Thursday, January 28, 2021

Apple’s Q1 2021 Results

Apple (Hacker News):

The Company posted all-time record revenue of $111.4 billion, up 21 percent year over year, and quarterly earnings per diluted share of $1.68, up 35 percent. International sales accounted for 64 percent of the quarter’s revenue.

Jason Snell (transcript, also: MacRumors):

Year over year, iPhone revenue was up 17% and set a new record, iPad revenue was up 41% to its best showing in six years, Mac revenue was up 21%, Services revenue was up 24% to a new record, and Wearables revenue was up 30 to a new record%.

Jason Snell:

There have been seven better quarters in the history of the iPad. It’s just that they were all between 2012 and 2015. This most recent iPad quarter was the best since the holiday quarter six years ago.

[…]

Here’s the message: The move to Apple silicon is going to spur Mac growth like never before. Apple’s M1 Macs were incredibly well received, but it’s still just the beginning of the transition. Cook is telling investors, and everyone else, that Apple expects the move to Apple silicon to put its competition in the PC market to shame and fuel a major boost to the Mac.

[…]

The integration of hardware with software is Apple’s secret sauce, or “the magic,” as Cook puts it. But look at the change to that recipe! It’s now the integration of hardware, software, and services.

Previously:

8 Comments RSS · Twitter

Whilst the raison d'etre of a company is to make profit - it will be interesting to see if Apple, which likes to position itself as progressive and socially aware, makes any changes in strategy due it's ever growing cash pile.

Tim Cook said last year, “But I do think, looking back, in the future, you will answer that question: Apple’s most important contribution to mankind has been in health.” When the cost of devices is the main barrier to some people benefiting from that 'contribution' it seems hard to justify such vast profits - especially now that hardware is an entry point to services revenue.

Apple uses privacy as a marketing tool plain and simple. Make everyone else look like information scavengers and Apple look like a savior of privacy. Apple's move to Apple silicon for Mac's only helps solidify Apple's control over all of its hardware. Pretty soon Mac users will probably be forced to install Apps only from their Apps store just like IOS devices do. Obviously no ability to natively run Windows on a Apple silicon Mac. I don't see the performance or efficiency of Apple silicon selling more Mac's. That was the same message Apple said when they moved to Intel from power PC's. In the end its more about what runs on a certain platform then anything. I don't see developers rushing to make apps for Mac's under Apple silicon. They see a bigger opportunity with Windows. Mac's have been stuck at their market share for some time. Obviously makes Apple money but I don't see Mac's becoming more popular.

>Apple uses privacy as a marketing tool plain and simple

I'm not sure if that's a useful criticism. Obviously, corporations do pretty much everything they do for their own benefit. At the same time, it's impossible to know what the motivations of all the individuals involved in these decisions are.

At the end, the question is whether what Apple does with regard to privacy is overall good for us, or bad, and I think it's pretty clear that it is good.

Ironically, although Apple says it cares about privacy, the environment it is creating for developers is quite different.

Apple has been locking down devices (App Store and notarization) which encourages developers to reconsider specialising on Apple devices: Apple can ban or Sherlock their work, and they'll lose all their investment in it. It can ban them, and they'll lose all the work they did learning Apple's proprietary APIs. On the web, they can simply move to another server hosting company.

Apple has moved from open x86 and Objective-C to proprietary "Apple Silicon" and not-as-good Swift. That means learning a whole new stack, incompatible with everyone else's, and wedding oneself even more tightly to Apple. It might be better to learn the Web stack instead, particularly now that WASM is a thing. On the web, your favourite language still works.

Apple has opened the Mac up to iOS apps. The only distinguishing feature of iOS apps is that there are tons of them, few are good, and the prices are so low that it's extremely hard to make money off of them. Most users only actually use Apps like Facebook's which are free and privacy disasters. If you're not the customer, you're the product. Because of low prices, many developers now sell users data to make up the difference. It is Apple which has driven down prices, pushing developers to make their software free, with "in-app purchases". It is Apple that requires a 30% cut for literally doing nothing developers can't do themselves: a Mafia like "service".

And now, Apple Silicon is going to be so much more performant. That means web-apps will be sufficiently fast to do most of what people need. If games can be streamed through a web-browser, many other things can be done that way too.

Since people do not change devices often, and are purposefully being locked in by Apple, I expect captured users won't leave their Apple devices quickly. Instead I expect more software to be offered only as "software as a service" delivered as web-apps. Think Figma instead of Sketch. These apps run on any computing device, decoupling the software maker from caring about Apple. This move has already started with Cordova, Electron, React "native" -- i.e. web-apps installed through the App Store.

In the long term, web-apps mean less privacy for the user. Personal devices will become terminals onto which some computation can be offloaded, but the data will be held on a server. That means less control for users: they won't be able to keep their data confidential, or share it easily to other software, the way one can if it resides on one's own machine. Since all computation occurs on a different machine, users will have no control what appears on their screens, will not know how their data is being monetized, etc. And all that data will be ripe for mining by ISPs, governments, hackers, and the like.

So I foresee a poorer and less safe ecosystem because Apple was greedy. The world, ironically, that Steve Jobs claimed would be a "sweet solution". The surprising twist, where the "Personal Computer" which had so convincingly killed the evil servers of yesteryear, finally dies replaced by the new world of cyberspace. Recall Tron, Neuromancer...

If Apple had really cared about privacy, not lock-in, their software would have been architected very differently and would be a lot more open. For instance, services such as Little Snitch would be encouraged. Building and selling software on one's own device would be encouraged, not milked for service revenue. Sharing data on iOS would be easier. Etc. But that's not the world we live in, and like other creatures, the developers who succeed will be doing what works in the environment that they inhabit.

Apple has moved from open x86 and Objective-C to proprietary “Apple Silicon” and not-as-good Swift. That means learning a whole new stack, incompatible with everyone else’s

Arguably, x86 was just as proprietary (effectively controlled by Intel, with some mutual patent-sharing going on with AMD, but effectively no chance for a third party to join) as ARM is; if you want something non-proprietary, RISC-V might be just around the corner.

Effectively, ARM has more diversity of CPU suppliers than x86. There doesn’t seem to be that much trouble porting stuff that already ran on ARM to compile on macOS for ARM. You can build stuff for a Raspi, and then with few adjustments if any have it compile on macOS as well.

As for the stack, I’m not sure what you mean by this.

It might be better to learn the Web stack instead, particularly now that WASM is a thing. On the web, your favourite language still works.

In practice, only stuff that people are willing to maintain works well. If a community effort wants to maintain Objective-C (I know they are a few, but they never gained that much foothold), they can do that.

Yeah, you could conceivably write WASM in Objective-C, but frankly, that sounds painfully esoteric.

If Apple had really cared about privacy, not lock-in, their software would have been architected very differently and would be a lot more open. For instance, services such as Little Snitch would be encouraged. Building and selling software on one’s own device would be encouraged, not milked for service revenue. Sharing data on iOS would be easier. Etc.

I don’t think privacy and lock-in have to be at odds, but yes, there are areas where Apple could stand to be a bit more open again.

Old Unix Geek

@Sören

Arguably, x86 was just as proprietary (effectively controlled by Intel, with some mutual patent-sharing going on with AMD, but effectively no chance for a third party to join)

Tell that to IBM, Cyrix, Centaur, Via, Rise, National Semiconductor, NexGen, Transmeta, Texas Instruments, NEC, SIS, Nvidia and many others. Also x64, the 64-bit version of x86, was invented by AMD. At the time Intel was distracted by Itanium and EPIC. So Intel ended up licensing it from AMD.

M1 relies on proprietary instructions to do fast math. That means your ARM SIMD code won't run as fast as Apple's. Either you accept this, or you accept the lock-in of reverse engineering and using M1's new math-matrix instructions, or the lock-in of using Apple's proprietary frameworks. Similarly with the Machine Learning APIs. Over time I expect the trend of using proprietary accelerators to continue: transistors are cheap, but transistors doing work are expensive since they cost power.

As to your Raspberry Pi point, not really: although a Raspberry PI 4 can run 64 bit code, most Linux support for RPIs is still 32 bit, which means that code will be compiled against a 32 bit target on Raspberry Pi, not a 64 bit one as on MacOS... Also, Linux and MacOS differ significantly as far as APIs go, so more than "a few adjustments" will be needed.

Objective-C is, as you say, unfortunately basically dying if not dead. If you want to get onto the corporate merry-go-round a second time around with Swift and M1, be my guest. As for me, I'm seriously contemplating getting off here, and bidding a sad farewell to Objective-C. I joined the Apple bandwagon because I was amazed by the first NeXT computers. But now that technology is dying, and I'm tired of relying on APIS, languages, etc that change every time a large corporation loses interest. It's a bit of a shame really, since I've spent the last 16 years on and off writing quite a bit of Mac/iOS stuff (both code and words). But there was a life before Apple, and there will be a life after it.

Finally, unlike you, I believe privacy and lock-in are at odds pretty much by definition. Lock-in means people are stuck with whatever protections Apple gives them, and cannot improve on them. That means a monoculture. Monocultures are much easier to hack: with a single zero-day, one gets access to most of Apple's devices in a particular category. Lock-in means you are trusting Apple 100% to look after your privacy for you. In my book, sharing my privacy with 150,000 Apple employees is not actual privacy. Think for instance of the fact iCloud backups are unencrypted.

Tschüß!

Tell that to IBM, Cyrix, Centaur, Via, Rise, National Semiconductor, NexGen, Transmeta, Texas Instruments, NEC, SIS, Nvidia and many others.

How many of those designed a x86 CPU in the past decade?

(And yes, I’m sure Nvidia has a thing or two to say about being bullied out of x86.)

Also x64, the 64-bit version of x86, was invented by AMD.

Which is why I said “some mutual patent-sharing going on with AMD”. Which you even quoted.

M1 relies on proprietary instructions to do fast math. That means your ARM SIMD code won’t run as fast as Apple’s. Either you accept this, or you accept the lock-in of reverse engineering and using M1’s new math-matrix instructions, or the lock-in of using Apple’s proprietary frameworks. Similarly with the Machine Learning APIs. Over time I expect the trend of using proprietary accelerators to continue: transistors are cheap, but transistors doing work are expensive since they cost power.

Yes, fair enough.

How many of those designed a x86 CPU in the past decade?

Elbrus, VIA and Zhaoxin.

x86's are now sufficiently complex that it takes serious money to make a competitive offering... which is why fewer companies design them than previously. Government support now seems commonplace, which I ascribe to the costs and risks involved. I should note that it is also true that few companies design their own ARM chips; most simply license a design from ARM.

Just missing your deadline, in 2008, the last startup that tried to design an x86 was Montalvo Systems but, AFAIK, it ran out of money and it was acquired by Sun.

The Elbrus-8S (Russian) provides x86 compatibility through binary translation to a VLIW architecture (very much like Transmeta). It was designed in 2014 to run Windows for the Russian Government. Since it runs an entire OS natively, rather than single apps, I would not compare it to Rosetta. They've just taped out their next generation Elbrus-16. Since it seems to be pretty much the same architecture, but with 16 cores, I assume, but do not know, that it will also run x86 software.

VIA/Centaur are still designing native x86 CPUs. In the past they also licensed their designs to Chinese Zhaoxin. However Zhaoxin's ZX-D, ZX-E and ZX-F seem to be home-grown architectures. Zhaoxin has the support of the Chinese government.

Leave a Comment