Last Wednesday we made the case that AI’s gains accrue to users and builders, not to a permanent margin explosion for the companies using it.

We’ll learn more this week.

To this point, we’ve seen the clear gains continue to accrue to the primary AI chokepoint for world chip supply, Taiwan Semiconductor. 

Last week they reported record revenue and profit up 77%. 

Gross margin expanded to 67.7%.  Operating margin was up to 60.3%. Net profit margin was 55.6%.  The year prior, those margins were 58.6%, 49.6% and 42.7%, respectively. That’s a massive scale business doing 9 to 13 points of margin expansion in one year.

Now, let’s look downstream.

Netflix told shareholders it used generative AI in roughly 300 titles this year. Battle scenes, crowds, entire sequences. Its co-CEO said one documentary’s AI footage came “twice as fast and at half the cost,” and that without the tools, productions “would have left out those key shots.”

So AI is everywhere in the product. But revenue growth is decelerating, from 16% to 13% to 12% guided. The gains went to viewers, who get bigger shows, and to creators, who get bigger tools. They did not go to margins.

And IBM, you’ll remember, warned on its quarter because customers redirected software budgets to buy scarce servers and memory chips.

So, in companies producing inputs that are scarce (advanced chips, memory, power) the margins are historic. Where AI is abundant (content, enterprise software) competition hands the gains straight to the customer. Jamie Dimon said exactly this on his earnings call last week — the ultimate beneficiary is the customer.

The productivity benefits are arriving for users and end consumers of AI, though it’s not hitting the income statement (at least yet) for most companies.

That said, it’s early in Q2 earnings season. We’ll learn more this week.

Wednesday, we’ll hear from Google (Alphabet). This is the first big hyperscaler to report. And it’s all about the capex plan.  The Wall Street community continues (for yet another quarter) to speculate about a capex slow down.

The suppliers of the most advanced chips in the world say otherwise.  

The TSM report shows pedal-to-the-metal, and if we listen to Nvidia’s CFO two months ago, she told us “AI infrastructure spending is on track to reach $3 to $4 trillion annually by the end of this decade.” That’s per year! 

Google will be a big share of that $3-$4 trillion.

They’re already on the record to spend $180-$190 billion this year, with a “significant increase” signaled for 2027. The infrastructure investment is accelerating, not slowing.