Imagine you’re at an auction where a $1 bill is up for bid. Sounds like easy money, right? You could bid $0.10, win the dollar, and pocket $0.90. But there’s a twist: the second-highest bidder must also pay their bid, and gets nothing.
This is the dollar auction, a game theory experiment first described by economist Martin Shubik in 1971. It starts innocently enough. Someone bids $0.05. Another person bids $0.10. Then $0.20. Each bid seems rational - you’re still making a profit on a dollar.
But then the bids reach $0.90. If you’re the second-highest bidder at $0.80, you’re about to lose $0.80 unless you bid $1.00. At a dollar, you break even, which is better than losing $0.80. But now the person who bid $0.90 faces the same calculation. They can either lose $0.90, or bid $1.10 to “only” lose $0.10. Each step of the way, the incremental decision is rational. Stop now and lose everything you’ve already bid, or keep going and minimize your losses.
The results are predictable. In experiments, the dollar often sells for $5, $10, or even $20. Two rational actors, each making rational decisions at every step, end up in a completely irrational situation. The key insight is that each individual decision was correct in isolation, but the system itself was designed to produce an absurd outcome.
Winner Takes All
This isn’t just a thought experiment. It explains real-world dynamics in technology markets.
In tech, we’ve seen the same pattern again and again. Microsoft dominated operating systems. Google owns search. Facebook captured social networking. Amazon runs e-commerce infrastructure. These aren’t close competitions - the winners capture the vast majority of the value in their markets.
This happens because of network effects and economies of scale. The more users a platform has, the more valuable it becomes, which attracts more users, which makes it more valuable. The biggest player can invest more in infrastructure, pay more for talent, and offer lower prices. At a certain point, the gap becomes insurmountable.
This creates a rational expectation that AI and Large Language Models will follow the same pattern. If there’s going to be one dominant AI platform that captures most of the value, then not competing for that position means losing everything. The rational move is to invest heavily, right now, to be that winner.
The AI Trap
This is exactly what we’re seeing with AI and LLMs. Companies are pouring billions into the space. OpenAI raised $6.6 billion at a $157 billion valuation. Microsoft has invested over $13 billion in OpenAI. Google, Meta, Amazon, and others are each spending billions on their own efforts.
The logic is sound. If winner takes all, and you don’t invest, you lose. If you do invest and win, you capture a market that could be worth trillions. The expected value calculation says to invest.
But here’s where it becomes a dollar auction. Once you’ve invested billions, stopping means losing everything you’ve put in so far. The rational move is to keep investing to protect your initial investment. Even if your chances of winning are declining. Even if the total amount being invested by everyone starts to exceed the realistic size of the market.
The math is starting to look uncomfortable. If you add up the investments and implied valuations across all the major players, the total easily exceeds $500 billion. That’s a lot of value that needs to be created just to break even. It’s possible - AI is genuinely useful - but it’s also possible that we’re collectively bidding $20 for a $1 bill.
The parallel to the dollar auction is exact. Each company’s decision to invest more is rational given what they’ve already invested. But collectively, the industry may be heading toward an outcome where the total invested far exceeds the total returns. The system has a built-in mechanism that drives rational actors toward irrational outcomes.
The Uncomfortable Truth
Unlike the dollar auction where you can choose not to participate, companies feel they have no choice. If you’re Microsoft, Google, or Meta, not investing in AI is not an option. Your competitors are investing, your shareholders expect you to invest, and the risk of being left behind is existential.
The question isn’t whether AI will create value. It will. The question is whether the investments will justify the returns, or whether we’re watching the tech industry bid $20 for a $1 bill. Each company is making the rational decision to stay in the auction. But that doesn’t mean the outcome will be rational.
In the original dollar auction experiments, the game ends when someone finally refuses to keep bidding. They take their loss and walk away. The question is: who will be the first to blink in the AI race, and how much will they have lost when they do?