AI is creating enormous value. But here’s the puzzle: the companies building AI aren’t necessarily the ones making money from it.

Who creates the value?
The AI labs — OpenAI, Anthropic, Google DeepMind, Meta. They spend billions training models that can write, code, analyze, and reason. They’re doing the hard scientific work. But most of them are still losing money. Training a frontier model costs $500 million to $1 billion. Running it costs more. Customers expect prices to keep falling.

Who captures the value?
Three groups, in this order:
First, the “picks and shovels” sellers. During the gold rush, the people who got rich weren’t miners — they were selling shovels and jeans. Today’s equivalent: Nvidia (chips), TSMC (chip manufacturing), and the cloud giants — AWS, Microsoft, Google Cloud. Every AI company, no matter who wins, has to buy from them. Nvidia’s profits exploded while AI labs burned cash.
Second, companies with proprietary data and customer relationships. Bloomberg, Adobe, Salesforce, and big banks own data and workflows that AI alone can’t replicate. They plug AI into what they already have and charge more.
Third, surprisingly, end users. A small business owner using ChatGPT for $20/month gets work done that used to cost ₹50,000. Most of AI’s value is leaking to consumers as cheaper, better services — not into corporate profits.
The losers?
Middlemen whose only job was processing information — basic coders, junior analysts, content writers, customer support agents. Their work is being automated faster than new roles emerge.
The lesson: In any technology revolution, value creation and value capture are different games. The smartest builders aren’t always the biggest winners. Position matters more than invention.