How Apple’s new AI strategy could leave OpenAI and Anthropic exposed

Apple has been widely mocked for being late to the AI party. But its emerging strategy suggests the company may have been playing a longer game: instead of racing to build the biggest model, Apple wants to own the interface layer—the place where billions of people actually access AI.

If that works, companies like OpenAI and Anthropic could find themselves massively exposed, competing for visibility inside Apple’s ecosystem rather than owning the user relationship directly.

Apple’s AI pivot: from model builder to AI gatekeeper

While rivals rushed to launch frontier models, Apple quietly chose a different path. Rather than spending tens of billions on training its own ChatGPT rival, Apple is positioning Siri and iOS as the front door to everyone else’s models.

The new Siri experience is expected to be powered primarily by Google’s Gemini as a foundation model, with the option to hand off specific questions to other providers like ChatGPT or Claude based on user preference. In other words, your iPhone becomes the place where you choose which AI brain you want to use for which task.

This is classic Apple: let others fight the expensive model war, then sit on top as the trusted interface that users actually interact with every day.

Why this leaves OpenAI and Anthropic “massively exposed”

On paper, OpenAI, Anthropic, and others have strong brands and huge user numbers. But Apple has something even more powerful: direct access to 1.5+ billion iPhones and a wider install base of around 2.5 billion devices across its ecosystem.

If Siri becomes the default way people talk to AI on those devices, model providers risk becoming interchangeable back-end utilities. Users might not care whether Siri routes a query to Gemini, ChatGPT, or Claude—as long as the answer is fast and good enough.

That dynamic leaves AI companies exposed in several ways:

• Loss of direct user relationships: Instead of opening the ChatGPT or Claude app, users just say “Hey Siri” and let Apple decide what happens behind the scenes.

• Commoditization of models: If models are swappable inside Siri, it becomes harder for any one provider to stand out or justify premium pricing.

• Platform dependence: If most consumer usage flows through iOS, AI companies become dependent on Apple’s rules, fees, and technical integration—similar to what happened with mobile apps and the App Store.

In that world, Apple doesn’t need to “win” the AI model race. It just needs to be the layer everyone has to go through.

Turning Siri into a revenue engine

Apple’s strategy isn’t just about control—it’s also about turning AI into a new cash cow.

Under the new iOS experience, you’ll likely be able to enhance Siri with your preferred AI service. You could either log in with an existing account (for something like Claude or ChatGPT) or subscribe directly through your iPhone via the App Store.

Whenever that happens, Apple takes a cut of the subscription—typically a monthly percentage for as long as you stay subscribed. Multiply that across millions of users subscribing to multiple AI tools, and you get a serious new revenue stream without Apple having to fund the huge compute costs of running those models itself.

For OpenAI, Anthropic, and others, that means:

• Apple sits between them and the customer.

• Apple takes a share of their revenue.

• Apple controls key discovery and promotion surfaces (Siri suggestions, system prompts, setup flows, etc.).

Owning the interface layer: Apple’s “nine-dimensional chess”

Many critics framed Apple as slow and reactive on AI. But waiting 12–18 months has given Apple a chance to watch how users actually behave, where models fail, and where the real pain points are.

Instead of chasing benchmarks, Apple is focusing on something more practical: making AI feel genuinely useful in everyday life. That means:

• Deep cross-app and cross-device integration.

• A single assistant (Siri) that can talk to your calendar, messages, email, maps, tickets, and preferred AI model.

• Use cases that look more like a real digital assistant—booking tickets, coordinating travel, lining up transport—rather than just answering questions in a chat box.

Apple’s goal isn’t to be the smartest model in the world. It’s to be the most convenient way to use AI without thinking about which model you’re using at all.

How Apple could reclaim its massive install base

Right now, a lot of AI usage happens in standalone apps or web interfaces: you open ChatGPT, Claude, or Gemini, type a prompt, and wait. Apple wants to collapse that friction into a single voice or text interaction with Siri.

With around 1.5 billion iPhones in circulation (and billions of Apple devices overall), Apple doesn’t need to be perfect. It just needs to be:

• Always there (on your phone, laptop, tablet, watch, and in your car via CarPlay).

• Slightly more convenient than opening a separate app.

• Good enough in quality for most everyday tasks.

There’s very little brand loyalty among AI users today. People switch between models freely depending on price, performance, and features. If Siri gives you access to multiple models in one place, many users will simply default to whatever’s easiest—and that favors Apple.

For more on how Apple is baking AI into upcoming devices, you can dive into 15 upcoming Apple products packed with AI and ‘Apple Intelligence’ features.

On-device AI and Apple’s silicon advantage

One underappreciated part of Apple’s strategy is its control over hardware. Apple designs its own chips end-to-end, and its latest silicon is among the best in the market for running AI inference directly on the device.

That matters for a few reasons:

• Speed and responsiveness: Many tasks can be handled locally without sending data to the cloud.

• Privacy: Sensitive data can stay on your device, which fits Apple’s long-standing privacy messaging—even if most users don’t think deeply about it.

• Cost savings: Offloading work to on-device models reduces the cloud compute bill, which is one of the biggest costs in running large-scale AI services.

Other players like Microsoft, Nvidia, Google, and Samsung are racing to push AI-focused chips into laptops and phones, but Apple already has a mature, vertically integrated hardware stack. That gives it a strong base to build AI-first devices and experiences.

New hardware: glasses, Vision Pro, and Siri at the center

Apple’s AI strategy isn’t just about phones. Rumors point to lighter, more everyday wearable devices—like glasses closer to Meta’s Ray-Ban line—sitting alongside or below the high-end Vision Pro.

In that world, Siri becomes the natural front end:

• You talk to Siri through your glasses, watch, or earbuds.

• Siri, in turn, talks to your favorite AI model (OpenAI, Anthropic, Google, etc.).

• The hardware is pure Apple; the intelligence behind it can be swapped in and out.

That’s a tough position for companies like OpenAI, which are exploring their own hardware concepts. As we explored in OpenAI’s rumored AI phone and why Apple should pay attention, building successful consumer hardware is brutally hard—even more so when you’re up against Apple’s design, supply chain, and retail machine.

Will AI agents kill standalone apps?

As AI agents become more capable, there’s a real risk that many apps turn into invisible back-end services. Instead of opening a travel app, you might just ask Siri to “book me tickets for Friday night’s game and sort the trains,” and Siri (plus your chosen model) handles everything via APIs.

That could mean:

• Fewer people opening apps directly.

• Less ad inventory for apps that rely on in-app ads.

• New discovery challenges—users might add “skills” or integrations to their assistant instead of browsing the App Store.

Developers may increasingly optimize for being a Siri-compatible “skill” rather than a standalone app with its own interface. That shifts power even further towards whoever owns the agent layer—in this case, Apple.

The money problem: will people trust AI with their wallets?

One big open question is how far users will actually let AI agents go. Many of the most powerful use cases—booking travel, buying tickets, ordering items—require spending money. Today, most people are understandably wary of letting an AI freely charge their card.

For Apple and others to unlock that next level of automation, a few things would need to happen:

• Clear, strong safeguards and spending limits.

• Transparent logs of what the agent did and why.

• Bank and regulator clarity on liability if something goes wrong.

Until then, it’s likely that users will want to approve purchases manually, even if an agent does all the searching and planning in the background.

Ads, monetization, and the next wave of AI business models

If assistants like Siri become the main way people interact with services, traditional ad models will be under pressure. Apps that used to rely on display ads may see their usage hollowed out as agents interact via APIs instead of user interfaces.

At the same time, AI platforms themselves are already experimenting with new ad formats. ChatGPT, for example, is testing ads inside its own interface. If that proves effective, it’s likely that other AI platforms—and possibly even assistant layers like Siri—will explore similar approaches.

That sets up a complex battle over who controls:

• The user’s attention.

• The recommendation surface (which app, service, or product the assistant suggests).

• The revenue from those recommendations.

Is Apple really “fixing” AI’s current pain points?

It’s hard to know how much of Apple’s AI strategy was deliberate long-term planning versus reactive course correction. The company clearly had internal ambitions around its own “Apple Intelligence” efforts and Siri upgrades that never fully materialized.

But by waiting, Apple has been able to:

• Watch where users struggle with current AI tools.

• See how expensive and complex running frontier models at scale really is.

• Decide to lean into its strengths: hardware, integration, privacy messaging, and ecosystem control.

Whether by design or by luck, Apple now sits in a position where a slow start could turn into a powerful advantage. If Siri becomes the default interface to AI for hundreds of millions of people, then companies like OpenAI and Anthropic may find that the real winner of the AI race isn’t the one with the biggest model—but the one that owns the gateway.