Apple has named John Ternus, its head of hardware engineering, as its next CEO. At the same moment, Apple is paying Google roughly a billion dollars a year to run the AI brain of Siri, because Apple’s own AI hasn’t been good enough.

Apple’s hardware is excellent but the AI is not. And the AI is where the next decade of computing is going to be decided. These two things are true at once, and I’ve been turning them over this week.

Steve Jobs faced a version of this problem, and got it wrong for most of a decade before he got it right. That’s the story I’ve spent four years reporting. After Apple pushed him out in 1985, Jobs started a company called NeXT and ran it for twelve years.

My book on those years, Steve Jobs in Exile, is out May 19. The ending is famous. Apple bought NeXT for its software, Jobs came back and the iPhone followed. What gets skipped is the eight years in the middle, when Jobs almost lost everything because he couldn’t hear what his own customers were telling him.

I keep thinking about one scene.

By the early 1990s, NeXT was burning through cash. The computers were beautiful and too expensive. The machine Jobs had poured five years into – a black magnesium cube with a fragile paint job and a finicky optical disc drive – wasn’t selling. What was moving was the software. The operating system called NeXTSTEP was around five to seven years ahead of anything else on the market. Developers and customers loved it, and a handful of firms were building their businesses on top of it.

One of those firms was O’Connor & Associates, a Chicago high-speed trading shop. NeXTSTEP let them write proprietary trading software fast, which gave them an edge on Wall Street. One day a principal from O’Connor showed up unannounced at NeXT’s offices in Redwood City, California, and asked to see Phil Wilson, NeXT’s human resources chief.

He closed the door.

“Phil,” he said, “you have to help me make Steve understand the value of NeXT computers to us is not in the hardware, it’s in the software.”

O’Connor had done the math. Without NeXTSTEP, they’d lose their trading advantage, and without the NeXT computer, they’d lose NeXTSTEP. They were so dependent on the software that at the end of every quarter they called NeXT and ordered the most expensive workstations on the price list, overspecced, overpriced, whatever it took.

“O’Connor basically made payroll,” Mike Slade, NeXT’s marketing VP at the time, told me. The firm was propping up the hardware business just to keep the software alive.

Phil Wilson went back to his office and wrote Jobs a long email laying it all out. NeXT should abandon hardware. Become a software company. The customers were telling them where the value was.

Jobs wrote back, “You’re an HR guy.” Then a subtle threat. If NeXT became a software company, he wrote, it would have to downsize, “and we won’t need a guy like you.”

It took Jobs two more years and two more rounds of layoffs and a near-bankruptcy to figure out what he needed to hear. On a day employees called “Black Tuesday” in February 1993, Jobs finally shut down the hardware division, laid off most of his employees, and sold what was left to Canon. Three years later, he told Terry Gross on Fresh Air that the door to start a new hardware company had closed before NeXT ever opened. He just hadn’t seen it at the time.

Here’s what I want to be careful about, because it’s easy to draw the wrong lesson from this. Jobs’s original NeXT vision, hardware and software built together as one integrated platform, was prescient. That vision is literally what saved Apple after he returned in 1997 and was later appointed CEO. The iPhone is the purest expression of it, hardware and software bound together, designed as one object, by a company that owns both stacks.

The mistake at NeXT wasn’t betting on hardware-plus-software. It was insisting on selling that vision on Jobs’s terms, in Jobs’s form factor, at Jobs’s prices, while customers were telling him what they actually needed was the piece they could afford and use. He wanted to show them his vision for the future, but they wanted a tool that solved their problems right now. It took him most of a decade to see that those were different things.

Now look at Apple in 2026. The hardware is still the best in the business. The integration – device to chip to OS – is the thing no competitor can match. That part of the original vision is intact. What’s failing, what has been failing for two years now, is the AI. Apple Intelligence landed with a thud and the promised personalized Siri got delayed. The company finally cut a check to Google for a custom Gemini model to paper over the gap.

And, in that moment, the board named a hardware engineer to run the company and promoted another hardware engineer, Johny Srouji, to oversee all the devices.

There is a version of the future where this is the right call. Apple’s silicon team is extraordinary. On-device AI may turn out to be where the next ten years are actually decided, and no one is better positioned for that than the company Ternus is about to lead. Maybe the Google deal is a bridge – the way Google Maps was on the first iPhone – and Apple is two years away from running the same integrated play on AI that it ran on chips.

That’s the bull case. I don’t dismiss it.

The story I keep coming back to is NeXT. A company doubles down on the thing it knows how to do, while its customers try to tell it something has shifted. The founder, or the board, doesn’t hear it. The problem is that it would require giving up a version of the company they’ve spent years building.

Tech author and veteran Asia news correspondent Geoffrey Cain‘s Steve Jobs in Exile comes out May 19 from Portfolio/Penguin. If this piece resonated, the book is the full story. Pre-order here. Subscribe for free to his Substack newsletter The Burner Files.