So what is Amdahl's Law?
The CEO of Anthropic, Dario Amodei, sat down with Nikhil Kamath for a provocatively titled podcast, The AI Tsunami is Here & Society Isn't Ready. Among other things, they discussed debated AI's impact on India's labor market and what it means for businesses there.
Dario used a 60-year-old nerdy idea from computer science to explain what AI does to business moats:
"There is this concept called Amdahl's Law. If you have a process with many components and you speed up some of them, the components that haven't yet been sped up become the limiting factor. When writing software becomes a lot easier, some of the moats that companies have will go away, but others will become even more important. The stuff we thought was really important before isn't as important — whereas these other advantages that we never really thought of as advantages are now super important."
tldr: When one bottleneck disappears, a different one takes its place.
What is it though?
Amdahl's Law is a computer science formula that calculates the maximum theoretical speedup of a program when only part of it is improved — usually through parallel computing.
The core idea: total speedup is limited by the part you didn't or couldn't speed up.
Say a task is 90% parallelizable and 10% serial. Even with infinite processors handling the parallelizable chunk, the best possible speedup is 10x. The 10% serial portion sets the ceiling. Adding more processors past a certain point yields almost nothing.
More processors, diminishing returns. The bottleneck always wins.
Why it matters for AI
The software part of building a company is getting dramatically cheaper and faster. Code that took a team months can now be scaffolded in days. Dario's point is that this doesn't make everything easier uniformly. It shifts where the constraint lives.
If writing software was 80% of what made a product defensible, and AI removes that advantage, it doesn't erase your moat. It exposes a different one. Distribution. Relationships. Proprietary data. Domain knowledge that took years to accumulate. Something.
Companies that assumed "we have engineers" was the answer are now finding out where the bottleneck in the other 20% of their business actually looks like.
That's Amdahl's Law in the real world. Speed up one component enough, and suddenly something else becomes the whole problem.