Today is the day the handoff happens. John Ternus takes the CEO seat at Apple, and Tim Cook — the man who turned a struggling computer maker into a four-trillion-dollar machine — steps up to executive chairman. It is only the third time in thirty years that the top job at Apple has changed hands, and the stakes are different this time. Cook inherited a company that needed supply-chain discipline. Ternus inherits a company that needs to figure out AI before the industry leaves it behind.

I have spent my career in server rooms and data centers, watching platform shifts happen in slow motion and then all at once. Apple's AI problem is not a product problem. It is a control problem — the same kind of control problem every infrastructure operator eventually hits. Apple does not own the models it depends on, does not own the infrastructure those models run on, and is only now waking up to what that actually means. This is the story of what Ternus is walking into, and why the next eighteen months will tell us whether Apple remains a platform company or becomes a hardware supplier to someone else's platform.

What Actually Changes Today at Apple

Ternus, 50, is a twenty-five-year Apple veteran who spent his career on the hardware side: iPhone, iPad, MacBook, AirPods, Apple Watch, and the Vision Pro. He was the engineer who championed the shift from Intel chips to Apple silicon — the most consequential infrastructure decision the company made in a decade. That move gave Apple control over the silicon in its devices, and it is the template for what Ternus now needs to do with AI.

Cook's numbers are the backdrop for every comparison. Since he took over in 2011, Apple's share price is up more than 2,000 percent. The company has sold more than 3.1 billion iPhones, and annual profit grew fourfold. Apple became the first public company to hit a trillion-dollar valuation in 2018 and now sits at four trillion. But the same era carried a persistent criticism: the product line stayed largely static, the Vision Pro failed to connect with buyers, and Apple's AI strategy was defined by what it licensed from others rather than what it built.

The AI Problem Ternus Inherits: Renting Intelligence

Here is the uncomfortable truth nobody at Apple likes to say out loud: Apple does not run a foundational AI model of its own. It does not train frontier models. It does not operate the kind of GPU fleets that Microsoft, Google, Amazon, and Meta pour hundreds of billions of dollars into every year. Instead, Apple rents intelligence.

Apple Intelligence — the feature set that includes image generation, text rewriting, notification summaries, and the long-delayed Siri upgrade — runs on Google's Gemini models. ChatGPT is integrated into the iPhone experience through a 2024 deal with OpenAI. In other words, the two most important AI capabilities on the most successful consumer device in history are supplied by Apple's direct competitors. ChatGPT and Anthropic's Claude rank first and second among free iOS apps. Gemini sits fourth. Apple's own AI products do not appear at all. Every time a user pays for a premium tier of one of those services, Apple takes a revenue cut — profiting, as one analyst put it, from competitors filling the gap Apple left open. That is not a strategy. That is a tax on someone else's strategy.

The OpenAI Deal That Turned Sour

The partnership path has not exactly been smooth. The 2024 deal to integrate ChatGPT into the iPhone led to a bitter dispute between the two companies, and earlier this year Apple sued OpenAI, alleging the AI company stole its trade secrets. You do not need to know who is right to understand what it signals: Apple's relationship with the companies powering its AI features is adversarial, and it is getting worse.

Meanwhile, Apple announced a deal with Google to use Gemini models for its own AI and the new AI-powered Siri. That puts Apple in the position of paying its biggest search rival to supply the brain of its virtual assistant — the same assistant that was supposed to be Apple's answer to the AI moment. IDC's Francisco Jeronimo put the risk plainly: Apple could become less relevant in a world where OpenAI and Anthropic lead in technology, in awareness, in services, and potentially in hardware. If OpenAI ever ships a device, that is a direct threat to the iPhone franchise in a way no Android phone has ever been.

The Capex Question: Sitting Out the Arms Race

Every serious AI player is spending like there is no tomorrow. Microsoft, Google, Amazon, and Meta are committing hundreds of billions of dollars annually to data centers and AI chips. Apple has largely sat that out. On one level, that is discipline — Apple does not need to own a cloud business to sell phones. On another, it is a bet that the on-device approach will win, and that bet is not yet proven.

Here is what the infrastructure crowd understands that the consumer press often misses: the on-device AI thesis only works if the models you ship in a phone are good enough without the data center. Apple has been building AI-capable silicon into its devices since 2017, and the Neural Engine in current chips is genuinely impressive. But frontier intelligence — the kind that answers hard questions, writes real code, and reasons through complex problems — still lives in the cloud. If that balance shifts further toward cloud, Apple's hardware advantage matters less and its missing infrastructure matters more.

The Hardware Bet: Chips Over Data Centers

Ternus's entire career points one direction: control the silicon, control the experience. The Apple silicon transition proved the playbook works — by designing its own chips, Apple stopped depending on Intel's roadmap and built the fastest laptop processors on the market. Ternus is now expected to apply the same logic to AI, and there are early signs. Bloomberg reported in January that Apple is accelerating development of three Siri-centered wearables: smart glasses, a pendant, and camera-equipped AirPods. Those are hardware plays — products only Apple could build because only Apple controls the chip, the sensor stack, and the operating system.

But there is a catch the hardware crowd does not like to dwell on: you cannot design your way out of a missing model. Apple can build the best AI glasses on earth, and they will still be running someone else's intelligence unless Apple either builds a frontier model or makes a credible case that on-device models are enough. Nothing Apple has shipped suggests it can match the frontier labs. That is the gap Ternus has to close, and he is starting from behind.

What September 9 Will Tell Us

Ternus makes his public debut as CEO on September 9 at Apple's autumn event, dubbed 'Surprise and Shine.' The expectation is a first foldable iPhone and a new AI-powered Siri. Both matter, but Siri matters more. A foldable iPhone is an incremental product — nice hardware, same platform. A Siri that actually works is a statement about whether Apple can ship intelligence that feels native instead of bolted on.

IDC's Jeronimo makes a sobering point for anyone expecting an instant reset: the September announcements were planned during Cook's tenure, and it will be 2027 before Ternus's own direction becomes visible. In other words, today is the beginning of a transition that will take two years to read. Treat everything Apple ships in the next four quarters as the old regime's roadmap.

What This Means

Strip away the Apple mystique and this is a classic infrastructure story. A company that controlled its entire stack — silicon, OS, distribution, retail — discovered that the one layer it does not control is now the most valuable layer in the industry. Apple's response is to double down on what it is good at: hardware and integration. That is a defensible bet, but it is a bet against the current shape of the AI market.

For people who actually run infrastructure, the lesson is uncomfortable. The most profitable hardware company in history could not buy its way into AI leadership, could not partner its way into it, and is now hoping its silicon advantage outlasts the cloud giants' capex advantage. If Apple's approach works, on-device AI becomes a real category. If it fails, Apple becomes a premium device maker in a world where the intelligence layer belongs to someone else.

What Comes Next

Watch three things. First, what Ternus says about AI on September 9 — specifically whether he talks about models, not just features. Second, whether Apple starts making real AI acquisitions or announces meaningful AI infrastructure investment; talk is cheap, but a GPU commitment is a strategy. Third, how the OpenAI lawsuit and the Google Gemini deal evolve, because Apple's AI future is currently defined by contracts with rivals, and those contracts are unstable.

John Ternus is taking over the most valuable company on earth at the exact moment its core product faces a platform shift it does not control. The next two years will determine whether Apple is the last great integrated hardware company or the first great one to miss the intelligence era. I know which way I am betting, and it is not on the company that rents its brain.

— Allan Ali, Sylt.ing