Key insights
- The article discusses the investment paradigm for AI, contrasting proven success with the challenge of profit outpacing cost. It uses an analogy of expanding a machine's capacity to illustrate how initial high costs for AI development and deployment can lead to exponentially increasing returns. While acknowledging potential long-term overhead risks like component depreciation, the author argues that AI's returns may not have a ceiling, making traditional valuation metrics difficult. The core argument suggests that AI investments, despite upfront costs, offer potentially limitless upside due to rapidly accelerating returns, making it a compelling, albeit complex, investment thesis.
Investor paradigms clash
- Proven success
Vs
- Profit never outruns cost
To put this in perspective for the layman: You have a machine. This machine out performs all other machines youve had in the past. But its limited right now by its capacity to produce. You know that by increasing capacity, you can make more money. So you spend more money to expand the capacity - returns start to increase. Every time you do this, the returns increase. Up front cost is high, so youre technically going further into debt aa you increase capacity, and it takes time for the profits to catch up.
The question is, when do you stop? Its like any buisness. If we expand our operation to another state, we make more money! So you expand.
But the difference with ai is how the returns dont seem to have a ceiling. That first capacity increase you bought for the machine has paid for itself 10x over and its rate of returns is more amplified every week. Its one thing to have a projected profit - and in nearly any buisness this can be be measured. But with ai its impossible, youre revising your ceiling daily, and it just goes higher and higher!
The only risk - as pointed out by mike burry - what about long term overhead? Components depreciate, they dont last forever. But is this a valid argument truly? If every dollar in initially has returned the investment already, then the worse that can happen is - you have to replace something which paid for itself already.
The most important part is relevance. If one provider suddenly offers something far greater - existing stuff could quickly be forgotten, but we are again in (most) situations, seeing returns by then anyway.