DeepSeek & Teddy Roosevelt

A Chinese company called DeepSeek just built a very good AI model that is open source and wiped a trillion dollars of market capitalization off of stock market value of several giant tech companies. Boohoo. A few interesting things about this.

The first wild thing is that a trillion dollars could evaporate overnight and it wouldn't matter much to the world. That's because there is just so much investable capital in the world now. It's a tough number to measure or even define (is the value of your house investable capital?) but if there was say thirty trillion dollars twenty years ago, there is five or ten times that now. And most of that increase has been among the extremely rich and who cares if some billionaires lose a small portion of their fortunes?

This Giant Pool of Money looks for places to flood, and will always speculate in the next big thing which is currently AI. The enshittification model that we live under works as follows: tech companies get insane volumes of cash, build fast, get tons of users who are locked in or too lazy to leave, they buy competitors, then kick back and make money while your service becomes obsolete and terrible. When these giant tech companies' market shares are comfortably in the double digits they all change from being champions of market competition to being champions capitalist protection of whatever it is they own or buy. From the latter it's a short hop to being monopolists. Whatever they own will lock in users and prevent competition, because that's the way to make tons of money if nobody stops you. And now nobody is stopping anybody; so for example, Google will not be forced to sell Chrome.

But in tech, open source alternatives mess with that arrangement and DeepSeek is open source. This is particularly true with AI, in part because AI writes great code that makes itself better. And smarter. And faster. Western companies built the best AI models, but China is right there and put out their own model in a matter of months. They pretended they did it for five million bucks, it was actually hundreds of millions, but that is still not much compared to the billions that investors are pouring into AI companies. And yet, as is so often the case, a great deal of the most interesting work is being done by individuals working on open source models. The giant tech companies in the US and Europe are used to being able to build walls around themselves. It's made them lazy. They will expend incredible efforts to build and keep those walls up to protect their monopolies but the acceleration of change that AI allows makes it an uphill battle.

The last time monopolies were tamed was in the period when giant fortunes were made in railroads and oil after the US Civil War. As income inequality rose due to these monopolies, enriching the rich and impoverishing the poor, political violence increased and the middle class increasingly blamed the monopolies, or trusts as they were then called. The methods used to bring them to heel were pretty simple. The first was to call them out publicly and there was a whole group of politicians and pundits over decades who simplified and explained the methods and meaning behind the monopolies. Then break them up and prevent new acquisitions, and finally, regulate them. It was popular and it worked for a time, but then went from the industrial to the financial sector, and that sector in turn had to be tamed during the Great Depression. That worked until the deregulation of the 80's and 90's. And so here we are again with stratospheric inequality and robber barons, but the path forward is tried and true if we decide to take it.

###