I use AI every damned day.
So this isn’t going to be one of those articles where some old geezer shakes his cane at ChatGPT and announces that computers have gone too far.
Quite the opposite.
I think AI is extraordinary.
But I also know something else.
Extraordinary can get bloody expensive.
And we’re beginning to get some remarkable numbers showing exactly how expensive.
Take Meta.
The company had a terrific second quarter on the surface. Revenue jumped 28% from a year earlier to $60.8 billion.
Then I got to the cash-flow numbers.
Meta generated $784 million in free cash flow during the quarter.
A year earlier?
$8.55 billion.
That’s not a typo.
Meanwhile, capital expenditures reached more than $31 billion for the quarter, and Meta now expects to spend between $130 billion and $145 billion on capital expenditures this year.
Meta isn’t going broke. Not remotely.
But when a company can increase revenue by 28% while free cash flow drops by more than 90%, I think we’re allowed to ask what the hell they’re spending all that money on.
A very large part of the answer is the infrastructure required for AI.
And Meta isn’t the biggest spender.
Amazon has now raised its expected 2026 capital spending to roughly $220 billion.
AWS is booming. Its second-quarter sales increased 37%, and Amazon says demand remains enormous.
But here’s the other number.
Amazon’s trailing 12-month free cash flow went from positive $18.2 billion to negative $7.6 billion.
Amazon itself says the enormous increase in property and equipment spending primarily reflects investment in artificial intelligence.
Think about that for a second.
These aren’t struggling startups desperately trying to catch the AI train.
These are some of the richest, most powerful corporations ever created.
And even they are discovering that building the future comes with one hell of a hydro bill.
Then I found my favourite number
$1.8 million.
According to reporting by the Financial Times, Amazon had an internal project using Anthropic’s Claude Sonnet to match author information with product listings.
Nothing terribly exotic.
It reportedly ended up costing about $1.8 million, ran roughly 860% over budget, and the overspending went undetected for months.
The project never shipped.
Amazon engineers were reportedly working afterward on automated guardrails to prevent similar runaway costs.
Now THAT got my attention.
Because if Amazon… the company that practically invented modern cloud computing… can lose track of what an AI project is costing, imagine what’s happening inside thousands of ordinary businesses currently being told they need an “AI strategy.”
Turns out we have some evidence.
KPMG surveyed organizations using AI and found only 26% had full real-time visibility into what their AI systems actually cost to operate.
Another 2026 survey, this one of 500 finance leaders at large American and British companies, found 79% had experienced AI-related cost overruns during the previous year.
There’s our story.
Not that AI doesn’t work.
Not that the AI boom is finished.
Not that Nvidia is about to start selling chips from the back of a van.
Something much more ordinary is happening.
The accountants have arrived.
For the first couple of years, the question was…
What can AI do?
Now another question is creeping into the boardroom…
What does the damn thing cost?
And eventually comes the question that matters most…
Did we make more money because we used it?
That’s a healthy development.
Because I’m sitting here at 74 years old using AI to research, write, brainstorm and build things I couldn’t have imagined doing this quickly a few years ago.
I don’t need somebody to convince me AI has value.
I see the value every day.
But usefulness and profitability aren’t the same thing.
The Internet changed the world.
It also produced Pets.com.
Railways transformed countries.
Plenty of railway investors still got their asses handed to them.
A revolutionary technology doesn’t guarantee that every company throwing money at it has suddenly become brilliant.
And that’s what makes this moment interesting.
The AI boom isn’t collapsing.
Hell, the Nasdaq closed at a record high today, helped by another burst of enthusiasm for AI stocks.
AMD just crossed the $1-trillion valuation mark for the first time.
Investors clearly haven’t abandoned AI.
They’re still throwing confetti.
But underneath the party, another conversation has started.
Companies are borrowing billions, building enormous data centres, buying mountains of chips and spending sums of money that would have sounded insane three years ago.
Sooner or later, those investments have to produce something more than impressive demonstrations and rising revenue.
They have to produce returns.
I don’t think we’re watching the end of the AI revolution.
I think we’re entering the part where the revolution has to start showing its receipts.
And frankly?
That’s when this gets really interesting.
The Recap…
AI isn’t running out of believers.
Meta, Amazon and the rest are still spending astonishing amounts building the infrastructure behind it.
But cash flow is getting squeezed, AI projects are blowing budgets, and many companies can’t even see their AI costs in real time.
The technology has proved it can do amazing things. Now comes the harder test… can businesses make the economics work?
The Gut-Punch…
AI passed the technology test.
Now it has to pass the accountant.
Source Credit
Financial results and corporate guidance: Meta Platforms Q2 2026 results and Amazon Q2 2026 results. AI spending and market context: Reuters. Enterprise AI cost visibility: KPMG AI Quarterly Pulse Survey, Q2 2026. AI cost-overrun survey: DoiT International. Amazon internal AI-project reporting: Financial Times.
If you enjoy thoughtful conversations, Canadian stories, and the occasional smart-ass observation about the world we’re living in, you’re in the right place.
Subscribe free and get new stories, insights, and observations delivered directly to your inbox.
No paywall.
No spam.
No nonsense.
Leave anytime with a single click.
I promise not to take it personally.