy

            Indeed we disagree on the assumptions. To me, especially looking at Daydream’s requirements and the current $400 price tag for a Daydream phone, AR and VR demonstrably don’t demand high-end phones today, and even the current half-high-endedness can be explained not so much by HW+SW specs requirements as by the very niche aspect of the market, it’s a PR checklist item, not a sales driver.

            Let’s discuss again in a few quarters, we’ll know if AR/VR has become a mass sales driver, a niche market, or just plain bombed. I’d still rather have glasses ^^

      • obarthelemy

        Thinking back on it, on the contrary, I’m wondering if the smartphone market isn’t getting close to a PC-like “commoditization” moment/inflexion point. The new high-end “features” (no jack, face unlock instead of finger unlock, AR/VR, dual cameras that still lag the Galaxy S8’s single one …) feel like inconsequential or even fake stuff.

        The low-end and midrange keep narrowing the gap: they now look cool, run the apps well enough (and often, for longer), and are beyond “good enough”, though not quite “excellent” (so… “really good” ?) on the qualitative stuff (screen, pics/vids, sound, feel…).

        Rather than AR/VR being one more reason to buy a high-end phone, is it not a non-reason on top of other vanishing reasons ? That leaves carrier subsidies, branding, and distinctive looks (quite random and dysfunctional at that, see Samsung slippery round sides… so actually part of branding and nothing more).

        I personally used to buy high-end phones, and still have mild premium envy (that LG V30 !! ;-p). When time comes to pay up, I always run into a) bad/missing features (FM radio, SD, screen too narrow and/or too small, tap-to-wake,…) b) an unjustifiable $400+ premium ($600+ instead of $200) for barely quality-of-life stuff (slightly smaller bezels, slightly better screen and pics) that in the end doesn’t matter.

        • Whether or not AR/VR remains a gimmick that general consumer,era would not appreciate remains to be seen. We have seen significant interest from developers, which is quite positive.

          Then current post is based on the assumption that it does go into the mainstream within a couple of years. If that isn’t the case, I need a different post.

          What I would note however is that the narrative that smartphones have been commoditised is in fact hugely dependent on things like this. The way to do it is to look for promising technologies that require expensive, specialised or bulky hardware, and think what would happen if you could make it work in smartphones. If you can find something, then that’s the route to escape commoditisation. In fact, that’s what happened to our cell phones when they were starting to be commoditised back in the mid-2000s: smartphones came on the stage with the promise of delivering PC-ish tasks on cell phones.

          Compared to that though, this time the new tasks have not been established yet on any platform, so you are right to say that it is riskier for AR than it was back then when phones started to do PC-ish tasks. Nonetheless, what’s happening today has many parallels.

          • obarthelemy

            My point is
            a) the probability of it taking off is low. If dev/OEM interest was a harbinger of success, we’d all be watching 3D movies on our Linux PCs ;-p
            b) if it does take off, the barrier to and cost of implementing it is low, my estimate is it is feasible today in a $150 phone. Even if I’m wrong by a huge $50, it’s still not a game changer re. high-end vs low/mid-range; and in a year or two, we’re back at $150 tops.

          • We’ll know the answer to the first question within this year.
            As for the second, we’ll see how quickly Google can bring this to other devices with less power. If they can start to go below current year flagships within this year, then I think you may be right.

          • obarthelemy

            Even mechanically, today’s $400 phone that supports Daydream will be $200 in 2 years, so we already have the answer, and that’s before economies of scale if mainstreaming does happen.

  • obarthelemy

    Stumbled upon that guy on Medium who has a series of interesting articles: https://medium.com/@mattmiesnieks