Musk Builds Rockets

Before Musk intervened, building rockets was considered a high-cost, long-cycle project with few private investors willing to participate.

Musk wanted to get into this business and consulted many experts, who gave the same conclusion, but Musk wasn't willing to accept it. One day on a flight, he sat alone studying material quotes.

After extensive research, he discovered an astonishing fact.

Rocket raw material costs accounted for only 2% of the total, while a staggering 98% were due to process issues, decision-making problems, and workflow inefficiencies.

He concluded this was unreasonable and that there was enormous room for improvement.

If I send an army to war and you tell me the general staff consumed 98% of the budget while the soldiers only used 2%,

how would you explain its reasonableness? It is clearly irrational.

So he approached the experts again and showed them the numbers, asking them to explain.

The experts told him, "You don't understand; this requires such complex processes and decision-making workflows."

Since the post-WWII era, when the US first started its space program, this is how it had been done—the so-called ancestral ways cannot be changed.

Musk just smiled and said:

"Except for the laws of physics and mathematical formulas, everything else can change."

How to change? By trying.

Old American rocket manufacturing was very similar to the chip design departments of domestic Chinese chip companies over a decade ago.

Back then, tapeouts were very expensive, costing a fortune each time, and all had to be sent abroad, often facing queue times of several months.

So chip design departments had no opportunity to experiment; they could only calculate repeatedly to ensure the first tapeout succeeded.

A huge amount of cost was piled onto this development stage.

In contrast, software teams faced no such constraints.

Why?

Software can be adjusted; the code I write, even without checking, can be run at zero cost.

Each run is like a tapeout for the chip design team—if there are issues, just fix them.

Run it once, and all bugs are reported; fix them, run again, fix again.

Thus, software development always runs ahead of chip design.

While you're still waiting for your tapeout, I've already finished debugging the OS, all the drivers on an FPGA, and have even started application-layer development.

Software always waits for hardware, and hardware always waits for tapeouts.

Rocket launches faced the same problem.

Everyone was afraid to experiment, fearing launch failures, so they全力以赴 to ensure success on the first attempt. This made materials ever more obscure and R&D workflows increasingly bizarre.

This cost gradually accumulated to account for 98%.

Musk adopted a different approach. He said, "Let's not treat rockets like hardware; let's approach them like software."

The question is, from a physics perspective, can stainless steel actually be used?

If it can, let's use it.

Leave those traditional ancestral rules aside; as long as the physics works, use it. The only goal is to reduce the cost per launch.

Once the cost per launch is lowered, a hardware debugging game becomes software debugging.

As long as my launch cost is low, failure is acceptable. Each explosion provides massive amounts of real-world data.

Then I fix bugs, iterate again, fail again, gather more data, fix again, and repeat.

After several rounds of rapid iteration, I become the most technologically advanced and competitive player in this field.

Because I turned a hardware game into software-like development.

This is a classic business case. Musk's cross-industry innovation, applying software iteration to hardware, has been studied worldwide.

His core philosophy can be summed up in two sentences.

1. Rapid iteration is king; it crushes everything.

2. To achieve rapid iteration, the top priority is reducing the cost per attempt.

As long as these two points are把握, your rapid iteration will roll forward like a snowball.