Tthe market’s view of artificial intelligence is that one of the big winners will be the model-makers,. after all those are the people who invented it. Markets — at least private markets — have assigned huge valuations to the companies building the models, predominantly OpenAI and Anthropic.
The history of great invesntions suggests investors should be careful with that conclusion.
Some of the most transformative inventions ever created generated enormous benefits for society but surprisingly little wealth for their inventors. The World Wide Web was released royalty-free. TCP/IP became an open standard. GPS was made freely available. Linux became one of the most important pieces of software in history while remaining open source.
Even technologies that were patented often failed to produce fortunes proportional to their impact. Bell Labs invented the transistor in 1947, but AT&T licensed the technology broadly under regulatory pressure. That decision helped create the modern semiconductor industry, spawning companies like Fairchild Semiconductor and, eventually, Intel. The economic value created by the transistor is probably in the hundreds of trillions of dollars. Only a tiny fraction accrued to its inventors.
The lesson is simple: inventing a general-purpose technology is not the same as capturing the value it creates.
That brings us to large language models.
Today, investors are valuing AI companies as though owning the best model will produce durable monopoly profits. That may happen for a time, but it’s worth considering another possibility: There are no patents in AI, secrets are hard to keep and distillation appears to produce models that are nearly as good.
If frontier models continue to become commoditized through open-source development, rapid distillation techniques and relentless competition, then paying a premium for the very best model may become increasingly difficult. Models that trail the leaders by a month or two could prove “good enough” for most commercial applications.
Moreover, there is a push for companies to control their own data, which may move them towards open source and away from the big models.
In that environment, pricing power becomes elusive.
It’s often argued that the real money will be made by application companies built on top of foundation models. That sounds plausible until you consider what AI itself does. One of its core capabilities is reducing the cost of creating software. If applications become dramatically easier and cheaper to build, competition increases and excess profits become harder to sustain there as well.
The same logic that commoditizes the model layer could eventually commoditize much of the application layer.
That raises a more interesting question for investors.
Perhaps the biggest beneficiaries of AI won’t be technology companies at all.
Instead, the winners may be businesses with deep competitive moats that operate in the physical economy. Manufacturers, industrial companies, logistics operators, utilities, mining companies and other capital-intensive businesses often earn modest margins despite significant barriers to entry. A company operating on a 2% margin that permanently expands to 3% through AI-enabled productivity has increased earnings by 50% without selling a single additional product.
Those gains are difficult for competitors to erase when the competitive advantages come from scale, infrastructure, regulation, geography or capital intensity rather than software.
This would represent a very different investment story from the one dominating markets today.
Investors may be fighting the last war and over-indexing towards the information economy of the past 30 years.
The internet produced extraordinary returns for software platforms because distribution itself became the moat. AI may prove different. If intelligence becomes abundant and inexpensive, it starts to resemble electricity more than enterprise software.
Electricity didn’t create the greatest fortunes for the scientists who uncovered its principles. Michael Faraday transformed physics without becoming fabulously wealthy. Even Thomas Edison and George Westinghouse, who successfully commercialized electrical systems, captured only a small fraction of the economic value electrification ultimately created.
Most of the benefits flowed to the broader economy through higher productivity, lower costs and entirely new industries.
AI could follow the same path.
That doesn’t mean AI won’t be one of the most important technological advances in history. I think it will become exactly that.
But the most important question for investors isn’t how valuable AI will become, it’s who actually gets to keep the value.
This article was written by Adam Button at investinglive.com.
