TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Amazon, Microsoft, Alphabet and Meta reported about $202 billion in combined property and equipment spending for July through September 2026, more than twice the year-earlier total. Their funding mix includes operating cash, debt, leases, partnerships and equity, while about $1.02 trillion in signed but not yet active leases adds future obligations. The figures come from company filings and calculations based on them; the timing and ultimate allocation of some costs remain uncertain.
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the three months through September, more than double the roughly $97 billion recorded a year earlier, according to the companies’ filings and a calculation combining them. The spending, much of it for data centers, servers and chips, is increasingly being financed through debt, leases and equity as well as cash generated by their businesses.
The figures underline the scale of the investment: Amazon reported $77.3 billion in property and equipment purchases, Alphabet $51.4 billion, Microsoft $39.4 billion and Meta $33.9 billion. These are company-wide capital spending figures, not a precise measure of AI spending. Amazon also builds warehouses; the company said its increase primarily reflected investment in artificial intelligence. Amazon expects about $220 billion in cash capital spending for 2026, while Alphabet maintained its 2026 forecast of $195 billion to $205 billion and said it expects spending to rise significantly in 2027.
Cash generation did not cover the quarterly spending at three of the four companies under the source’s comparisons. Amazon’s cash purchases exceeded its operating cash flow by $34.9 billion, while Alphabet reported quarterly free cash flow of negative $9.1 billion and Meta negative $2.1 billion. Microsoft reported $58 billion in operating cash flow against $39.4 billion in property and equipment purchases. These comparisons use different cash flow measures in some cases and should not be read as a complete accounting of how each company funds its investments.
Borrowing and other arrangements are adding to the funding mix. Amazon, Alphabet and Meta raised about $225 billion in long-term debt in the first nine months of 2026, according to a sum of their cash flow statements. Microsoft sold $25 billion in bonds this week, including debt with a maturity extending to 2066. Meta said on October 26 that it plans to fund AI investment through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships and equity capital. It also sold $25 billion of common stock and $12.5 billion of preferred shares this week, according to the source material.
Capital flows · AI infrastructure · October 2026
Where Does the Money for the AI Data Center Boom Come From?
Big Tech’s build-out is being paid for through a mix of business cash, borrowing, leases, partnerships and share sales. The spending is accelerating; the ultimate cost and return remain uncertain.
01 / The scale
More than twice the prior-year spend
Company-wide property and equipment figures include AI infrastructure, but also other facilities and equipment. They are not a precise measure of AI spending.
02 / Company filings
Who spent what in the quarter?
Reported purchases of property and equipment for July–September 2026, alongside the same period a year earlier.
Up from $35.1B. The company said the increase primarily reflected investment in artificial intelligence.
Up from $24.0B. The company expects spending to rise significantly in 2027.
Up from $19.4B. Its July–September quarter was fiscal Q1 2027.
Up from $18.8B. Meta described a balanced financing mix for AI investment.
Figures are company-wide capital spending, not an exact AI-only total. Amazon also builds warehouses. Microsoft’s reported operating cash flow was $58B against $39.4B in purchases; comparisons of cash coverage across firms use different measures.
03 / The funding stack
Five channels pay for the build-out
Operating profits have long funded infrastructure. At today’s scale, companies are combining several sources of capital and commitments.
Operating cash
Cash generated by advertising, software, commerce and cloud businesses.
Borrowing
Bond and loan investors provide funds in exchange for repayment and interest.
Leases
Facilities can be used now while payments extend into future years.
Shared ownership
Landlords or special-purpose companies may fund buildings used by tech firms.
New shares
Stock sales raise capital; existing shareholders bear potential dilution.
“through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital” Meta · statement dated October 26, 2026
04 / Who carries the cost?
Financing shifts risk across people and time
The headline lease estimate is a future commitment. It is not equivalent to debt already owed or cash already paid.
Shareholders
May absorb dilution when companies issue new stock.
Lenders
Take credit risk when they lend or buy company bonds.
Property owners
May finance facilities and depend on long-term tenants.
Power systems
Utilities and customers may face effects from new supply and grid demand.
05 / The timing mismatch
Equipment can age before obligations end
Servers: about 5–6 years
Equipment may be written off or outdated within a relatively short period. New generations can change the economics of an installed fleet.
Leases & power: 15–20 years
Some financing and service obligations last much longer than the equipment. Slower AI revenue growth could leave companies and partners paying for capacity after assets lose value.
At the sector’s higher-cost edge, the source cites CoreWeave senior notes with stated rates of 8.5%–9.75%. Those rates alone do not establish repayment ability or the likelihood of losses.
06 / What remains unknown
Spending totals cannot settle the bubble question
Filings describe spending and financing. They do not show how quickly AI products will earn revenue or whether returns will cover the build-out.
Allocation is still unclear
The source does not provide a company-by-company breakdown of the $1.02T lease estimate here. It is also unclear how much reported spending supports AI versus other facilities and equipment.
Demand will decide
The outcome depends on future demand, infrastructure utilization, financing terms and equipment value. The source figures do not establish how costs will ultimately be divided among companies, investors, property owners, utilities and customers.
07 / What to watch next
Filings will reveal the next commitments
Subsequent earnings reports and financing disclosures can show whether capital plans and obligations keep expanding.
Spending outlooks
Track whether Amazon’s roughly $220B 2026 cash spending outlook changes and whether Alphabet’s 2027 spending rises as indicated.
Cash, debt & leases
Look for how much investment is covered by operating cash and whether borrowing and lease commitments continue to grow.
Revenue & utilization
Investors and customers will watch whether AI services generate enough revenue to support the infrastructure being built.
Who Carries the Infrastructure Cost
The financing choices matter because they distribute the cost of AI infrastructure across different parties and over different time periods. Shareholders absorb dilution when companies issue stock; lenders and bond investors take credit risk; and landlords or special purpose companies may finance buildings that technology firms use under long-term agreements. Utilities and electricity customers may also be affected as large data center projects require new power supply and grid capacity, though the scale and allocation of those costs are not established in the source figures.
The source estimates that the four companies had about $1.02 trillion in future lease payments for signed leases that had not yet begun as of September 30, based on their quarterly filings. The commitment figure is not the same as debt already owed or cash already spent. Still, it shows how financing commitments can grow beyond the liabilities most visible on a balance sheet.
The central risk is a mismatch between the life of the equipment and the length of the obligations. The source says servers are typically written off over five to six years, while data center leases and power contracts can last 15 to 20 years, and some bonds extend for decades. If revenue from AI services grows more slowly than expected, companies and their financing partners could be left covering costs after particular equipment has lost value or become outdated.
Products Worth Considering
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
Anker SOLIX C1000 Gen 2 Portable Power Station
DJI Mic 3 Wireless Microphone System (2 Transmitters, 1 Receiver, Charging Case)
As an affiliate, we earn on qualifying purchases.
From Cash-Funded Builds to Mixed Finance
For years, the largest technology companies could fund data center construction from profits generated by advertising, software and online commerce. In 2026, the scale of announced infrastructure investment has led them to use several funding channels at once. The source describes five: operating cash, borrowing, leases, building-owning side companies and new shares, with long-term power contracts shaping operating costs as well.
Quarterly spending rose at all four companies compared with the same period a year earlier: Amazon from $35.1 billion to $77.3 billion, Alphabet from $24.0 billion to $51.4 billion, Microsoft from $19.4 billion to $39.4 billion, and Meta from $18.8 billion to $33.9 billion. The combined comparison is calculated from company-reported figures. Microsoft’s July–September quarter is its fiscal first quarter of 2027.
The risks may vary across the industry. The source points to newer GPU cloud providers such as CoreWeave, whose senior notes carry stated interest rates of 8.5% to 9.75%, as one example of higher-cost financing at the sector’s edges. Those rates alone do not establish a company’s ability to repay or the likelihood of losses. The source’s broader point is that the eventual risk depends on who owns the facilities, lends against them, or has contracted to pay for power and capacity.
“through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital”
— Meta, in a statement dated October 26, 2026
How Fast AI Revenue Will Grow
The filings and announcements describe spending and financing, but they do not establish how quickly AI products will generate revenue or whether that revenue will cover the investments. The $1.02 trillion lease total is a combined calculation of future payments under leases that had not started by September 30; the source does not provide a company-by-company breakdown here. It is also unclear how much of each company’s spending is specifically for AI rather than other facilities and equipment.
Whether the investment amounts to a bubble cannot be determined from spending totals alone. The outcome will depend on future demand, infrastructure utilization, financing terms, and the value of equipment over time. The source figures do not show how potential costs will ultimately be divided among technology companies, lenders, property owners, investors, utilities and customers.
Filings Will Show the Next Commitments
The next evidence will come from companies’ subsequent earnings reports, financing disclosures and lease commitments. Those updates can show whether planned capital spending rises or falls, how much is funded from operating cash, and whether debt and lease obligations continue to expand. Alphabet has said it expects a significant increase in capital spending in 2027; Amazon’s stated 2026 cash capital spending outlook is about $220 billion.
Investors and customers will also be watching whether AI services produce enough revenue to support the infrastructure being built. Until those results become clearer, the reported spending and commitments establish the scale of the build-out, while its long-term returns and final allocation of costs remain open questions.
Key Questions
How much did the four companies spend in the third quarter of 2026?
Amazon, Microsoft, Alphabet and Meta reported a combined about $202 billion in property and equipment purchases for July through September 2026, based on a sum of their filings. The measure includes non-AI spending as well as AI infrastructure.
Where is the money for data centers coming from?
The companies are using operating cash, debt, leases, partnerships and equity issuance. Long-term power contracts also affect the cost of running data centers.
What does the $1.02 trillion lease figure mean?
It is the source’s combined estimate of future payments on signed leases that had not started as of September 30, 2026. It is a measure of future commitments, not cash already spent or debt already due.
Does the spending prove there is an AI investment bubble?
No. Spending totals show the scale of investment, but do not establish whether it will earn an adequate return. That depends on future AI revenue, infrastructure use, financing costs and the value of equipment over time.
Source: Thorsten Meyer AI
