Intel Didn't Blow the iPhone in One Meeting. It Had Already Sold the Door.

The story has the shape of a parable, which is the first clue it's wrong. A CEO sits across from Steve Jobs, hears the price Apple wants to pay for the chip inside a phone nobody's seen yet, runs the math in his head, and says no. A few years later that phone reorders the world economy, and the man who said no spends his retirement explaining the most expensive 'no' in the history of computing. Paul Otellini did give that interview, and he did say his numbers were wrong by an order of magnitude.¹ But the parable buries the more useful fact under a tidy morality tale. By the time the iPhone existed, Intel had already sold the building it would have needed to say yes.

The official memory is that Intel passed on the iPhone in a single fork-in-the-road decision. The record says Intel had been backing out of the relevant business for months before the iPhone was public, and the man at the center of the legend wouldn't even fully cosign it.

Read that line slowly. The CEO at the center of the legend can't decide whether his company turned the deal down or lost it. That hedge is the whole story, because a clean 'pass' implies a door that was open and a choice to walk away. The truth is the door was already being bricked up. Intel's thesis here is uncomfortable: this was not one man's failure of imagination. It was an institution that had decided, on the record, that it didn't want to be in the room the iPhone would be built in.

The decision that came before the decision

On June 27, 2006, Intel announced it was selling its entire ARM-based chip business — the XScale communications and application-processor line — to Marvell for $600 million. The stated reason was to let Intel 'focus its investments on its core businesses, including high-performance, low-power Intel Architecture-based processors.'² Marvell's own filing dates the asset purchase agreement to June 26.³ Now line that up against the calendar that matters: the iPhone was not announced until January 2007, and its chip didn't ship until June 29, 2007.⁴ Intel sold the ARM business roughly six months before the world knew the iPhone existed. The popular framing — a discrete, reversible decision by a CEO who could have just said yes — requires a business unit that Intel had already handed to a competitor for cash.

It was never going to be an Intel chip anyway

Here is the architectural detail the morality tale skips. People imagine Intel could have slotted an x86 chip into the iPhone, the way it slotted x86 into the Mac the year before.⁵ But the discussion was never about x86. The phone needed a low-power ARM application processor — exactly the XScale category Intel was selling off. And the chip that actually shipped, the APL0098, was an ARM11 design fabricated by Samsung on a 90nm process.⁴ Intel's XScale parts came from a different ARM lineage on a different node. There was no drop-in substitute sitting on a shelf. Saying yes wouldn't have meant signing a purchase order; it would have meant committing to a fresh design and a manufacturing ramp in a business Intel was, at that very moment, paying to walk away from. The 'pass' wasn't declining a sale. It was declining to reverse a strategy.

And the reason for that strategy was identity, not arithmetic. Intel saw itself as the company that designs the chip and owns the architecture — x86, top to bottom. Being an ARM licensor, building someone else's instruction set to someone else's spec at someone else's price, sat crosswise to everything that made Intel Intel. That is why selling XScale felt like clarity rather than retreat: it returned the company to its self-image.⁸ The price Otellini cited — Apple's offer 'below our forecasted cost,' not recoverable on volume¹ — wasn't a freak miscalculation by one executive. It was the natural output of a cost structure built to make high-margin processors for PCs, pointed at a consumer-mobile market that runs on razor-thin margins and ferocious volume. Intel didn't have the wrong number. It had the wrong machine for generating numbers in this market at all.

But Otellini admitted it was a blunder — so wasn't it?

The honest counter is that Otellini himself conceded the loss, and bluntly: the forecasted cost was wrong, and the volume turned out to be '100x what anyone thought.'¹ If the man who made the call says he blew it, who are we to launder it into a structural inevitability? Fair. But notice that admitting the outcome was a disaster is not the same as admitting the decision was a single, reversible choice. Two independent accounts crack the price-only story open further. Steve Jobs's telling is that Apple rejected Intel partly because Intel was organizationally too 'slow,' and because Apple didn't want to strengthen a competitor it was already buying Mac chips from.⁷ And an adversarial 2024 reconstruction argues Intel may not even have known the chip was for a phone — the spec may have arrived dressed as an iPod opportunity — making 'Intel passed on the iPhone' a sentence that smuggles in knowledge nobody at Intel had at the time.⁶ When the buyer says you were too slow, the seller says he can't decide if he passed or lost, and a careful re-read suggests neither side knew what was being built, the clean parable disintegrates. What's left is harder and truer: a company that had already chosen not to be there.

The temptation is to imagine an alternate Intel that said yes in the meeting and rode the iPhone to a trillion dollars of mobile silicon. But there was no such meeting to win, because there was no such Intel. The company that could have built the iPhone chip was the one Intel sold to Marvell for $600 million, to get back to being the company it had always wanted to be. The miss wasn't a moment of bad judgment. It was a company being exactly itself — and discovering, a few years late, what that cost. The most dangerous 'pass' isn't the one you regret. It's the one you'd make again, for the same proud reasons, having learned nothing.