Does the company that makes the best hardware in the world actually need to win the AI model race? Or does it just need to own the device you’re holding when you talk to one?
That’s the question I keep turning over as we process the biggest leadership transition in tech since, well, since Steve Jobs handed things to Tim Cook. John Ternus is now running Apple. Tim Cook has stepped aside. And the timing tells you everything about where this industry is headed.
A Hardware Guy Takes Over at the Exact Right Moment
Let me be blunt: Apple has been behind in AI. Not embarrassingly behind — more like fashionably late to a party it didn’t think was worth attending. Siri has been a punchline for years. Apple Intelligence, when it finally showed up, felt more like a catch-up move than a conviction bet. The company has been playing from behind in the model space, and everybody knows it.
But here’s what most AI pundits keep getting wrong: they keep scoring Apple on a test it isn’t trying to pass.
John Ternus is a hardware executive. He’s the guy who oversaw the transition to Apple Silicon, the M-series chips that turned MacBooks from overpriced fashion accessories into legitimate machines that developers and creators actually prefer. He understands supply chains, product design, and the physical layer of computing better than almost anyone alive. And in 2026, with Apple’s stock up roughly 17% on the year and the company overtaking Nvidia in valuation, the market is telling us something: the next phase of AI might not belong to the model builders. It might belong to the device makers.
Nvidia’s Whole-Stack Gamble
On the other side of this story, Nvidia is doing what every dominant infrastructure company eventually does — it’s reaching further up the stack. Training chips weren’t enough. Inference chips weren’t enough. Now Nvidia wants the software layer, the deployment tools, the developer ecosystem, the full vertical. It’s a bold strategy, and I won’t pretend it’s stupid. Jensen Huang doesn’t make stupid bets.
But vertical integration is a different discipline than selling GPUs to hyperscalers who will buy anything that makes their models train faster. When you try to own the whole stack, you start competing with your own customers. Cloud providers notice. Startups notice. And the moment a viable alternative appears — whether it’s AMD, custom silicon from Google and Amazon, or something we haven’t seen yet — those customers start diversifying their supply chains fast.
Apple overtaking Nvidia in valuation isn’t just a stock price footnote. It signals that investors are recalibrating. The gold rush phase, where the picks-and-shovels seller dominates, may be giving way to a phase where the companies that actually put AI in front of billions of users start commanding the premium.
Why This Matters for AI Tools and Agents
I review AI tools and agents for a living. And what I’ve seen over the past year is that the model layer is commoditizing faster than anyone predicted. GPT-5, Claude, Gemini — they’re all remarkably capable, and the gaps between them are shrinking. The real differentiation is moving to two places:
- The application layer: Who builds the best agents, the best workflows, the best user experiences on top of these models?
- The device layer: Who controls the hardware where these agents actually run, where they access your data, where they earn your trust?
Ternus inherits a company that dominates the second category. Over a billion active iPhones. A silicon team that’s arguably the best in the world. An ecosystem that developers can’t afford to ignore, even when Apple’s own AI efforts have lagged. If Apple gets even moderately competent at on-device AI — and the M-series trajectory suggests they will — the implications for how agents interact with users are enormous.
Imagine an AI agent that lives on your device, processes sensitive data locally, and integrates natively with every app in your life. That’s the play. And it doesn’t require Apple to build the best foundation model. It requires them to build the best foundation for models.
My Honest Take
I’m not an Apple fanboy. I’ve trashed Apple Intelligence in previous reviews for being half-baked and overly cautious. But I also know how to read a chessboard. The Ternus era begins with Apple bigger and richer than it’s ever been, catching up in AI capability while already owning the most valuable distribution channel in consumer tech. Nvidia starts this chapter still dominant in infrastructure but facing the classic innovator’s dilemma of expanding into territories where it has no natural moat.
For those of us building with and reviewing AI agents, the message is clear: watch the edges, not just the cloud. The next generation of AI experiences won’t be defined by who has the biggest model. They’ll be defined by who owns the last foot between the model and the user. And right now, that’s Apple’s game to lose.
🕒 Published: