In partnership with

In today’s email

• Breaking down what happened in this week’s September event

• How Apple reframed the AI race and stepped on its competition

MADE POSSIBLE BY

Get software delivered with financial guarantees, focusing on your goals

With ELEKS' product-oriented delivery, we guarantee that your software vision is realised in a superior solution implemented within your timeline or budget constraints. We prioritise your success and focus on maximising your product's business value.

Our team provides industry-leading expertise across your entire SDLC and takes full responsibility for the implementation roadmap, budget, quality metrics, and process setup, ensuring your strategic goals are achieved.

Schedule your consultation

Hi reader.

As you may (or may not) have noticed, I haven’t written anything in the last couple weeks. I appreciate your patience; I’ve been swamped with many things, among them being a detailed research piece on the investability of AI agent infrastructure in cahoots with a top growth-stage investor and multiple-exit founder.

I can’t wait to share this with you all. For now, let’s dig into Monday’s Apple event.

WHAT HAPPENED

The September event

This Monday, Apple held their annual September event. Unlike the flagship summer WWDC conference wherein software takes center stage, Apple uses September events to release new iPhones.

The most exciting WWDCs unveil new Macs, while a worthy September event showcases iPhones that eclipse their predecessors in design, performance, efficiency, and features.

Apple will often spice things up, sprinkling in new subsidiary product updates throughout both events (iPad, AirPods, Apple TV, subscriptions, etc.).

At the event

I miss the magic of live, in-person Apple events. There’s something to watching Craig Federighi beam with excitement on stage while the crowd oohs and aahs. When COVID bludgeoned the life out of in-person events, Apple began releasing perfect, pre-recorded, highly-edited videos to replace the unpredictability of live demos and the anxiety of speaking live to millions.

This year’s September event began with an accessibility showcase. Apple highlighted the efficacy of their health-focused features, from iPhone’s car crash detection automatically paging first responders to Apple Watch’s EKG sensor detecting heartbeat abnormalities.

They’ve invested heavily into clinical studies to improve these features, namely for hearing and for the heart. They’ve uncovered a data goldmine for training models and building features that deeply affect consumers.

Thanks to a comprehensive otic study, Apple revamped AirPods to serve as FDA-approved hearing aids. Over a billion people suffer from degraded hearing, while Apple has sold hundreds of millions of AirPods in the last 8 years.

It’s a fantastic idea with top-drawer execution. To power noise cancellation, AirPods sample their surroundings 48,000 times per second. This, I’d say, is exactly what makes AirPods effective substitutes for in-ear, clinically prescribed hearing aids.

After a few more anecdotes of life-saving predictive features, Apple moved into their product releases. Most Apple enthusiasts agree that the product lineup upgrades were nothing to write home about.

• AirPods 4

  • Gets many features previously exclusive to the “pro” tier AirPods Pro 2

  • Standardized USB-C connectivity

• Apple Watch Series 10

  • Wider, larger display with a thinner casing and lighter housing

• iPhone 16 & 16 Pro

  • A new side haptic button (a fake button that feels real) to control the camera

  • A “big boost” in battery life

  • New chips (A18 and A18 Pro) with better performance and efficiency in the graphics and machine learning departments (more on this later)

  • Improved camera quality and precision

• A few new colors here and there

As a fanboy who could watch an Apple event daily, this 98-minute spectacle could’ve been presented in 15.

With that said, there were a few key disclosures particularly related to AI software, chips, and data integration that were generally overlooked by the tech ecosystem, but of pivotal importance to consumers, to their competitors, and to the modern AI arms race.

THE APPLE FLYWHEEL

So you forgot I’m king

How much of your life is on your iPhone? How about your Mac? My answer is “almost all of it.”

The core pillars of my digital life—texts, emails, documents, and workspaces—live on my iPhone and Mac. All the rest lives in apps, with which Apple ensures their software is deeply connected.

The seamlessness of the Apple ecosystem is second to none. Quick-switch FaceTimes and AirPods handoffs are “simple” examples of integration… In reality, Apple devices are interconnected down to the bits pulsing between transistors.

Apple controls the entire stack. They own the hardware, the software, the integration and distribution on third-party apps, the on-device data, and the consumer (the blue bubble still reigns supreme).

In the modern AI arms race, we’re consumed by our desire for new models. We define winners and losers by who creates these models (OpenAI, Anthropic, Meta, etc.) and by who infrastructs them (Nvidia).

We focus less on who can best leverage an AI system with their existing products, ecosystem, and distribution.

My thesis: Apple is generally understood to be a laggard in the modern AI race. I agree with this take, but only through the lens of how we currently value existing AI behemoths (more models, more features). But Apple has decided to avoid that game entirely. Don’t sleep on their ability to innovate with targeted airstrikes as opposed to scorching the Earth.

Here’s how I see them doing it.

The chip

It all starts with the chip.

In 2020, Apple made an announcement that shocked the world. They decided to phase out of a relationship with Intel, who’d previously supplied Apple with chips for 15 years.

Intel is the king of the CPU. Their flagship i3, i5, i7, and i9 lineups have provided millions of devices with performant central processing. They’ve always struggled with graphics, a critical faction of computing useful for video editing, gaming, crypto mining, and training/running machine learning models. (Shoutout to anyone else who suffered with the MacBook Pro’s Intel Iris 650 excuse of a graphics unit.)

Apple began designing their own SoCs (silicon-on-chip), signing a partnership with Arm lasting through 2040. Arm specializes in efficient and customizable chipsets. (Remember when Nvidia tried buying them for $40 billion?)

These chips are “unified,” meaning that on the same housing, they provide a device with CPU cores, graphics cores, and memory. Because all components are on the same chip, they are extremely fast.

The fact that these chips are designed by Apple is key. They are built solely for Apple software, Apple-compatible apps, Apple hardware, and Apple data.

Video editing with Apple’s Final Cut Pro on a Mac with an Apple SoC is a buttery, blissful experience.

And yet, Apple didn’t compromise on compatibility, developing Rosetta 2, an impressive under-the-hood translator that allows Intel-based applications to still run on the new Apple SoCs. In other words, most users never notice the difference while still enjoying the benefits.

So now, central processing and graphics are not only covered, but better than everyone in the industry.

But what about machine learning?

Neural behavior

Apple’s chips have long had “Neural Engines” for on-device machine learning. I’ve always felt these were over-powerful and underutilized.

It’s well known that Apple prefers to move slower than the competition but outclass them on quality and reliability. Still, the most exciting applications of on-device neural processing have come slower than expected.

I believe Apple kicked off this September event with a focus on accessibility to call attention to their computational cohesiveness. They ran the studies, they have the data, they built the chip, they own the hardware, and they control the software. Everything is built to work perfectly with everything.

It’s showboating.

Now, we’re seeing some impressive utilizations of this power and interconnection.

Of course, 24/7 automatic crash detection, fall detection, sleep monitoring, heartbeat monitoring, and hearing hazard prevention are impressive and worthy uses of the Neural Engine. My favorite application, though, is Siri, powered by a beautifully mind-blowing revamp of the cloud.

The architecture

My biggest takeaway from this week’s 98-minute event came in just 10 seconds—Private Cloud. This summer, Apple brought their ecosystem-wide privacy focus to datacenters, built with their own chips. You read that correctly—an Apple cloud and datacenter powered by Apple chips. (Talk about controlling the entire stack.)

Apple Private Cloud Compute (PCC) is built to extend the processing power of Apple devices while ensuring that personal user data is not accessible to anyone other than the user—not even Apple. This property is independently verifiable.

So, when your iPhone needs a computational boost to run a language model or process a horde of data, PCC is on standby. These jobs are processed by Apple chips in one of Apple’s datacenters, providing the same performance, privacy, and efficiency benefits as when Apple replaced on-device Intel chips with their own.

How is nobody talking about this?

Siri

When she was first unveiled, Siri was a feat of engineering. But while she could hold basic conversations, she always struggled to understand a user’s context and make effective decisions. (”Here’s what I found on the web for do I have new emails.”)

Now, thanks to Apple PCC, on-device and PCC-based language models, the interconnection of apps, and control over every layer of the stack, Siri has massively leveled up.

She’s aware of texts, photos, emails, documents, and more, all with the Apple promise of privacy (scary, yes, but highly-scrutinized and verifiable). She can handle requests like “book an Uber to the airport when my mom’s flight arrives,” searching for your mom’s flight information, tracking it, and scheduling an Uber (if enabled). And all of this comes with conversational awareness, meaning she understands language and no longer requires specific wording to reach an outcome.

On-device awareness and user-focused intuition are incredibly exciting, truly useful applications of AI. And because Apple controls everything in their expansive ecosystem—”my way or the highway”—they can facilitate this vision better than anyone else.

• Because Apple controls on-device data, they can encode and query it better than anyone. This allows Apple Intelligence to find specific moments in hidden videos, ex. “show me when I pocketed the winning 8-ball in last year’s pool tournament.”

• Because they tightly control the software, they can easily interface with on-device operations (settings, etc.) and apps. Despite having strict software requirements, Apple has always provided developers with SDKs (integration access). Almost every service, from iMessage to speech recognition to sensors, is connectable. Apple would allow external apps (i.e. Uber, in the earlier example) to facilitate a contextually-aware connection to Siri’s intelligence.

• Because they control the hardware, using these new features is a seamless experience. Apple’s devices arguably have the best distribution of any high-end product in history. No stupid “AI pins” required.

• Because they control the chip, their models run blazingly fast and efficiently, and are encoded to work perfectly with the software they’re used in.

• Because of PCC, they can extend the abilities of on-device processing infinitely, running large language models and data-intensive jobs securely and efficiently in the cloud. And because PCC also uses Apple chips, they have the same interconnection, performance, and efficiency benefits as on-device SoCs.

Granted, I’ve always been an Apple fanboy, but as a follower of trends in AI infrastructure and consumer applications of agents, I’m ecstatic to see Apple play the long game. They understand that as inference costs trend to zero, the best business model is providing true value to their end users. And given their deep control over every single layer of the stack, from app to data to software to hardware to chip to cloud, they’re in the perfect position to execute.

ABOUT RTF

Return the Fund 🚀 

• One startup a week poised for 10x growth; market deep dives with actionable insights for builders and investors.

• Technical breakdowns of advanced new tech to empower informed decision-making

• Find your next prospect, product, job, customer, or partner. 🤝 Written by undercover pioneers of the field; trusted by builders and investors from Silicon Valley to NYC. 💸 

Last week, I covered my learnings from a summer spent in Seattle.

As you know, we’re hell-bent on uncovering future unicorns cruising under the radar. Preeminent companies are lean, quiet, and driven before reaching their watershed moments. By the time people start talking about them, it’s too late.

You won’t get editions from us very often. Weekly at best. Two reasons:

1. We’re balancing full-time VC/startup work with Return the Fund.

2. We prioritize depth, insight, and value. This is not a daily news publication… We hope that when you do get an email from us, it’s dense, valuable, actionable, and worth saving.

Thanks for reading today’s RTF. Feel free to reach out to us at [email protected]. 🤝 

Psst: None of our company picks are ever sponsored. All research and opinions are completely held by the Return the Fund team.