Microsoft, Alphabet, Amazon and Meta now plan to invest roughly $720 billion to $745 billion this year. Until recently, they could cover almost all major spending from their own cash generation. Now they are issuing huge bond packages and relying more heavily on leases and project financing. This does not mean they are close to collapse. It means that capital-light technology companies are becoming capital-intensive infrastructure businesses.
The largest technology companies are not close to insolvency. They still make tens of billions of dollars and most of them have strong balance sheets.
What has changed is the nature of their business.
Mag 7 companies used to be valued as capital-light digital platforms. Software could be built once and sold to millions of customers at a relatively low additional cost.
AI is different. A data center needs land, electricity, cooling, networks, servers and chips that can become obsolete quickly. Big Tech is starting to look less like a software industry and more like power, rail or telecommunications. It has to build expensive infrastructure before it knows how intensively customers will use it.
The main question is therefore not whether AI demand will grow. It is whether AI revenue will grow fast enough to pay for today's construction race.
Futuro Invest has already covered longer depreciation schedules, hidden obligations and circular financing. The new evidence goes further:
This is no longer only an accounting debate. The pressure is visible in cash flow, new debt and the market price of credit risk.
The Mag 7 are Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia and Tesla. The biggest AI infrastructure bill is currently being paid by Microsoft, Alphabet, Amazon and Meta.
Apple follows a different AI investment model. Tesla combines AI spending with factories, energy and robotics. The roughly $720 billion to $745 billion total therefore refers mainly to the four largest cloud operators, not all seven companies.
The total is not purely AI spending either. Amazon also includes logistics, robotics and satellites. Even so, the companies themselves identify cloud and AI infrastructure as the main reason for the sharp increase.
Current plans are approximately:
The combined total is roughly $720 billion to $745 billion in one year.
The definitions are not identical. Microsoft offers the clearest example. It first expected about $190 billion of capital expenditure. It then extended the estimated useful life of data center buildings from 15 to 25 years. Some future data center leases will consequently move from finance leases to operating leases. Reported capex falls to about $175 billion, even though Microsoft explicitly said that its underlying investment plan has not changed.
For an ordinary investor, the lesson is simple. A lower reported number does not necessarily mean that less infrastructure is being built. Part of the bill may simply appear in a different accounting line and at a different time.
Free cash flow is, in simple terms, the cash generated by the business after paying for the buildings, servers, chips and other assets needed to operate it.
Alphabet reported excellent growth. Google Cloud revenue rose 82 percent. At the same time, quarterly capex reached $44.9 billion and free cash flow fell to negative $5.9 billion.
Amazon's AWS revenue grew 37 percent, its fastest pace in 18 quarters. Yet trailing twelve-month free cash flow moved from a positive $18.2 billion to a negative $7.6 billion.
Meta still produces large advertising profits, but second-quarter free cash flow was only $784 million, down from $8.5 billion a year earlier.
Microsoft remains the most resilient of the four. It generated $55.4 billion of operating cash and $19.6 billion of free cash flow in the quarter. Capital expenditure still reached $41 billion, and roughly two thirds went into shorter-lived assets, mainly CPUs and GPUs.
The figures do not cover exactly the same periods, so they are not a ranking. They do show the same trend. Investment is growing faster than ordinary cash generation.
A bond is a loan split among many investors. The company receives cash today and promises to pay interest and return the principal later.
Issuing bonds is not a sign of distress by itself. A strong company may rationally borrow long term and preserve its own cash reserves. Debt also avoids diluting current shareholders.
The current scale is unusual:
The documents generally describe the proceeds as being for general corporate purposes, not exclusively AI. The timing alongside record infrastructure spending is nevertheless clear.
The Bank of England noted that five leading AI hyperscalers represented only about 3 percent of outstanding US investment-grade debt at the end of 2025. By early May 2026, they represented more than 15 percent of new issuance for the year. Barclays estimated that AI could require about $240 billion of investment-grade bond financing in 2026 alone.
The AI race has moved from income statements into the bond market.
Many data centers are not built directly on the balance sheet of Microsoft, Amazon or Meta. A separate project company may own the facility. A private credit fund, insurer or institutional investor finances it. The technology company then signs a long lease or minimum capacity commitment.
The cloud company avoids paying the full cost today. A large upfront investment becomes a long series of future payments.
The BIS warns that these structures connect hyperscalers with private credit funds, insurers and banks. If the project works, that is not a problem. If capacity remains unused, losses can travel through several links at once.
This is where the story connects with our article about private credit moving into insurance companies. The risk did not disappear. It moved from a visible bank loan into leases, project companies, private funds and potentially insurance balance sheets.
Investors can no longer focus only on revenue growth and earnings per share. They will increasingly ask how much cash remains after investment.
When free cash flow falls, there is less money for buybacks, dividends and acquisitions. High valuations become more sensitive to disappointment.
It would still be wrong to treat all companies alike. Google Cloud grew 82 percent, AWS grew 37 percent and Azure grew 43 percent. These are real revenues, not only promises. If that growth persists, today's investment could create powerful future cash flow.
The market is therefore likely to become more selective. The winner may not be the company with the largest data center, but the company that turns each invested dollar into recurring, profitable revenue fastest.
The flood of technology bonds has three effects.
First, investors have limited capital. When hundreds of billions of dollars of new debt arrive, issuers may eventually need to offer higher yields to find buyers.
Second, technology firms become more sensitive to interest rates. A company funded mostly from internal cash can tolerate high rates more easily than one that regularly needs new bonds and leases.
Third, losses can travel beyond technology stocks. The debt is held by pension funds, bond funds, private credit, banks and insurers. If data center utilization disappoints, the damage will not appear only in Nvidia's share price.
This is not a direct default threat to US Treasuries. However, heavy corporate issuance competes for the same investor capital. Together with greater demand for energy, construction and power generation, it may help keep long-term yields higher than technology companies would prefer.
Favorable: AI demand keeps growing quickly. Data centers fill up, compute prices remain healthy and revenue catches up with investment. Debt stays a cheap tool and the buildout proves justified.
Middle path: AI grows, but returns take longer. Free cash flow remains weak for several years, bond issuance continues and stocks split between real winners and companies holding expensive unused capacity.
Stress: Construction runs ahead of demand. Compute rental prices fall, weaker operators cannot meet obligations and guarantees are activated. CDS spreads widen, bonds fall and the largest firms cut capex. The shock reaches chipmakers, power suppliers, construction companies, private credit and insurers.
No single indicator settles the argument. But falling free cash flow, rising debt and larger long-term obligations together form a signal that can no longer be ignored.
AI can be the most important technology shift of this generation and still be a poor investment for some of the companies building its infrastructure. Both can be true at the same time.
The Mag 7 are not close to running out of money. Their free cash flow has, however, reached the limit of what the current construction pace can finance internally. That is why bonds, leases, project companies and credit guarantees are expanding.
The crucial point is still ahead of us. The company that spends the most will not automatically win. The winner will be the one that fills new capacity quickly and converts it into real, recurring cash flow.
Note: This article is for informational purposes only and does not constitute investment advice. Investing in financial markets involves risk, and investors should conduct their own analysis before making any investment decision.