Skip to content

Nate B Jones AI Apple Business Hardware

Apple WWDC 2026: The AI Story Everyone is Missing

Nate B Jones reads Apple's WWDC 2026 announcements as one bet wearing three costumes: a new Siri AI, a Google Gemini model alliance, and Private Cloud Compute expanding onto Google Cloud and Nvidia GPUs. The thread tying them together is a single question he thinks decides the first AI trillionaire, namely where AI does your work when it works all day: a chatbot tab, a giant cloud, or the device you already own. Apple's answer is device first with private cloud for overflow, with the OS, apps, App Intents, and personal context as the surface AI sees and touches. Jones argues Apple is deliberately commoditizing the model to own the trusted action surface, the place a billion people touch AI through, and the trust that comes with it.

Published Jun 11, 2026 18:34 video 25 min read Added Jun 16, 2026 Open on YouTube →

At a glance

Apple used WWDC 2026 to ship three headlines that look unrelated, and Nate B Jones argues they are one bet wearing three costumes. Headline one is a new Siri and a new Apple Intelligence. Headline two confirms that the next generation of Apple foundation models was built with Google using Gemini family tech. Headline three expands Private Cloud Compute beyond Apple's own data centers into Google Cloud running Nvidia GPUs. Jones reads all three as Apple answering a single question that he thinks decides who becomes the first trillionaire in AI: when AI does real work for you all day, where does that work actually run? In a chatbot tab, in a giant cloud that burns tokens and power, or inside the computer you already bought? Apple's answer is the device first, then private cloud behind it, with the operating system, the apps, and your personal context as the surface AI sees and touches. The model, he says, is being commoditized on purpose. The prize Apple is chasing is ownership of the place a billion people touch AI through.

The three headlines hiding one bet

Jones opens by laying the three announcements side by side so you can see the trick. Apple showed Siri AI. Apple confirmed the Google Gemini alliance underneath part of its model stack. Apple expanded Private Cloud Compute into Google Cloud on Nvidia GPUs. Three separate headlines, he says, except they are not. They are Apple answering the question that may decide who becomes the first trillionaire in AI history.

The question is concrete. When AI starts doing real work for you all day long, where does that work run? Does it run in a chatbot tab? Does it run in a giant cloud service where every serious task burns tokens and GPU time and power and data center capacity? Or does more of it start inside the computer you already bought: the iPhone in your hand, the Mac on your desk, the chip inside that Mac, the operating system, the apps, the files, the photos, the messages, the screen, and then private cloud behind it only when the device is not enough? That, he says, is Apple's vision, and that is the WWDC story.

Sure, the feature list is long: Siri AI, Apple Intelligence, Gemini family tech, App Intents, foundation models, visual intelligence. But the list is not the story. The story is that Apple is trying to turn AI from something you rent in the cloud into something built into the computer you bought. That is why this WWDC matters, he says, because it hands you a clean map to the future without forcing you to become a model provider spreadsheet. So stop asking which model is ahead or whether Siri is finally catching up. Ask who owns the place where AI sees your work, touches your apps, remembers your context, and does something.

He frames the stakes by audience. If you run a team, this changes the budget conversation: it is no longer just should we buy ChatGPT or Claude or Gemini, it is where does our work live, which systems can AI safely touch, what has to stay private, and who gets permission to act. If you build software, it changes the product conversation: bolting a chatbot onto your app is not enough, the app has to become legible to the operating system. And if your day is drowning in email and tabs and files and passwords and copy and paste, that is exactly the fight Apple picked at WWDC.

Figure 1. The framing of the whole video. Three announcements that read like separate news items, a smarter Siri, a Gemini supply deal, and Private Cloud Compute renting Google Cloud and Nvidia GPUs, collapse into a single strategic question. Jones says it is not who has the best model but who owns the surface where AI sees your work, touches your apps, and is allowed to act.

Every AI announcement, mapped

Before the strategy, Jones lays out the actual list so nothing is lost. Apple announced the next version of Apple Intelligence, a new Siri AI, new Apple Foundation Models, new on-device models, new server models running through Private Cloud Compute, Google Gemini family tech underneath part of the model stack, and the Private Cloud Compute expansion into Google Cloud with Nvidia GPUs. On the developer side: App Intents becoming much more important to Siri and Apple Intelligence, Foundation Models as a developer framework, a Core AI path so developers can run local models in apps using Apple silicon, and Xcode agents with model choice. Add natural language Shortcuts, Safari features that can watch web pages and organize tabs, and Passwords doing more adjunctive work to fix weak accounts. Then the normal WWDC platform stuff: new OS versions, design refinements, performance and search improvements, betas now and final releases in the fall.

So many announcements that it is easy to lose track, he says, but the AI ones all point the same way. Apple is trying to make AI part of the computer again. Not a separate chat tab, not a model picker, not a cloud product you visit. The computer. The thing in your hand, the apps you use, the files, the photos, the messages, the passwords, the calendar, the browser, the screen context where your life already is. That is a very different strategy than build the best chatbot, and Jones thinks that is exactly where most of the coverage misses the point.

Siri is the face, not the strategy

The easy WWDC question is did Siri get smarter, and he grants it is a fair one. Siri matters, because normal people will judge Apple AI through Siri, and if Siri feels dumb, as it has, the whole thing feels dumb. But Siri is not the AI strategy. Siri is just the face. The better question is what Siri now sits on top of, and the answer is where it gets interesting.

Siri sits on top of personal context, screen awareness, app actions, the Spotlight semantic index, Apple Foundation Models, Private Cloud Compute, and App Intents. That is not a voice assistant story. That is Apple trying to make the operating system itself feel agentic, and Jones means something deliberately simple by that. Can the system see your screen? Can it understand your files and photos? Can it talk to the apps where your work actually happens? Can it take action without spraying your life into a random cloud service?

That is the product, he says, not a benchmark and not a flashy demo. The real product is whether your computer can finally take the hint: find the thing, move the file, watch the page, build the shortcut, draft the message in the app where it will actually be sent. This is what ordinary consumers want. It is what his aunt wants. She does not want to manage six models or think about tokens or context windows or local versus cloud. She wants the computer to do the right thing and not leak her life in the process. And that, he argues, is a better problem for Apple to own than beating OpenAI at frontier model speed.

Here is the load bearing claim. Apple has not had the best assistant. Apple has not had the best model. It does not have those today. But you do not need to be the best frontier lab if you own the place where personal AI becomes useful. The AI industry trained everyone to think the model is the product. For most consumer AI, Jones says the product is the model plus context plus permissions plus interface plus actions plus trust, delivered as one package you believe in. That bundle is what Apple is trying to own.

Figure 2. Why "did Siri get smarter" is the wrong question. Siri is a thin face on top of a deep stack. Underneath sit personal context, screen awareness, app actions, the Spotlight semantic index, App Intents, the on-device and server foundation models, and Private Cloud Compute at the base. Jones's point is that Apple is not shipping a better voice assistant, it is trying to make the operating system itself agentic.

App Intents and apps the OS can call

The developer story is the least sexy part of the keynote and, Jones argues, maybe the most important. App Intents is how developers make an app's contents and actions available to the system. In plain English, it is how an app tells Apple Intelligence here is what I have, here is what the user can do with it, and here are the actions you are allowed to take. That matters because an AI assistant that cannot act inside apps is not really an assistant. It can suggest, advise, write, and summarize, but all the actual work lives somewhere else.

Then the historical irony. Apple spent the last 15 to 20 years teaching every company on Earth to become an app, and you still see it now: people are vibe coding, the number of apps in the App Store is skyrocketing, but usage is not. Apple knows that. So now it is trying to teach the operating system to do the thing the app used to do. Jones calls it Apple trying to self-destruct. The old world was open the app, learn the interface, tap around, do the thing. The agentic OS world is ask the system, and the system uses the app for you. That is Apple moving past the app world it built.

But it cannot move too far, because the app ecosystem is also the tollbooth. So App Intents is the compromise architecture. Apple gets to make apps callable by the OS while keeping the app, the developer relationship, the permission layer, the App Store and all that money, and distribution inside Apple's orbit. That is why this is not Apple killing apps. It is Apple turning apps into things the operating system can call.

And that changes what developers should care about. For the last couple of years, a lot of AI product strategy has been you have to at least add a chatbot. Jones says he has literally sat in rooms where people say, at a minimum we have to do that, and so everything gets AI-washed: put AI in the headline, write a press release, done. That, he says, is lazy, and Apple's version is the opposite of lazy. It is structural. If the OS is becoming an AI surface, your app has to become legible to the OS. So your data model matters, your permissions matter, your actions matter, and your integration with Spotlight, Siri, and Shortcuts matters a lot. The winning apps may not be the ones with the flashiest chatbot. They may be the apps whose data and actions are clean enough that Apple Intelligence can actually use them. Not exciting to demo, he admits, but enormously important in practice.

Foundation Models matter for the same reason. Apple is opening model access through a native Swift framework: on-device Apple models, Private Cloud Compute models, and other providers that conform to that framework. Apple is not saying we built every model and you should only use us. Apple is saying we want to own the native model interface on Apple platforms. It is a different kind of control. Core AI matters because it gives developers a path to run other local models on Apple silicon. Xcode agents matter because Apple is pushing the same agent story into the developer workflow itself. None of this is a Siri story. It is Apple trying to make AI native across every surface at once: consumer, developer, app, and device. The deeper read, Jones says, is that Apple is turning everything in the system into a pipeline that enables an AI layer for consumers over the existing Apple OS, and betting that is enough to self-disrupt. He admits he does not know whether it is.

The Google and Nvidia twist

That uncertainty leads straight to the Google Gemini piece. A lot of people will treat it as humiliating for Apple, and at some level it is, Jones concedes. Apple would obviously prefer the clean mythology: our hardware, our software, our chips, our models, our magic. Instead the story is that the next generation of Apple Foundation Models was built in collaboration with Google using Gemini family tech. That is significant.

But the cheap take, that Apple failed and had to use Google, is weaker than it sounds. The stronger take is that Apple may not care who supplies raw model capability, because model capability is commoditizing. Apple wants to own the layer the user touches: the device, the OS, the app platform, the permission prompts, the Siri surface. So let Google provide the model capability and let Nvidia provide the private cloud infrastructure. Apple still wants to own the experience, and it is betting it can. The line that anchors it: you can source model capability, but you cannot easily source a billion devices, a mature operating system, a developer ecosystem, and the trust people place in the computer they carry around.

Private Cloud Compute is where the argument gets more complex. The simple version of the Apple hardware thesis is that inference moves off the cloud and onto the device: Apple silicon with unified memory and neural engines, local models, all fixed cost. That is still part of the story. But WWDC made clear the larger version is device plus private cloud. Apple's pitch is run what you can on the device, and when the request is too hard, trust us, we will route it to Private Cloud Compute. And now Private Cloud Compute is expanding beyond Apple's own data centers into Google Cloud on Nvidia GPUs for the really hard workloads, including agentic tool use and complicated reasoning.

That puts Jensen Huang in an interesting position. Nvidia may still be inside the infrastructure, Google may still supply the model capability, the cloud may still handle the hardest workloads, but Apple wants to be the front door and stay the front door. The device decides what runs locally. The OS decides when context is available. The app layer exposes the actions. Private Cloud Compute is relegated to handling overflow.

Figure 3. The device plus private cloud routing Apple described. The user only ever touches the device and OS, which Apple owns outright. The default is to run work locally on Apple silicon at fixed cost. Only the hardest requests, agentic tool use and heavy reasoning, fall through to Private Cloud Compute, which now leans on Google Cloud and Nvidia GPUs. Apple rents the expensive bottom layer while keeping the surface, and the trust, at the top.

The two bottlenecks and the trillionaire question

Here Jones names the real prize. AI has at least two major bottlenecks. One is raw compute: GPUs, power, data centers, networking, memory bandwidth, all the things Nvidia is incredible at. That bottleneck is very real. The other is the trusted action surface: where the AI meets the user, where it touches apps, where it gets permission to act. That bottleneck is real too, and it is the one Apple is trying to own.

And that, he says, is a trillionaire question. Not who has the cleanest demo, not who has the best frontier model, but who owns the default meter for everyday intelligence. If the future of AI is mostly bigger models and bigger data centers, Jensen wins and keeps winning, and Nvidia becomes the tax collector on intelligence. That is one path. But if a huge amount of useful personal AI happens through the device and the operating system, the economics get a lot more complicated for Jensen. The device becomes the default, the cloud becomes a specialist, and the thing in your pocket becomes part of the larger AI compute experience. That is a very different world.

If Apple pulls it off, it changes who gets paid at scale. Nvidia still wins in frontier training, enterprise inference, robotics, scientific computing, and data centers, and that build-out is real. But Apple can shift a meaningful part of the consumer AI value chain toward hardware it sells, software controls it owns, and services it can meter or bundle through iCloud and the App Store. That is why this is trillionaire level territory. The first trillionaire is probably not the person with the smartest model. It is the person who owns the meter when intelligence becomes economically unavoidable. Maybe that is Jensen, because every path runs through GPUs. Maybe it is Apple, because personal AI becomes native to Apple devices and Apple turns the iPhone upgrade cycle into the AI upgrade cycle. Maybe it is both.

Bottleneck	Raw compute (Nvidia's game)	Trusted action surface (Apple's game)
What it is	GPUs, power, data centers, bandwidth	where AI meets you, touches apps, gets permission
Who is strong	Nvidia, the hyperscalers	Apple owns device, OS, apps, trust
If this wins	Nvidia is the tax collector on intelligence cloud	device is default, cloud is a specialist device
Apple's move	rent it: Google Cloud + Nvidia GPUs for overflow	own it: the front door for personal AI
The prize	own the default meter for everyday intelligence, the trillionaire question

Figure 4. The two bottlenecks Jones says the AI race is really fought over. One is raw compute, where Nvidia and the hyperscalers dominate and Apple is content to rent. The other is the trusted action surface, the place AI is allowed to see your work and act on it, where Apple already owns the device, the OS, the apps, and the trust. Whoever owns the default meter when intelligence becomes unavoidable is, in his read, the first trillionaire candidate.

What it means if you actually use this stuff

Jones closes by translating the strategy into stakes for real people. WWDC, he says, is trying to give us a flashlight through the fog, clarity about what ordinary folks around the Thanksgiving dinner table will talk about as AI. If you want to get ahead of it, watch the surfaces: device surfaces, OS surfaces, browser and search surfaces, how files are handled. That tells you more about where personal AI is going than another leaderboard argument or another WWDC press release.

If you already use ChatGPT, as nearly a billion people do, or Gemini, and you are trying to turn it into real work, the takeaway is blunt. Apple is not all the way there. But WWDC is about building the rails for a default experience. Because if your day is full of context switching, email and Slack and documents and tabs, going back and forth, the boring features that make life feel seamless are the ones that matter. The value of AI is not I wrote a paragraph. The value is getting more work done with less context switching, less handoff, fewer administrative papercuts. The computer notices the page changed. The password was weak, so the computer fixed it. The shortcut gets built in plain English so you can see what the computer is doing. None of this is AGI, he stresses. It is just the machine becoming less useless at the work in front of you.

Then the part he flags as the real moat: Apple's whole product claim is that the computer can now know a lot about you without making you feel stripped mined for data. And he asks directly, do you trust OpenAI with that? Do you trust Anthropic like that? People answer differently, he says, but that trust lane is the lane, and it gets more valuable as AI agents start touching more of your work.

For builders, the future of an app on Apple platforms is not can I get it launched and approved. It is can I expose the actions, clean up the permissions, make the workflow safe, expose the objects so App Intents work. The apps that win are not the ones with the flashiest demos. They are the ones whose data and actions are clean enough for the operating system to operate them. And from a brand perspective, the question becomes are you stuck enough in people's heads that they ask for you by name when they talk to Apple.

So the surface story is Google supplying the AI features behind Siri. The bigger story is Apple trying to turn the iPhone, Mac, and iPad into the default place where personal AI runs, sees, decides, and acts. And if it is default for consumers, it may become default for workers, because we all bring our own devices everywhere. The question then becomes: if it is that seamless on Apple, will you start demanding that seamlessness at work? That, he says, is the play. Not Apple built the smartest model, not Apple killed Nvidia, not everything runs locally. The play is Apple wants to own the computer where personal AI becomes useful, and by extension the computer where AI is valuable.

If that works, the AI race stops being only about who has the biggest cloud cluster. It becomes about who owns the trust in the system and the surface that agents work against. That is why Jensen should be watching, and why the other major AI players should have been paying attention. Because the first trillionaire, Jones predicts, will not be decided by who IPOs this summer. The more useful question is who owns the surface a billion people touch AI through. Apple has a path to that, it is building that path, and WWDC exposed the roadmap. Pay attention, he says, or you will get distracted by tomorrow's frontier model headline and the next day's and the next day's. Apple wants to be synonymous with AI for a billion people, and if it is, the rest of the race changes entirely, because Apple will have won the last mile that drives actual trust.

Key takeaways

The three WWDC headlines, new Siri AI, the Google Gemini model alliance, and Private Cloud Compute expanding onto Google Cloud and Nvidia GPUs, are one bet: own the surface where personal AI runs and acts, not the model.
Apple's strategy is to make AI part of the computer again. Not a chat tab or a model picker, but the device, OS, apps, files, photos, messages, and screen context where your life already is.
Siri is the face, not the strategy. What matters is the stack underneath it: personal context, screen awareness, app actions, the Spotlight semantic index, App Intents, foundation models, and Private Cloud Compute.
You do not need the best frontier model if you own the place personal AI becomes useful. For consumers, the product is the model plus context plus permissions plus interface plus actions plus trust, as one package.
App Intents is the compromise architecture: it makes apps callable by the OS while keeping the App Store, the developer relationship, and the permission layer intact. Apple is turning apps into things the OS can call, not killing them.
The winning apps will be the ones whose data and actions are clean enough for Apple Intelligence to operate, not the ones with the flashiest chatbot.
Foundation Models, Core AI, and Xcode agents extend the same move across developer and device surfaces. Apple wants to own the native model interface, not necessarily build every model.
The Google Gemini deal looks humiliating but reads as deliberate commoditization. Apple will source model capability from Google and infrastructure from Nvidia while keeping the experience and the trust.
Apple's compute model is device first, private cloud for overflow. Private Cloud Compute now rents Google Cloud and Nvidia GPUs for the hardest reasoning and agentic work, while Apple stays the front door.
AI has two bottlenecks: raw compute, where Nvidia wins, and the trusted action surface, where Apple is fighting. Whoever owns the default meter for everyday intelligence is the first trillionaire candidate.
The practical advice: watch the surfaces (device, OS, browser, search, files), not the leaderboards. The value of AI is less context switching and fewer papercuts, and the moat is trusting a computer that knows you without strip mining you.

Chapters

Timestamps are clickable. Click one and the player jumps there and keeps playing while you read.

0:00 The three headlines hiding one bet
2:26 Every AI announcement, mapped
6:04 App Intents and apps the OS can call
9:40 Moving inference from cloud to device
16:18 Owning the default surface for personal AI

Notable quotes

That sounds like it's three separate headlines, doesn't it? But I think it is Apple answering the question that may decide who becomes the first trillionaire in AI history. Nate B Jones, 0:08

Apple is trying to turn AI from something you rent in the cloud into something built into the computer you bought. Nate B Jones, 0:45

Siri's not the AI strategy for Apple. Siri's just the face. The better question is what is Siri sitting on top of? Nate B Jones, 2:09

You don't need to be the best frontier lab if you own the place where personal AI becomes useful. Nate B Jones, 2:51

This is not actually Apple killing apps. It is Apple turning apps into things the operating system can call. Nate B Jones, 3:42

Apple is not saying we built every model and you should only use us. Apple is saying we want to own the native model interface on Apple platforms. Nate B Jones, 4:28

You can source model capability. You cannot easily source a billion devices, a mature operating system, a developer ecosystem, and the trust people have in the computer they carry around. Nate B Jones, 5:19

The other bottleneck is the trusted action surface. Where does the AI meet the user? Where does the AI get permission to act? And that is a trillionaire question. Nate B Jones, 11:42

The first trillionaire is probably not just the person with the smartest model. It is the person who owns the meter when intelligence becomes economically unavoidable. Nate B Jones, 13:18

Apple wants to own the computer where personal AI becomes useful and by extension Apple wants to own the computer where AI is valuable. Nate B Jones, 16:40

Resources mentioned

Nate B Jones, the channel (AI News and Strategy Daily), for sober, non binary reads on AI markets and adoption.
Apple WWDC 2026, the developer conference where every announcement in this video was made.
Apple Intelligence, Apple's personal intelligence system and the umbrella for the AI features discussed.
Siri, the assistant Jones calls the face, sitting on top of the new agentic stack.
App Intents, the framework that lets the OS call an app's contents and actions, the architectural core of the strategy.
Foundation Models framework, the native Swift interface to on-device, Private Cloud Compute, and third party models.
Private Cloud Compute, Apple's private server inference, now expanding onto external infrastructure.
Apple silicon, the on-device compute with unified memory and neural engines that anchors the local-first thesis.
Xcode and Swift, the developer workflow and language where Apple is pushing model choice and agents.
Spotlight, Safari, Shortcuts, and Passwords, the system surfaces being wired into Apple Intelligence.
Google and Gemini, the model partner whose family tech underpins the next Apple Foundation Models.
Google Cloud and Nvidia, the external infrastructure and GPUs now powering Private Cloud Compute overflow, with Jensen Huang the figure whose position the video weighs.
iCloud and the App Store, the services through which Apple can meter or bundle AI value.
OpenAI / ChatGPT and Anthropic / Claude, the frontier labs Apple is contrasted against on model capability and, crucially, on trust.

Where it stands

This is a strategy read, not a product review, and Jones is careful to say he does not know whether Apple's self-disruption works. A few honest footnotes. The strongest idea, that value migrates from a commoditizing input (the model) to the surface built around it (the device, OS, and trusted action layer), is a durable pattern in technology and fits cleanly here, and the "trusted action surface" framing is the real contribution. The softer parts are the leaps. Casting the Gemini deal as deliberate rather than forced is a generous interpretation that the public facts do not settle either way, since a supply partnership can be both a strategic choice and a sign of a model gap. The trillionaire framing is rhetorical scaffolding, not a forecast, and Jones treats it as such. The claim most exposed to being wrong is the implicit bet that consumers will route serious AI work through an Apple OS layer rather than through the chatbots they already open by reflex, which is precisely the open question the video is built to flag. And the on-device thesis has a quiet tension he names himself: if the hardest work overflows to Private Cloud Compute running on Google Cloud and Nvidia, then "it runs on the computer you bought" is true for the easy cases and aspirational for the hard ones. The framing holds; the timeline and the win do not yet.

Full transcript

Apple used WWDC to show Siri AI, confirmed the Google Gemini Apple alliance story, and to expand private cloud compute into Google Cloud using Nvidia GPUs. And that sounds like it's three separate headlines, doesn't it? But I think it is Apple answering the question that may decide who becomes the first trillionaire in AI history. When AI starts doing real work for you all day long, where does that work run? Does it run in a chatbot tab? Does it run in a giant cloud service where every serious task burns tokens and GPU time and power and data center capacity? Or does more of it start inside the computer you already bought? The iPhone in your hand, the Mac on your desk, the chip inside that Mac, the operating system, the apps, the files, the photos, the messages, the screen, and then private cloud behind it when the device is not enough. That is Apple's vision. That is the WWDC story. Sure, the feature list is Siri AI and Apple Intelligence and Google Gemini family tech and app intents and foundation models and visual intelligence. The list goes on. But the story is that Apple is trying to turn AI from something you rent in the cloud into something built into the computer you bought. And that is why this WWDC matters. Because if you were trying to understand AI without turning your life into a model provider spreadsheet, this gives you a clean map to the future. So, stop asking which AI model is ahead or if Siri is catching up. Ask who owns the place where AI sees your work and touches your apps and remembers your context and does something. If you run a team, this changes the budget conversation. It is not just should we buy ChatGPT or should we buy Claude or Gemini? It's where does our work live? Which systems can AI safely touch? What has to stay private? Who gets permission to act? Those are the larger questions for 2026. If you build software, this changes the product conversation. It's not enough to bolt AI onto the app. Obviously, the app has to be legible to the operating system. That's the new challenge, and that's the challenge Apple is going after with WWDC. If you're drowning in email and tabs and files and passwords and copying and pasting, then this is the part that matters. That's the fight Apple picked at WWDC. So, this is the story of what Apple announced, what Siri now sits on top of, why the developer layer matters, why Google and Nvidia make the story more interesting, and what this changes if you are trying to make AI useful in actual work. Let's start with the announcements. Apple announced the next version of Apple Intelligence, new Siri AI, new Apple Foundation models, new on-device models, new server models running through private cloud compute, Google Gemini family tech underneath part of the model stack, and a private cloud compute expansion into Google Cloud with Nvidia GPUs. They also announced App Intents becoming much more important to Siri and Apple Intelligence, Foundation Models as a developer framework, Core AI so developers can run local models in apps using Apple silicon, and Xcode agents with model choice. Natural language shortcuts, Safari features that can watch web pages and organize tabs, passwords doing more adjunctive work to fix your weak accounts, and then they did the normal WWDC platform stuff, too. New OS versions, design refinements, performance improvements, search improvements, betas now, final releases in the fall. So, yes, so many announcements it's easy to lose track of it all. But the AI announcements all point in the same direction. Apple is trying to make AI part of the computer again. Not a separate chat tab, not a model picker, not some cloud product that you visit. The computer. That's Apple's vision. The thing in your hand, the apps you use, the files, the photos, the messages, the passwords, the calendar, the browser, the screen context where your life already is. That's the AI. And that's a very different strategy than build the best chatbot. And that is where most of the coverage seems to be missing the point. Because the easy WWDC question is did Siri get smarter? It's a fair question. Siri matters. Normal people are going to judge Apple AI through Siri. If Siri feels dumb as it has, the whole thing feels dumb. But Siri's not the AI strategy for Apple. Siri's just the face. The better question is what is Siri sitting on top of? And the answer gets really interesting. Siri now sits on top of personal context and screen awareness and app actions and spotlight semantic index and Apple foundation models and private cloud compute and app intents. And that's not just a voice assistant story. That's Apple trying to make the operating system itself feel agentic. And by that I mean something really simple. Can the system see your screen? Can it understand your files and photos? Can it talk to the apps where your work happens? Can it take action and do it without spraying your life into a random cloud service? That is the product they're trying to build. It's not some benchmark they're hitting. It's not a demo they want to make flashy even though WWDC has demos. The real product is whether your computer can finally take the hint. Can it find the thing? Can it move the file? Can it watch the page? Can it build the shortcut? Can it draft the message in the app where it will actually be sent? This is what ordinary consumers want. It's what my aunt wants. They don't want to manage six models or think about tokens or context windows or local versus cloud. None of that matters. They want the computer to do the right thing and not leak their life in the process. Apple's always been good about it. That is an Apple problem. It's a better problem for Apple than beating OpenAI at frontier model speed. Look, Apple has not had the best assistant. Apple has not had the best model. They don't have those today. But you don't need to be the best frontier lab if you own the place where personal AI becomes useful. And that's what's going to matter to my aunt and my cousins and people who are not deep in AI. The AI industry has trained us to think the model is the product. But for most consumer AI, I think the product is the model plus context plus permissions and interface and actions and trust in a package you believe in. That's what Apple is trying to own. And that brings us to the developer story, which is probably the least sexy part of the keynote and maybe the most important one, too. App Intents is how developers make their apps' contents and actions available to the system. In plain English, it is how an app tells Apple Intelligence, here's what I have, here's what the user can do with it, and here are the actions you are allowed to take. That matters because an AI assistant that cannot act inside apps is not really an assistant, is it? It can just suggest things, advise on things, write things, summarize things, but all the work lives elsewhere. Apple spent the last 15 or 20 years teaching every company on Earth to become an app. And you see that even now when people are vibe coding and the number of apps in the App Store is skyrocketing. But usage isn't. Apple knows that. Now it's trying to teach the operating system to do the thing the app used to do. Apple's trying to self-destruct. The old world was open the app, learn the interface, tap around, do the thing. The agentic OS world is ask the system and the system uses the app for you. That is Apple trying to move past the app world it built. But the catch is this, it can't move too far because the app ecosystem is also the tollbooth. So App Intents is the compromise architecture. Apple gets to make apps callable by the OS while keeping the app, the developer relationship, the permission layer, the App Store with all that money, and the distribution inside Apple's gray. That is why this is not actually Apple killing apps. It is Apple turning apps into things the operating system can call. And this changes what developers ought to care about. For the last couple of years, a lot of AI product strategy has been you have to at least add a chatbot. And if you think I'm making that up, I have sat in rooms where people say, well, at a minimum we have to do that. And it really is something people have done, and everything gets AI-washed. Now we can put AI in the headline, we can write a press release, we're done. So much of that is lazy. The Apple version of AI here is a lot more than a lazy press release. Apple is getting very structural with their approach to AI. If the OS is becoming an AI surface, your app has to become legible to the OS. So, for developers, your data model matters, your permissions matter, your actions matter, your integration with Spotlight and Siri and shortcuts matters a lot. The winning apps might not be the apps with the flashiest chatbot. They might be the apps whose data and actions are clean enough that Apple Intelligence can actually use them. And that is not super exciting to demo, but it's super important in practice. And foundation models matter for the same reason. Apple is opening model access through a native Swift framework. On-device Apple models, private cloud compute models, and other providers that conform to that framework. Apple is not saying we built every model and you should only use us. Apple is saying we want to own the native model interface on Apple platforms. It's a different kind of control. Core AI matters because it gives developers a path to run other local models using Apple silicon. Xcode agents matter because Apple's pushing the same agent story into the developer workflow itself. So, this is not a story about Siri. It is Apple trying to make AI native across the platform, consumer surface, developer surface, app surface, device surface. The deeper WWDC story is that Apple is turning everything in the system into a pipeline to enable an AI layer for consumers over the existing Apple OS. And they're betting that that's enough to self-disrupt. I don't know if it is. And the question around whether Apple is doing enough brings us to the Google Gemini piece of this story. A lot of people are going to treat this as humiliating for Apple. At some level, it is. Apple would obviously prefer the clean mythology, the clean story. Our hardware, our software, our chips, our models, our magic. Instead, the story is the next generation of Apple foundation models was built in collaboration with Google using Gemini family tech. That's significant. But the cheap take is not as strong as it sounds. The cheap take is Apple failed and had to use Google. I think the stronger take is Apple may not care who supplies raw model capability. It's commoditizing. Apple wants to own the layer the user touches. Who owns the device, the OS, the app platform, the permission prompts, the Siri surface. Apple wants to own that. So, yeah, they'll let Google provide the model capability. Nvidia can provide the private cloud infrastructure. Apple still wants to own the experience and they're betting they can. You can source model capability. You cannot easily source a billion devices, a mature operating system, a developer ecosystem, and the trust people have in the computer they carry around. Now, private cloud compute is where that argument gets a little bit more complex. Follow me here. The super easy version is the Apple hardware thesis is local AI inference moves off the cloud, it moves onto the device. You have Apple silicon with unified memory and neural engines. You have local models, it's all fixed cost. That is still a part of the story. But WWDC made it clear that the larger version of that story is really a device plus private cloud. Apple is saying run what you can on the device and when the request is too difficult, trust us, we'll route it to private cloud compute. And now private cloud compute is expanding beyond Apple's own data centers into Google Cloud using Nvidia GPUs for really hard workloads, including agentic tool use and complicated reasoning. This puts Jensen in a very interesting position. Nvidia may still be inside the infrastructure, Google may still supply the model capability, the cloud may still handle the hardest workloads, but Apple wants to be the front door and stay the front door. The device is going to decide what runs locally. The OS is going to decide when context is available, the app layer exposes the actions, private cloud compute is relegated to handling overflow. Apple is going for one of the two core bottlenecks in AI here. In AI we have at least two major bottlenecks. One bottleneck, of course, is raw compute. GPUs and power in data centers and networking and memory bandwidth, all that stuff that Nvidia is incredible at. And that's a very real bottleneck. The other bottleneck is the trusted action surface. That's what's getting fought over now. Where does the AI meet the user? Where does the AI touch apps? Where does the AI get permission to act? And that bottleneck is real, too, and Apple's trying to own that one. And that is a trillionaire question, people. It's not who has the cleanest demo, it's not who has the best frontier model, it's who owns the default meter for everyday intelligence. If the future of AI is mostly bigger models and bigger data centers, Jensen wins and keeps on winning. Nvidia becomes the tax collector on intelligence. That is just one path. But if a huge amount of useful personal AI happens through the device and operating system, the economics get a lot more complicated for Jensen. The device becomes the default, the cloud becomes a specialist, the thing in your pocket becomes just part of the larger AI compute experience, and that's a very different world. And if Apple pulls that off, it changes who gets paid at scale. Sure, Nvidia still wins in frontier training and enterprise inference and robotics and scientific computing and data centers, and that build-out is real, but Apple can shift a meaningful part of the consumer AI value chain toward hardware it sells, toward software controls, and toward services it can meter or bundle through iCloud and the App Store. This is why we're talking about trillionaire level territory here. Because the first trillionaire is probably not just the person with the smartest model. It is the person who owns the meter when intelligence becomes economically unavoidable. Maybe that's Jensen because every path runs through GPUs. Maybe that's Apple because personal AI becomes native to Apple devices and Apple turns the iPhone upgrade cycle into the AI upgrade cycle. And maybe it's both. WWDC is trying to give us a flashlight to get through that fog. It's trying to give us clarity around what ordinary folks around the Thanksgiving dinner table are going to talk about as AI. And if you want to get ahead of that story, I would encourage you to watch the surfaces. Watch device surfaces, OS surfaces, browser and search surfaces, how files are handled. This tells you more about where personal AI is going than another leaderboard argument or even than the press releases around WWDC. If you already use ChatGPT as a billion people do just about, or Gemini, and you're trying to turn that into real work, the takeaway is even more blunt. Look, Apple is not all the way there. But WWDC is about building the rails to make that kind of default experience possible. Because if your day is full of context switching, if it's full of email and Slack and documents and tabs and you're always going back and forth, those boring features that make life feel seamless with AI really matter. Because really the value of AI is not I wrote a paragraph, the value is intelligence that lets you get more work done with less context switching, less handoff, fewer administrative papercuts. The computer just notices that the page changed. The password is weak, so the computer fixed it. The shortcut gets built in plain English so you can understand what the computer is doing. None of this is AGI. It is just the machine becoming less useless at the work sitting in front of you. Apple's whole product claim is the claim that the computer can now know a lot about you without making you feel like you're being stripped mined for data. And I got to ask, do you think OpenAI is someone you trust with that? Is Anthropic a company you trust like that? I don't know. People tend to have different answers to those questions. But that's the lane we're talking about. And that trust lane is extremely valuable as AI agents start touching more and more of our work. And if you're building software in this space, if you're building in the Apple ecosystem anywhere, the future of an app on Apple platforms is not going to be can I get this app launched and get it approved? It's going to be can I expose the actions? Can I clean up permissions? Can I make the workflow safe? Can I expose the objects so app intents work? The apps that win are not the ones with the flashiest demos. It's going to be the apps that have data and actions clean enough for the operating system to actually operate them. And the question from a brand perspective is are you stuck enough in people's heads that they will actually ask for you by name when they talk to Apple. So sure, the surface WWDC story is about Google supplying the AI features in Siri AI. The bigger story is that Apple is trying to turn the iPhone and Mac and iPad into the default place for consumers where personal AI runs and sees and decides and acts. And by the way, if it's default for consumers, it may well become default for workers because we all bring our own devices everywhere. And so the question is going to become if it's that seamless on Apple, are you going to start demanding that kind of seamlessness at work? That's the play. It's not Apple built the smartest model. It's not Apple killed Nvidia. It's not everything runs locally. The play is Apple wants to own the computer where personal AI becomes useful and by extension Apple wants to own the computer where AI is valuable. And if that works, the AI race stops being only about who has the biggest cloud cluster. It's about who owns the trust in the system. Who owns the surface that agents work against. And that is why Jensen should be watching. And frankly, the other major AI players should have been paying attention here. Because I don't think the first trillionaire is going to get decided by who IPOs this summer. I think it's going to be more useful to ask ourselves, if you want to create lasting wealth from AI, who is going to own the surface that a billion people touch AI through? Apple has a path to that. Apple is building that path. WWDC exposed for all of us the roadmap they're using to build that path. Pay attention. Because if you don't, you're going to get distracted by the headlines tomorrow and the next day and the next day about frontier model this and that. Apple wants to be synonymous with AI for a billion people. If they are, the rest of the AI race is going to change entirely because they will have won the last mile that drives actual trust. So, let me know what you think in the comments and I'll see you next time.