I have been watching something strange happen in the AI industry and I think there is a historical pattern that explains it better than any analyst hot take. The big AI companies like OpenAI and Anthropic and Google are valued at hundreds of billions because they have massive language models that nobody else has. But those models are becoming commodities at an incredible speed. Just 12 months ago GPT-4 was miles ahead of anything else and today free open-source models are catching up in months at a fraction of the cost.

This has played out before. In 1769 James Watt patented his improved steam engine design and for the next 31 years he and his partner Matthew Boulton had a complete monopoly. They charged enormous fees equivalent to millions today for anyone who wanted to use their technology and business boomed. They were the OpenAI of their era and everyone needed their engines and they controlled the future. Then 1800 arrived and the patents expired. Within months competitors flooded the market and Richard Trevithick built high-pressure engines that were ten times more powerful than Watts designs while being smaller and lighter and cheaper. The technology that Watt had kept locked away for decades suddenly became something anyone could build and improve and steam engine adoption exploded. Watts company had to pivot from being the technology leader to just another manufacturer competing on price and service.

The part that should worry proprietary AI companies is that the winners of commoditization were never other technology firms. When steam engines became commoditized the real money shifted to infrastructure companies like railroads and factories and ship builders who could build them cheaply and integrate them into existing systems. They did not need to invest billions in R and D. They just took the blueprints and built businesses around them. We are seeing the same pattern with AI where Meta released LLaMA as open source because they understood that making the technology free lets them sell the infrastructure to run it.

Proprietary AI companies are valued as if their technological moat will last forever but history shows that the valuation usually crashes when technology becomes commoditized. Watt’s company went from having a monopoly on the most important technology of the Industrial Revolution to being just another steam engine maker and their premium disappeared. Today’s AI leaders face the same fate at a much faster pace because what took decades in the 1800s now happens in months.

Every month open-source models get closer to proprietary ones and the cost to train and run them keeps dropping and more companies realize they can download powerful AI for free instead of paying premium prices. The companies valued at hundreds of billions today are building on something that is rapidly becoming a commodity. The question is how fast the valuation correction will happen.

History does not repeat itself exactly but the AI industry is playing out the same pattern that the steam engine went through two centuries ago. The future belongs to people who understand that the real value is in knowing what to do with the technology when everyone has access to it.

Tweet me your responses. I am @troysk704.