How Apple Can Own I/O to Own the Universe

Plus! The Great Inflation; Two-Way API Businesses; Zoom and Cities; From ETrade to eToro; Hardware-as-a-Service

How Apple Can Own I/O to Own the Universe

This piece is a guest post from Evan J. Zimmerman. Follow him on Twitter and Medium.


Another year, another iPhone.

The question many, including Apple, are asking themselves is what’s next. What comes after the iPhone is the biggest question facing Apple and arguably the entire computing industry. While we’ve gotten rumors of what the answer might be—a new AR headset, perhaps?—at Apple’s most recent launch event we got another smartphone that seemed in many ways to be a continuation, not a revolution.

Plus, of course, Apple kept the lightning port, which has been a source of angst for reviewers for years.

This huge question and small question seem unrelated, but they’re actually intimately connected in Apple’s strategy for the new computing paradigm. One way to frame tech companies’ efforts to commoditize the complement is that the winners want to own the meta, and leave the object-level to others. In apps, Apple wants to define what kinds of models are acceptable, to offer a highly-taxed platform on which particular apps can compete—and then to outsource to developers the task of figuring out which apps consumers will pay for. Earlier, Microsoft and Intel effectively outsourced PC development, and Amazon’s third-party market is, indirectly, a way for Amazon to harvest the results of every merchant’s R&D and marketing. An important part of the way Apple does this today is to create not only a software but also a hardware ecosystem mediated by its tight control of the software/hardware interface; today, this integration is considered one of the defining characteristics of Apple’s business model.


While many people are familiar with Moore’s Law, far fewer have heard of Bell’s Law. Roughly put, if computer speeds double every 2 years, then every decade computers will be 32x more powerful. A 10x improvement indicates a paradigm shift, and so in 1972 Gordon Bell at DEC predicted that, every 10 years, there would be a new computing paradigm. A 32x improvement is roughly the increment after which some specific task goes from barely possible to completely trivial, at which point it’s time to build something that’s newly barely possible. The result is a new kind of computing platform that also introduces a host of new capabilities and contexts.

And so far, he’s been right. Bell formulated his law during the rise of the mini computer. A decade later, right on time, Apple would play a key role in the transition to the post-minicomputer paradigm. And ever since, Apple has ridden each successive wave of computing to the next wave of Bell’s Law.

An underappreciated aspect of Bell’s Law is that new computing also means a new explosion of device forms and data, which would imply that all parts of the von Neumann architecture should change. And while it’s been less studied, it’s true. Each new computing paradigm has come with a new way to transfer data, or I/O. Workstations used tape, but personal computers used floppy disks; laptops used discs and USB; and personal computing devices like smartphones used standardized alternatives. Call it Bell’s I/O.


Apple somewhat stands apart. As is its wont, Apple has vertically integrated its I/O strategy too. While Apple didn’t own floppy discs, in the 1990s it tried to push its own alternative to USB, called Firewire. In the 2000s Apple abandoned Firewire, but it found a new light by riding the wave of personalized computing with the iPod. Instead of using a standard port, it used a completely proprietary port called the 32-pin connector. Apple used this same connector for its iPhone until smartphones became physically too small and it introduced what we now know as Lightning. These new I/O ports are extremely proprietary. Apple even patents them!

In the background, Apple has turned its Bell’s I/O strategy into a monetization scheme. Because it patents its I/O ports, you can only create compatible tools by getting a license. This infrastructure is called MFi (originally, “Made for iPod”). This is especially huge because Apple’s proprietary ports are the only way to connect to the iPhone with the elimination of the headphone jack.

MFi serves three core strategic objectives for Apple:

  1. Primarily, MFi gives Apple control over its ecosystem, not just devices. This is especially important for quality, since Apple is famously the "it just works" company.
  2. Competitive knowledge. Through license fees, Apple knows everyone who makes an accessory, or who plans to, in advance of when they start producing. And thanks to unit-based licensing, Apple also knows roughly how well they are doing.
  3. Incremental revenue. At $4/device, this is not insignificant at Apple's scale. Plus proprietary ports means proprietary adaptors, which Apple famously overcharges for.

So what will the new computing paradigm be? The industry seems to have converged on what Rick Osterloh at Google called “ambient computing,” which is a bigger version of IoT: pretty much everything will be a computer, and pretty much everything will be connected all the time in a way that doesn’t necessarily require human intervention. Unlike all previous computing paradigms, ambient computing won’t have a standard form factor—in fact, it’s defined by the lack of form factor. It’s like a different state of matter, with the solid discrete computers sublimating into a fog of ubiquitous computing.

If Bell’s I/O holds, we're in for a new port, which means a new proprietary connection from Apple.

The problem? MFi is a program for licensing physical port patents. And if the future is ambient computing, physical ports will look like this:

retail box

That just won’t work for most form factors. You can’t expect your phone to plug in to your smart fridge or your smart toaster.

At the same time, wireless connection is difficult. It can be unreliable, slow, and is hard to use. Not great for the “it just works” company.

Enter Apple Silicon.


In 2016, Apple released the Apple Airpods. They have gone on to be a monster hit, selling $6 billion in 2019 alone. People shell out hundreds of dollars, primarily, because when you connect Airpods they do this:

As Steve Jobs would say, because it “just works.”

But why do the Airpods “just work?” It’s because Apple Silicon has secretly become Apple’s new port.

Airpods are powered by special wireless chips: the W1 for regular Airpods and the H1 for Airpod Pros. These headphone chips combine Bluetooth and Wi-Fi to solve the biggest problem people had with wireless cans: to make a connection that is extremely fast and stable. It also combines NFC and custom software for easy pairing. In other words, Apple has solved the problem everyone has with not just headphones but any wireless connection.

But weaving these things together isn’t easy. Until now, it’s only been available on the iPhone as the Airdrop feature. To make this work on less powerful machines, Apple needed to create a custom chip to run these functions, which is what the W1 and H1 are. Amazingly, Apple replaced a dumb wireless standard with a whole mini computer.

There is some similarity to Microsoft’s “embrace, extend” strategy, since the Apple wireless approach is based on taking open standards and packaging them, similar to what Microsoft did with countless software projects from browsers to Java. It also explains why it started as Airdrop, since smartphones are also just computers themselves. It also guarantees backwards and forward compatibility with any device Apple ships in the future.

Apple appears to be doing something similar with its App Clips product:

The release of Apple’s wireless approach is so perfectly timed that Apple must be thinking about the wireless ecosystem. Airpods and the W1 were released the exact year that wireless headphones became more popular than wired ones. As wireless replaces wired for almost all connections, it seems obvious that putting Apple chips in every pair of headphones would be a no-brainer. It would make every wireless headphone in the world work better with the iPhone, building unprecedented ecosystem lock-in.

Today, only Apple devices use any of these chips.


The successor to Apple’s MFi program, we’d argue, should be Apple Silicon Wireless.

There are several approaches Apple could take, but perhaps the one that makes the most sense is selling chips made with its own designs while also taking a patent fee. Manufacturers would register with MFi, maintain certain standards, and get an instruction set they could use to make wireless everything that works better with iPhone. Imagine, for example, if Apple’s new car key function was even more secure because it used an Apple Silicon design made for cars (we’ll call it the C1). Why should smart locks and smart lightbulbs use their own bridges like zWave instead of just connecting through a trusted bridge made using custom Apple Silicon?

And with Apple looking to diversify its revenue, why not get further into the chip game? Apple could plausibly pull it off because it has a huge ecosystem. JAMF, which merely helps companies manage Apple devices, is a public company!

This would be advantageous for Apple’s strategy for several reasons:

  1. The future is wireless, and not just with headphones. So to keep the same level of quality control and visibility into its ecosystem, Apple needs to have some control over wireless connection, not just wired.
  2. The number of accessories that will connect wirelessly will vastly exceed all electronics in the past combined. So it increases the Apple ecosystem size dramatically.
  3. It is a meaningful advance that creates lock-in. Wireless is difficult—Apple’s wireless chips are a real technical improvement. It also makes devices easier to set up with Apple devices, which makes everyone’s whole life easier by using Apple products, which is a compounding lock-in: every incremental Apple-compatible purchase makes more sense than the last one, and as Apple’s footprint grows, it makes sense for more niche smart home devices to choose Apple compatibility. Yet, this involves mostly open standards, so it shouldn’t be too scary to Apple’s competitor-customers. Think about iMessage: it uses SMS, but it also has iPhone-to-iPhone specific features that drive device lock-in.
  4. It will help shore Apple's position up in some of its weaker areas. Imagine fitness connectors, smart home devices, and much more using Apple Silicon as a hook into Healthkit or Homekit. Plus it would make smart appliances and similar products easier to use, which could reduce churn for OEMs among Apple users, who are usually very valuable.
  5. It is chips, so it is a further way to expand Apple Silicon by driving economies of scope.
  6. Apple can probably get more than $4 per chip, which is more than MFi; plus most of the users of a potential Apple Silicon Wireless are currently paying Apple nothing.

To top it all off, there’s a precedent for the model of monetizing a change in computing form factor by selling chips and collecting royalties: Apple’s nemesis, Qualcomm! Qualcomm is also a fabless chip company, filling orders through a pure play foundry like TSMC. And every time Qualcomm sells you a chip it also makes you pay for a patent license. In fact, according to their 2019 10-K, it’s 40% of their revenue and 54% of their profits.

So why hasn’t Apple done this already? It’s possible that they would find this too difficult or early for technical reasons that are not publicly known—perhaps each chip needs to be too custom for each use case right now to be worth the trouble. They might not feel that their designs are mature enough yet, or they might not want to reveal what they see as trade secrets, though this could be solved by just selling black-box chips and licensing patents, which are already public. The most likely answer, however, is probably that they are just too vertical right now and want to keep the revenue of easy Apple headphones to themselves.


But that would be ignoring the lesson of its own history. Apple will never build everything, and it’s never expected itself to. That is why Apple has a worldwide developer conference every year. And it is also why it has gone to such great lengths to control the primary I/O into its devices.

So let Apple sell its own headphones. But with Apple Silicon Wireless, Apple is aligned with its customers. Setting up and managing wireless devices that work with our central brain, our smartphones, is terrible. And as that becomes our main way of connecting to our ever-expanding computer ecosystem, Apple is losing its proprietary connection to those devices, and hence its control.

This makes it likely that Apple Silicon Wireless will become available to many manufacturers soon. Apple’s ecosystem needs it.

This piece is a guest post from Evan J. Zimmerman. Follow him on Twitter and Medium.


The Great Inflation

An interesting twitter thread breaking down the early stages of the 1960s-1970s inflationary cycle. Inflation was a concern before the oil crisis—a very common one, because there wasn't a good monocausal explanation. One cause of inflation was demographics (reflected, in this case, in higher clothing prices; this was a time when US apparel manufacturing was moving south for cheaper labor, but wages around the country were rising) and because 60s-era health programs took a long time to ramp up. "Inflation" is an aggregate phenomenon, but even when there's a high-level cause like growing deficits and elevated borrowing, the effects show up in different industries at different times.

Two-Way API Businesses

There's a broad category of companies that represent literal or figurative APIs that help other companies plug a new feature into their products. Twilio is a literal example; use their API, and your product can now interact with customers through text, phone, email, and messaging. And DoNotPay is a more figurative case study, wrapping paperwork- or phone call-intensive tasks into an app. Sivo, the "Stripe for debt" is interesting because it's a two-way example of this. Sivo lets companies add financial services to their products; they cite the example of helping Uber offer drivers vehicle loans and advances. But Sivo itself is converting its customers into a ready source of easily-acquired customers, which is generally the limiting factor for lenders. From Uber’s perspective, Sivo is an API for adding features; from Sivo’s perspective, Uber is an API for getting leads. The model of bootstrapping on someone else's customer acquisition has worked very well for Afterpay and Affirm, and it makes sense to apply it outside of consumer lending.

Zoom and Cities: The Five-Century-Old Case Study

Anton Howes has a piece looking at how improved sailing technology might explain the growth of London in the sixteenth and seventeenth centuries. As he notes, the city grew by nearly 10x in that period, an extreme change that demands an explanation. Faster ships (with better guns) were a major part of this. What's interesting is that cheaper transportation's first-order effect is to make geography matter less; if there's something that can be built more cost-effectively outside of London, the city doesn’t have to make it But the second-order effect of that is to make city-level agglomeration effects more important; when more can get outsourced, the city can focus even more on what it does best. The comparison to telework is imperfect in many ways, but it's an interesting possibility: technologies that make people more productive no matter where they are can have the most outsized impact on people's productivity when they're in the same place as all the other high-productivity people.

From ETrade to eToro

Marc Rubinstein has a great piece comparing the public market debuts of E*Trade and eToro. eToro figured out that there's a social component to risk-taking. The two things traders want are:

  1. The ability to make any bet, on any asset, with any amount of leverage, and
  2. Outside guidance telling them exactly what to trade and when.

eToro accomplishes the latter by letting traders follow other traders and clone their transactions on the site. One trader, for example, has users with $130m in assets following his trades. This is a very interesting model since it meshes typical consumer-facing approaches—get users to generate content other users will enjoy—with an exaggerated version of the tendency for institutional investors to trade ideas and end up with very similar-looking portfolios.


Smartphone companies in India are selling phones on credit, and using apps to disable phone features when buyers don't pay. This is getting attention because some phone buyers resell their phones without disclosing the payment plan, but the model makes sense; it's a way to charge based on usage, and a way to automatically collateralize a loan without depriving the user of access to their collateral. They can also make collections a more gradual process:

Within three to four months, [the app will] have gathered enough information about these people to figure out what’s the risk profile...Datacultr uses a laundry list of techniques to force borrowers into paying. The app starts by sending audiovisual prompts in regional languages as reminders. If the user misses their first repayment, it forcefully changes the wallpaper on their cellphones...Juriasingani said that, of all of Datacultr’s nudges, forced wallpaper changes are the most likely to result in a payment. Replacing wallpaper images of family or friends with messages like, “Your EMI is due” makes the reminder difficult to ignore. “We see almost 50% of these people pay within three days of just putting up the wallpaper,” he explained, “and almost 70% in seven days.”