“Who is now making the profits? Who is taking it all home? And that question to be answered is super easy: it is starting at the bottom.”
That’s what I said in the video. My main article says the money in AI lands first at the bottom of the ladder: chips, power and cooling. That’s why I own a data-centre infrastructure company. Here is the strongest honest case against that view.
“The tokens will become a lot cheaper.” That can hurt the builders.
My own bet is that within five years tokens will cost half or less of today’s price. Think through what that means.
The research group Epoch AI found that the price of reaching a given level of AI performance has been falling between 9 and 900 times per year, depending on the task.
In the video I gave the example myself: instead of 10 tokens to complete a task, you might need five. If AI keeps getting that much more efficient, the world may need fewer chips, fewer data centres and less cooling than today’s plans assume. Then the bottom of the ladder isn’t the safest rung. It’s the one with the most to lose.
The spending is enormous, and it has to pay back
Alphabet, Amazon and Meta alone plan to spend roughly $545–570 billion on capital investment in 2026. Much of that becomes revenue for chip and equipment suppliers today.
Companies don’t keep spending at that pace forever. If AI revenue at the top of the ladder doesn’t grow fast enough, budgets get cut, and suppliers feel it first.
“Heavily subsidised”? The subsidy cliff may never come.
In the video I said we are now heavily subsidizing AI usage. The evidence is mixed.
Bloomberg and the Financial Times have both reported that Anthropic made an adjusted operating profit in the second quarter of 2026. In Nvidia’s August results, CEO Jensen Huang described the tokens its systems produce as productive and profitable.
If the labs can make money at today’s prices, there may be no dramatic stage two, just a slow, competitive business.
“A backlog of two years”? Less visibility than it seems.
That’s what I said about Vertiv. Its last published backlog was $15.0 billion at the end of 2025, roughly one year of its expected 2026 sales. In its first- and second-quarter 2026 earnings releases and presentations, the company didn’t publish an updated backlog figure.
So investors know less about the next few years than a “years of orders” story suggests. And orders can be delayed or cancelled when customers change plans.
“Cloud computing is next.” Maybe it’s the real winner.
I said that if you look up the ladder, cloud computing will be next. There’s a stronger version of that: the long-term winners may be the companies that sell AI as a service. They profit when AI gets cheaper, while hardware suppliers get squeezed.
In the latest quarter, Google Cloud grew 82%, Azure 43% and AWS 37%. Those businesses sit higher on the ladder and capture recurring revenue.
The bottom line of this side
The “picks and shovels” idea worked brilliantly in the build-out phase. But if AI keeps getting cheaper and the spending wave slows, the bottom of the ladder could be where the disappointment lands first.
Keep reading
- The War Was the Headline. The Bond Market Was the Story.
- What You Actually Own When You Buy a Gold ETF
I’m not a financial adviser. This is my own investing journey, not financial advice. Any investment decision is your own.