Intel Did the Math on the iPhone and Said No. The Math Was Right.

In the summer of 2006, Intel sold a working ARM chip business - the XScale line, processors built for smartphones and PDAs - to Marvell for $600 million, so it could pour everything back into the x86 chips that made it rich.¹² Months later Apple put an ARM chip in the original iPhone and changed how the world computes. The two events look, in hindsight, like the dumbest sequence in modern business: Intel held the future in its hand and traded it for cash. That story is told as a parable of blindness. It is the wrong parable. Intel saw mobile coming with perfect clarity. It did the arithmetic, decided the arithmetic said no, and was right about the arithmetic.

The popular version is that Intel lacked vision - that it couldn't imagine a computer without a fan. Almost the opposite is true. The company knew exactly what it was giving up. It gave it up on purpose, because protecting the cash cow was the textbook-correct move, and the one number that would have changed the answer was the one number nobody got right. (The 2006 XScale sale was framed explicitly as a focus decision — exit ARM, double down on x86 and servers¹ — but the margin logic that made that focus feel safe was the same logic that later paralyzed Intel's Atom strategy.⁶)

The chip Apple wanted was never an Intel chip

Start by killing the cleanest part of the legend. The famous tale has Steve Jobs asking Intel to build an x86 chip for the iPhone and Otellini refusing. That isn't what was on the table. The processor Apple was weighing from Intel was the ARM-based XScale - and there was simply no way an x86 chip could fit inside an iPhone at the time.⁵ Intel couldn't have shipped its crown-jewel architecture into that phone if it had wanted to; the physics weren't there. Otellini himself never claimed a heroic refusal. Years later he put it in terms of regret rather than triumph: 'We ended up not winning it or passing on it, depending on how you want to view it. And the world would have been a lot different if we'd done it.'⁴

So the decision that mattered wasn't a phone call in 2007. It was the choice in 2006 to get out of the ARM business altogether - to sell XScale and double down on x86. That choice was not made in a fog. It was made by people staring straight at a spreadsheet.

The margin logic that made saying no correct

Here is the thesis, stated plainly: Intel didn't miss mobile because it couldn't see it. Intel refused mobile because mobile threatened to cannibalize the most profitable chips in the world, and the rules of margin defense told it not to let that happen. The mechanism is brutally simple. An x86 server or PC chip carried a fat price and a fat margin. A mobile chip was cheap, low-margin, and aimed at a device whose economics, at the volumes anyone forecast, looked like a money-loser. Every dollar of cheap mobile silicon risked pulling demand - and engineering attention, and fab capacity - away from the dollars that actually paid the bills.

That is why Intel could never bring itself to push x86 down into phones with real force. It 'basically couldn't allow low-margin x86 products to cannibalize its high-margin x86 products, so it was stuck with half-measures.'⁶ When Atom finally arrived in April 2008, Intel's own annual report positioned it for netbooks and mobile internet devices - not smartphones.³ Atom was a fence, not a spear. Its job was to keep ARM from creeping upmarket into laptops, not to carry x86 down into the iPhone's territory.⁶ A company defending a cash cow doesn't build a weapon to attack the new market. It builds a wall to keep the new market out of the old one.

The error wasn't the decision. It was one number.

Run Intel's logic forward and it holds up cleanly - on one assumption. The cannibalization math only points to 'no' if mobile stays small. A loss-making chip at a million units is a rounding error you walk away from. A loss-making chip at a hundred million units is a market you cannot afford to be absent from, because at that scale the cheap chip funds the fabs, sets the standards, and trains the next generation of engineers. The difference between the two answers isn't strategy. It's a single input cell. And Otellini, after the fact, named exactly where the cell was wrong: the volume was, in his words, '100x what anyone thought.'⁴¹⁰

This is what makes Intel a genuine case study rather than a punchline. The decision framework was sound. The margin discipline was real. The refusal to cannibalize was the move every business-school case rewards. The catastrophe lived entirely in the forecast - in the gap between a market that looked tiny and a market that became the largest computing platform on earth. Intel didn't fail to think. It thought carefully, with the wrong number.

But couldn't Intel have just bought its way in later?

The fair objection is that a giant with Intel's fabs and cash could simply force its way into mobile once the scale was obvious - subsidize the chips, buy the volume, and let the manufacturing edge do the rest. Intel tried exactly that, and the attempt proves the deeper point. Under a later CEO, Brian Krzanich, Intel set out to ship 40 million tablet chips in a single year by heavily subsidizing Atom; it actually shipped 46 million - and the effort hurt profitability so badly that Krzanich decided not to repeat the strategy for smartphones.⁷ In other words, when Intel finally did the thing the cannibalization rule had forbidden - eat your own margin to win the market - the rule was vindicated. It really did lose money, and Intel really did pull back. The honest counter to 'Intel should have just paid to win mobile' is that Intel paid, and discovered the price was higher than the prize, by which point the market had already standardized on ARM. The window had been open in 2006. By the time Intel was willing to spend through it, it was a wall.

And the cash cow Intel defended so carefully? It held - for a while. Intel's x86 share peaked around 89% in 2014, then began a structural slide as AMD's Ryzen took ground; by Q4 2025 Intel held 70.8% of the x86 market, according to Mercury Research data.⁸⁹ The fortress was real, the margins were worth protecting, and the discipline that protected them was correct. None of that saved Intel from the one thing margin logic could not price: a market it was rationally right to skip, that turned out to be the future. Intel did the math, and the math was right. It just left out the only line that mattered.