Musk's Terafab bets up to $119 billion on doing the entire chip supply chain in one building

The building that wants to do everything

Terafab is not just big. It is maximal. One campus is meant to do chip design, etching, memory, packaging, and testing - every step the rest of the industry hands off to a different specialist. It is meant to make several chip variants at once, at the hardest process technology, for cars, robots, and orbit.

Read the plan and you can feel the appetite behind it. Not “let’s make one thing well,” but “let’s do all of it, ourselves, at the largest scale ever tried.” That appetite has a name, and it has caught better builders than Musk.

The pattern has a name

In 1975, a software engineer named Fred Brooks described what he called the second-system effect. The first system a designer builds is usually lean. They are unsure, so they keep it small and cut hard. It works.

The second system is the dangerous one. Now they have proof they were right. Every feature they cut the first time, every idea they parked, they add back - all at once. The result is bloated, late, and often dead on arrival.

This is not the same as saying you cannot design something complex. Plenty of complex things get built on purpose. The trap is narrower and sneakier: it strikes the person who has just won.

Why success is the trap

The danger here is not inexperience. It is confidence.

A beginner keeps things small out of fear. A winner has had the fear burned off. Their first restraint looks, in hindsight, like caution they no longer need. So the guardrail that kept version one shippable gets removed exactly when the ambition gets bigger.

Every added piece seems reasonable on its own. Own the memory too. Add the packaging. Support four chip variants. Each is a small yes. Together they are a scope no single team can hold, and the weight lands all at once, near the end, when it is hardest to cut back.

The industry already learned this

Here is the part that poses as plain fact but is really a hard-won choice. The chip industry is split across specialists - foundry, memory, packaging, testing - not because no one thought to combine them, but because combining them broke the companies that tried.

Intel is the warning next door. It was America’s original all-in-one chipmaker, designing and building under one roof. Doing both at the leading edge nearly sank it, and it has spent years recovering by narrowing what it tries to do. The specialised chain is not a lack of ambition. It is the scar tissue from everyone who reached for everything and could not hold it.

Where else it catches you

Once you see the second-system effect, it is everywhere, and you are somewhere inside it.

It is the software rewrite: a clean first product replaced by a “version two” that tries to fix everything and ships years late, if ever. It is the hit debut followed by the overloaded second album. It is the modest civic success that spawns a grand plan, over budget and stalled. It is the small business that wins, then bets the whole thing on doing ten new things at once.

The people who feel the collapse are rarely the ones who ordered it. The staff who burn out on an impossible scope. The town promised jobs on a plant that may never date its later phases. The customer waiting on the thing that keeps not arriving.

What the whole picture asks of us

The lesson is not that big bets are wrong. Sometimes the maximal version works, and sometimes caution is just falling behind. The lesson is smaller and harder: the moment you are most sure you can hold everything is the moment to check whether you can.

No single seat sees the whole chain - that is why the industry cut it into pieces each team could master. The confidence that follows a win is real information about the past, and almost none about the load ahead. We are all somewhere on our own second system, adding one more reasonable yes. The humble move is to remember that the first one worked because of what it left out.