.svg)
The demand is real and spending is enormous. The math on where the payback comes from is the part nobody is talking about.
The question behind the headlines
Most AI ROI conversations are about individual companies. Who is winning, who is cutting costs, who is shipping faster. Those stories are real, and companies like Revature are clearly benefiting from AI, as are many others.
But I want to zoom out. What is the return on the whole ecosystem? As of September 2026, global AI spending stands at $2.7 trillion. That is a number we have never seen before. So where does the payback come from?
To think it through, I looked at these five numbers.
Five numbers that frame everything
1. 105x growth in two years.
Anthropic grew from a $1 billion run rate in January 2025 to an unexpected $105 billion run rate by December 2026. That is unheard of. Nobody else is even close.
2. $2.7 trillion.
Is the global AI spend as of September 2026. The majority is going into infrastructure, services and software.
3. $720 billion.
Is the combined annual labor cost of the 36 enterprises that can potentially consume AI services the most.
4. Coding.
Is the first proven ROI scenario. The payback is real, and we see it day in, day out. Coding agents are the most successful and proven AI use case we have.
5. Productivity gains.
How much ROI they deliver is debatable. But are there gains? Absolutely, yes.
Those last two are the practical ROI that we can point to today.
Demand is Real
Starting with Anthropic. A 105x run-rate increase in two years is ludicrous. These are real, massive revenues, which means the consumption of AI services is real. And that is just one company.
Spending is real too, and it is enormous. The $2.7 trillion is flowing mostly into infrastructure, services and software.
But, here is what got me thinking.
Who would be the biggest consumers of AI services?
The 36 companies
Here is my list of the biggest consumers of AI services:
- The top 10 system integrators, the likes of Accenture, Infosys, Cognizant and TCS
- The Big Four consulting firms
- The top seven global banks
- Five healthcare companies
- Insurance providers
That is 36 companies. Their entire annual labor cost, combined and global, is $720 billion. That is just one-fourth of total global AI spend.
Read that again.
If you took the entire payroll of the most likely AI buyers on the planet, it would cover about a quarter of what the world has already spent on AI.
The Coding Reality Check
Coding is the best AI use case we know. The top 10 system integrators do most of the world's coding services, and they probably spend around $100 billion a year on programmers.
That does not even account for one company's revenue, Anthropic's. Our number one proven use case does not fully cover even one of the model builders.
Now take the hypothetical that will never happen: replacing every employee at all 36 companies. It would still account for just 27% of the $2.7 trillion.
Cost replacement alone cannot close the gap.
So where does the return come from?
I believe the only path to real ROI is AI-infused revenue acceleration.
Not doing the same work cheaper, but growing revenue in ways that were not possible before.
Will it happen?
Absolutely, yes.
Is it happening now?
No, it is not.
And that leaves me with the question I cannot shake:
Will the market have the patience for companies to start seeing revenue acceleration with AI?
Will the market wait, or will it pull back?
Nobody knows.
That is my thought. I would love to hear yours.
Do you think the market is patient enough?
Is revenue acceleration realistic?
Tell me in the comments.
.webp)
.webp)
