Vertical Integration

Vertical integration is the decision to make the hard parts of a product yourself instead of buying them from outside suppliers. For Elon Musk that means SpaceX building its own rocket engines and avionics, and Tesla building its own battery cells, self-driving chips, and even the giant single-piece body castings under its cars. The reason is not pride of ownership. It is that owning the part means owning how fast you can change it, and owning the cost.

What is Vertical Integration?

Most manufacturers are really just assemblers. They buy engines, chips, and other pieces from a web of specialist suppliers and bolt them all together. Vertical integration runs the other way: you pull the critical pieces in-house and make them under your own roof. The trade-off is real. Doing it yourself is harder and more expensive at the start, because you have to build the factory and learn the craft from scratch. But the payoff is control. When you make the part, no supplier gets to set your price, your timeline, or how fast you are allowed to get better.

Why It Matters to Elon Musk

The idea was forced on SpaceX from the very start. In 2001 and 2002 Elon Musk tried to simply buy rockets, traveling to Russia to purchase refurbished missiles, and came home turned down and disgusted by what suppliers wanted to charge. So SpaceX hired engineers and built its own instead. Walter Isaacson's biography records the mental tool that came out of that fight, which Elon Musk calls the idiot index: the cost of a finished part divided by the cost of its raw materials. A high number means the difference is waste, a signal that the part should be redesigned and made in-house for less. Out of the same instinct comes his five-step algorithm, drilled into teams at both companies: question every requirement, delete the part if you can, simplify what is left, speed it up, and only then automate.

How it Works in Practice

At SpaceX the engine is the clearest case. The Raptor is designed at the company's own facilities, its turbopumps and injectors were 3D-printed early on, and its manifolds are cast from superalloys SpaceX developed itself rather than ordering from a metals vendor. Because the engine is theirs from end to end, the team can change a design and run the new version on the test stand on their own schedule, not a supplier's. Tesla uses the same thinking on the car. Rather than buy its self-driving computer, Tesla designed its own. Rather than buy battery cells from someone else, it builds the 4680 cell on its own lines. And instead of welding a rear underbody together from dozens of stamped pieces, it casts the whole section as a single part on a giant press.

The Evidence

The numbers show why these parts were worth taking in-house. When Tesla unveiled its self-driving computer at Autonomy Day in 2019, it reported 2,300 frames per second of image processing against just 110 on the bought hardware it replaced, about a 21-fold jump from a chip built for its exact job.

The casting story is about deletion rather than raw speed. A rear underbody that used to be put together from roughly seventy separate stamped and welded pieces becomes one aluminum casting, which Tesla credited with cutting that section's production cost by about forty percent. Across the front and rear castings and the structural battery pack, Tesla counted hundreds of parts removed from the car entirely.

The broadest number is SpaceX as a whole. By making its own engines, avionics, and structures, the company builds close to ninety percent of a rocket in-house, far above the buy-and-assemble norm of the launch industry.

Vertical Integration Today

The principle now runs through every Elon Musk company. SpaceX builds its Raptor engines, its avionics, and the Starlink satellites it launches. Tesla makes its self-driving computer, its 4680 cells, and its body castings, and trains its self-driving software on computing hardware it increasingly owns outright. The pattern holds everywhere: wherever a supplier controls the price or the pace on a part that matters, the answer has been to learn to make that part yourself.

What Comes Next

The next frontier is the most specialized layer of all, the silicon itself. Tesla and SpaceX have both signaled plans to bring more chip and computing manufacturing in-house rather than lean entirely on outside chip factories, the same move applied one level deeper. If the past is any guide, the goal is not to make absolutely everything, but to own whichever part is too costly, too slow, or too important to leave in someone else's hands.

The Bottom Line

Vertical integration is a bet that control beats convenience. It costs more to start and demands that you master things other companies are happy to buy. But for products as hard as a reusable rocket or an affordable electric car, owning the engine, the chip, and the casting is how Elon Musk turns a supplier's markup and a supplier's calendar into his own.

Related

Keep reading: Capital Allocation, First Principles Thinking. Zoom out to the State of Elon overview, or open the Promises Tracker.