Wall Street keeps celebrating every new AI spending announcement.
Another data center.
Another GPU order.
Another capex increase.
Another quarter where investors cheer because management promises to spend even more.
Everyone is focused on revenue.
Almost nobody is watching the cash.
That may become the biggest mistake of this AI cycle.
The AI boom is no longer being funded entirely out of profits.
It is increasingly being financed with debt, long-term contractual obligations, and future cash flows that have not been earned yet.
The numbers are becoming difficult to ignore.
The world’s largest technology companies are expected to spend around $700 billion on AI infrastructure during 2026 alone, with estimates continuing to climb as the AI arms race accelerates.
But capex is only part of the story.
The hidden number is much larger.
According to Nikkei, five of America’s biggest technology companies have accumulated roughly $1.65 trillion of off-balance-sheet AI commitments, an amount that has exploded roughly eightfold in just four years and now exceeds their reported debt. These include long-term data center leases, chip purchase agreements, cloud infrastructure contracts, and other obligations that still require cash even if they don’t immediately appear as traditional debt.
Investors usually watch debt.
They rarely watch commitments.
Eventually both have to be paid.
The financing model is already changing.
Instead of relying primarily on internally generated cash, hyperscalers have dramatically increased borrowing.
Debt issuance has climbed to roughly four times the pre-AI average, with more than $120 billion raised this year alone to help finance AI expansion.
The cash flow picture is changing just as quickly.
Alphabet shocked investors by reporting its first negative free cash flow since becoming a public company, roughly $5.9 billion negative, despite delivering record revenue and 82% cloud revenue growth.
Instead of slowing down, management raised 2026 capital expenditure guidance again, from $180–190 billion to $195–205 billion.
The stock immediately sold off because investors suddenly cared less about revenue and far more about how much cash AI is consuming.
Alphabet isn’t alone.
Amazon continues pouring enormous sums into AI infrastructure.
Microsoft continues expanding data centers globally.
Meta keeps increasing AI investment.
Oracle has become one of the most leveraged beneficiaries of the AI buildout while aggressively expanding infrastructure.
Tesla is now rapidly increasing AI-related capital spending as Elon Musk says the company should spend “as fast as we can,” even after reporting negative free cash flow.
Individually these companies still have strong balance sheets.
Collectively, however, something important has changed.
For decades Big Tech generated more cash than it knew what to do with.
Today much of that cash is immediately being reinvested into AI.
When internally generated cash no longer covers expansion, companies borrow.
When borrowing no longer covers expansion, they issue equity.
When both become expensive, projects slow.
That is how solvency risks begin.
Not because companies suddenly go bankrupt.
Because balance sheets slowly become dependent on future success instead of current cash generation.
Everything now depends on one assumption.
That AI will eventually generate enough profits to justify today’s spending.
That assumption is carrying an extraordinary amount of weight.
Every new data center.
Every long-term lease.
Every GPU cluster.
Every power agreement.
Every financing package.
Every debt offering.
Every off-balance-sheet commitment.
They all assume AI demand keeps growing fast enough to produce returns large enough to pay for an infrastructure buildout unlike anything the technology sector has ever attempted.
If those returns arrive, today’s spending will look brilliant.
If they don’t, today’s investments become tomorrow’s liabilities.
Markets usually recognize solvency problems long before companies actually run out of money.
Investors don’t wait for bankruptcy.
They sell when they realize cash generation is no longer keeping up with commitments.
That is why this story matters.
The AI boom has created some of the richest companies in history.
Ironically, it is also forcing those same companies to spend cash faster than they can replace it.
The market still believes AI spending automatically creates value.
History suggests infrastructure booms rarely end because people stop believing in the technology.
They end when the financing behind the boom becomes harder to sustain.
That is the risk investors should be watching.
Not whether AI changes the world.
Whether the companies racing to build it can keep paying for it.