The Lost Decade: How Intel Handed TSMC the Most Valuable Position in Technology

In May I wrote about TSMC’s 30 billion dollar moat and how capital intensity made it unassailable. That piece answered how TSMC built the position.

It did not answer the more useful question, which is who vacated it.

For roughly forty years, Intel held the most defensible position in technology. It designed chips and it manufactured them, and it was better at manufacturing than anyone alive. That combination is called integrated device manufacturing, and while Intel held the process lead it was close to unbeatable.

In the first quarter of 2026, TSMC reported 35.9 billion dollars in foundry revenue. Intel Foundry reported 5.4 billion. Of Intel’s 5.4 billion, roughly 174 million came from external customers. That is about 3%.

Intel is now a foundry that mostly manufactures for itself, competing against a foundry that manufactures for everyone.

The Pain Point: A Moat You Have to Keep Paying For

Some advantages are structural. A network effect gets stronger while you sleep. A brand accumulates. A switching cost compounds, which is why Oracle can ship worse software and keep the customer.

Process leadership is not one of those.

Manufacturing leadership is an advantage you re-earn every eighteen to twenty-four months, at enormous capital cost, with a technical execution risk that does not go away because you succeeded last time. Miss one cycle and you have not lost a quarter. You have lost the cycle, and your competitor gets to build volume, yield and customer relationships on a node you cannot match.

The dangerous property of this kind of moat is that it looks identical to a structural one right up until the moment it fails. Intel in 2014 looked as safe as any company in technology.

The Radical Story: One Node, Six Years

Intel ran a cadence it called tick-tock. Tick was a new manufacturing process, tock was a new architecture on that process, alternating on a predictable rhythm. It was the operating system of the company and the industry set its watch by it.

The 10 nanometre process was delayed in 2015.

That single sentence contains the whole story, but the consequence is the part worth studying. Intel did not skip a generation and move on. It stayed on 14 nanometres and shipped generation after generation of products on the same underlying process, adding architectural features and clock speed to compensate for a node it could not deliver.

Meanwhile TSMC moved 7 nanometres into high volume. Then 5. Then 3.

And here is the mechanism that turned Intel’s delay into TSMC’s monopoly. TSMC did not have to win the customer relationship by being a better designer. It only had to be available. AMD had already separated design from manufacturing, so when TSMC reached 7 nanometres, AMD moved. When TSMC reached 5, AMD moved again. Apple, Nvidia and Qualcomm did the same thing.

Intel’s competitors did not need to solve Intel’s manufacturing problem. They rented someone else’s solution, and every year Intel spent fixing its own process was a year those competitors spent shipping on a better one.

Why it broke

The integrated model was Intel’s greatest strength and it became the trap.

A fabless competitor with a manufacturing problem changes supplier. Intel with a manufacturing problem has to fix the factory, because the factory is the company. There is no alternative supplier when you are the supplier.

So the same vertical integration that made Intel unassailable while it was winning made it uniquely unable to route around failure. The advantage and the fragility were the same structure viewed from different sides.

What the Position Is Worth Now

The 2026 numbers show what the vacated position was actually worth.

TSMC ran a 58.1% operating margin in the first quarter of 2026. Intel Foundry ran a 2.4 billion dollar operating loss in the same quarter. That loss is narrowing, by 72 million on the prior quarter, on better yields across Intel 4, Intel 3 and 18A, and foundry revenue grew 20% sequentially and 16% year on year. The direction is right. The gap is still the width of an industry. Nodes at 7 nanometres and below account for roughly 74% of TSMC’s wafer revenue, with 3 nanometres alone contributing 25%. TSMC is the default manufacturing partner for Apple, Nvidia, AMD and Qualcomm, which is to say it manufactures for the four companies most responsible for the current computing cycle.

Intel’s 18A process is real progress and should not be dismissed. But yields tell the commercial story more honestly than node names do. Intel’s 18A reached roughly 55% by the middle of 2025 against a target of 65 to 70%. TSMC’s N2 was around 70%, heading for 75%.

Yield is cost per usable chip. A customer choosing a foundry is choosing a yield curve, not a marketing name, which is why technical recovery and commercial recovery are different events separated by years.

The Playbook: Knowing Which Kind of Moat You Have

1. Classify the advantage honestly

Ask one question about your moat: does it get stronger if we do nothing for two years, or weaker? Network effects and switching costs strengthen. Process leadership, cost leadership and speed advantages decay. They are rented, not owned, and the rent is due every cycle.

2. Treat a missed cycle as a strategic event, not an operational one

Intel handled 10 nanometres as a delay to be managed. It was a change in competitive position. When your rented advantage lapses, the correct response is to re-plan the business, not to re-plan the schedule.

3. Vertical integration cuts both ways, and you only feel one side at a time

Owning the whole stack removes supplier risk and adds concentration risk. While you are the best, integration is a moat. The moment you are not, it is the reason you cannot substitute. Know which regime you are in.

4. Your competitors do not have to solve your problem

They only have to find someone who already has. This is the part Intel’s strategy underweighted. The existence of a credible external supplier turns your internal failure into their opportunity at almost no cost to them.

5. Recovery in engineering is not recovery in the market

Intel may well close the technical gap. Winning back Apple or Nvidia is a different project, with a different timeline, against an incumbent that now has scale, yield and a decade of trust. Getting the technology back is necessary. It is nowhere near sufficient.

Intel did not lose because it was outmanoeuvred. It lost because it treated a decaying advantage as a permanent one, and by the time the difference was obvious, the position had someone else standing in it.

Every advantage you have is either compounding or decaying right now. Very few leaders can tell you which of theirs is which.