Apple Is Not Positioned for Future Endeavors

Apple's new AI partnerships are the right correction—but until they become broadly shipped capability, its closed-system advantage is becoming a speed risk.

Apple's best rebuttal to this article is already visible. Its newest foundation models were built with Google. Its most capable cloud model runs on NVIDIA hardware in Google Cloud. Its developer stack now makes room for Claude and Gemini.

Those are serious moves. They are also evidence that Apple's old answer—own the stack, control the pace, reveal the result when it is ready—was no longer enough.

I do not think Apple is incapable of building good AI. I think the operating model that made Apple exceptional is moving at the wrong speed for this particular market. The company is starting to correct that. The unanswered question is whether it can turn those corrections into ordinary, broadly available user capability before the gap becomes part of what people expect from Apple.

The Siri Apple has promised is not the Siri most people have

On June 8, Apple announced a new Siri AI: more capable, more aware of personal context, able to understand what is on screen, and able to act across apps. That sounds much closer to the kind of agent people now expect.

But as of August 25, that product is not broadly shipping. Apple's own availability notes say the new Siri AI is available for developer testing and is planned as a user beta later in 2026. The distinction matters. A preview can show direction. It cannot close a capability gap in the phone already in my hand.

Apple has earned a reputation for refusing to ship unfinished technology merely to win a news cycle. That discipline still has value. But AI capability does not improve on an annual product-calendar schedule. Models, tools, and agent systems are improving continuously. People who use them every day are not comparing Apple's future Siri with last year's Siri. They are comparing today's Siri with today's frontier systems.

That comparison is rough on Apple.

The closed-system advantage has acquired a speed cost

Apple's integration is real. The hardware, operating system, services, privacy controls, and developer ecosystem can be designed together. When it works, an Apple feature feels less like a third-party attachment and more like part of the device.

Apple also carries nearly all of the coordination cost when it does not work.

In its 2025 Form 10-K, Apple says its markets are defined by rapid technological change and that successful competition depends on a timely flow of new products and technologies. The filing also says Apple designs nearly the entire product solution: hardware, operating systems, applications, and related services.

Put those statements beside each other. End-to-end control gives Apple consistency. It also means that a delay in one layer can hold back the whole experience. In a market where the model layer may improve several times inside one hardware cycle, the closed loop protects the experience only if the loop can absorb outside progress fast enough.

That is the part I do not yet trust.

Apple silicon is an advantage, not a sufficient boundary

The broad claim that the rest of the industry is simply "moving away from Apple chipsets" does not hold up. It is too imprecise to support this argument. Apple silicon is excellent at the job Apple designed it to do: efficient, private, tightly integrated computing across Apple's devices and servers.

The narrower argument is stronger. Apple cannot allow the boundaries of its own silicon strategy to become the boundaries of the intelligence available to its users.

Apple appears to recognize that. Its description of the third generation of Apple Foundation Models says the on-device and several cloud models are optimized for Apple silicon. For AFM 3 Cloud Pro—the model intended for the most demanding reasoning and agentic work—Apple worked with Google and NVIDIA to extend Private Cloud Compute to NVIDIA GPUs in Google Cloud.

That is exactly the kind of infrastructure pragmatism Apple needs. It also reveals the constraint. Privacy-preserving Apple silicon remains valuable, but frontier reasoning cannot be limited to the compute Apple owns, the models Apple trains alone, or the cadence of devices Apple controls end to end.

Apple is opening, but a correction is not a comeback

Apple's partnership story is now much better than the caricature.

The company says its third-generation model family was custom-built with Google. Its new intelligence frameworks let developers choose Claude, Gemini, or another provider that supports Apple's language-model protocol. Xcode brings agents from Anthropic, Google, and OpenAI into the development workflow.

I want to be precise about what that proves. It proves Apple can open important parts of the system. It proves the company sees value in outside models and heterogeneous infrastructure. It does not yet prove that users will receive a coherent, dependable, system-wide AI experience at the pace the market now demands.

Partnership announcements are inputs. The product is the output.

Apple still has to make outside intelligence feel native without hiding it behind a maze of fallbacks, region restrictions, device limits, or future-tense promises. It has to let the best available model handle a task when Apple's own model is not the best fit. It has to preserve privacy without treating privacy as an explanation for weak capability. And it has to update the intelligence layer faster than a traditional operating-system cycle.

Those are execution problems, not branding problems.

A deliberately rough IQ analogy

Here is the bluntest way I experience the current gap: Claude Fable 5 feels like the equivalent of 120–140 IQ, while today's Siri often feels like 65–85 at most.

Those numbers are an analogy, not psychometric measurements. There is no credible, comparable IQ test behind them, and Anthropic's official Fable 5 material reports task benchmarks and product capabilities—not a human IQ score or a controlled comparison with Siri. I am using the ranges to describe the difference between a system that can sustain difficult, tool-using work and an assistant that too often fails to carry an ordinary request across context and action.

The analogy is subjective. The product gap it points toward is observable.

I can give a frontier agent a messy goal, let it inspect relevant files, watch it choose tools, and require it to verify the result. Siri still frequently behaves like a voice interface that can answer, route, or trigger a narrow command. Apple's announced Siri AI is designed to change that. Until it ships broadly and survives ordinary use, it is evidence of intent rather than evidence of parity.

Privacy is the strongest counterargument

Apple's privacy architecture deserves more than a dismissive footnote. On-device processing can keep sensitive context close to the user. Private Cloud Compute is an ambitious attempt to extend those guarantees when a task needs more compute. Deep operating-system integration could give Siri something a general chatbot lacks: safe, structured access to the personal context and applications that make an assistant genuinely useful.

That combination could become Apple's advantage. It is also why I am not arguing that Apple should abandon integration and bolt a generic chatbot onto every screen.

I am arguing for a more permeable system: Apple defines the privacy contract, the permission model, and the user experience, while using the best outside models and infrastructure when they are materially better. Control the standard, not every ingredient.

What would change my conclusion

I will believe Apple is positioned for what comes next when four things are observable rather than announced:

1. Siri AI is broadly available and reliably completes multi-step, cross-app work for ordinary users.
2. Outside models are first-class capabilities, not obscure developer options or emergency fallbacks.
3. Apple's privacy promises survive the use of outside infrastructure without making the useful experience dramatically weaker.
4. The intelligence layer improves on an AI-market cadence instead of waiting for the next annual platform reveal.

Apple has the distribution, hardware, capital, developer ecosystem, and trust to recover. Few companies could make a new capability available across so many daily devices so quickly.

But potential is not position. Right now, Apple is showing that it understands the direction of travel. Its new partnerships are meaningful. Its architecture is becoming more pragmatic. Its future Siri sounds far more capable.

Now Apple has to ship the future before everyone else turns it into the present.