Something about the AI race keeps bothering me.
Every earnings season, the spending gets bigger.
The timeline for earning it back stays fuzzy.
Microsoft, Meta, Amazon, and Google are pouring hundreds of billions of dollars into AI infrastructure, building data centers, buying GPUs, expanding power capacity, and upgrading networks at a pace the industry has never seen.
The market keeps rewarding them because revenue is still growing.
But there’s a difference between growth and return on investment.
Take Microsoft.
Azure continues posting strong growth, but AI infrastructure isn’t cheap.
GPUs, data centers, networking, and electricity all put pressure on margins and free cash flow.
The same story is playing out across the hyperscalers.
No CEO wants to be the company that underinvested while everyone else built the next computing platform.
So the spending race keeps accelerating.
That creates an interesting problem.
The AI winners may not be the companies spending the most.
They could be the companies selling the picks and shovels.
Equipment makers like Applied Materials, Lam Research, KLA, and ASML are getting paid whether one AI model wins or ten.
Their order books stay full while hyperscalers wait for AI investments to generate enough revenue to justify the massive bills.
This is what makes today’s AI boom different.
The money is being spent now.
The returns are expected later.
Maybe much later.
If AI adoption keeps accelerating, the spending will look brilliant in hindsight.
If monetization takes longer than investors expect, today’s record capital spending could become tomorrow’s margin problem.
That’s the question Wall Street is wrestling with right now.
Not whether AI changes the world.
Whether it earns back the hundreds of billions being spent to build it.