🔍 Read the full analysis: The Economics Of AI Data Centers: Funding, Leases And Power Bills on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the third quarter of 2026, more than twice their spending a year earlier, according to figures summed from company filings. The companies are combining cash, bonds, loans, leases, partnerships and equity to fund data centers, while long-term leases and power contracts can leave investors, landlords and electricity customers exposed if AI revenues lag.
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the third quarter of 2026, more than twice their spending a year earlier, according to a total calculated from their filings. As the AI data-center build-out grows beyond what operating cash covers for several of the companies, they are using new borrowing, leases, partnerships and equity alongside cash to pay for infrastructure.
The quarter’s reported property-and-equipment spending was $77.3 billion at Amazon, $51.4 billion at Alphabet, $39.4 billion at Microsoft and $33.9 billion at Meta. Those amounts compare with roughly $97 billion combined in the same quarter a year earlier. Microsoft’s figure rises to $45.2 billion when new finance leases are included. The totals and comparisons are based on company cash-flow statements and filings, rather than a separately reported industry measure.
Recent financing moves show how the companies are covering the bill. The source report says Meta sold $25 billion of its own stock and $12.5 billion of preferred shares, Microsoft issued $25 billion in bonds, including debt extending to 2066, and Amazon disclosed a loan of up to $80 billion secured by servers it plans to buy. Meta said on October 26 that it expects to fund AI investment through “a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.”
Cash generation differs across the group. In the quarter, operating cash flow was reported at $42.4 billion for Amazon, $42.3 billion for Alphabet, $58 billion for Microsoft and $32.8 billion for Meta. Comparing those figures with cash property-and-equipment spending, Microsoft had about $18.6 billion left by the report’s arithmetic; the other three spent more than their quarterly operating cash flow on that category. These comparisons do not include every corporate cash need or financing arrangement.
Infographic 01 · One quarter, four companies
The bill doubled. The cash did not.
What Amazon, Microsoft, Alphabet and Meta paid for buildings, servers and chips from July to September 2026, set against the cash their businesses brought in.
AmazonQ3 2025: $35.1B spent
Spent $34.9B more than came in
AlphabetQ3 2025: $24.0B spent
Free cash flow −$9.1B
MicrosoftQ3 2025: $19.4B spent
$18.6B left over
MetaQ3 2025: $18.8B spent
Free cash flow −$2.1B
Cash purchases of property and equipment as reported; not all of it is for AI (Amazon also builds warehouses). Gaps for Amazon and Microsoft are my arithmetic; free cash flow for Alphabet and Meta as the companies report it. Microsoft’s quarter is its fiscal Q1 2027 (July–September 2026).
Sources: Amazon results, 29 Oct · Microsoft 10-Q, 28 Oct · Alphabet results, 28 Oct · Meta results, 21 Oct. Debt raised: cash flow statements, nine months 2026.
Powered by Thorsten Meyer AI
Who Carries the Data-Center Costs
The funding mix matters because it determines who bears the risk if AI services do not generate revenue quickly enough to support the infrastructure being built. A company can borrow directly, rent a facility, or work with a separate owner that finances the building. Each approach changes where the obligation appears and who must meet payments, but does not remove the underlying cost of land, servers, construction and electricity.
The report calculates that the four companies had about $1.02 trillion in future lease payments for leases that had not yet started as of September 30, based on its sum of their quarterly filings. That figure is more than twice their combined long-term debt, according to the source’s calculation. It is a measure of future contractual commitments, not a bill due all at once. The distinction matters for investors trying to judge how much flexibility companies retain if demand or construction plans change.
Power adds another layer. Data centers need reliable electricity, and the source describes power contracts lasting 15 to 20 years, alongside leases that can run as long as 20 years. Those obligations can last well beyond the useful life of a server, which the report says is typically written off over five to six years. If revenue falls short or equipment becomes outdated before longer contracts expire, companies and their financing partners may still owe payments.
Products Worth Considering
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
From Cash-Funded Growth to Financing
For years, the largest technology companies could fund data centers from profits generated by advertising, software and online shopping. The source report says that pattern has changed as spending on AI computing capacity has accelerated. Amazon expects about $220 billion in cash capital spending for 2026; Alphabet maintained guidance of $195 billion to $205 billion and expects a “significant” increase in 2027. Microsoft expects about $170 billion for calendar 2026, while Meta forecast $135 billion to $145 billion, including finance-lease payments.
These company forecasts are not identical accounting measures: some include lease payments, and the companies build infrastructure for businesses beyond AI. Amazon, for example, also invests in warehouses, although it said the recent spending increase “primarily reflects investments in artificial intelligence.” The figures therefore indicate the scale of the investment cycle, but should not be read as a precise, like-for-like tally of AI-only spending.
Financing also extends beyond the four largest firms. The report points to newer GPU cloud provider CoreWeave, whose senior notes it says carry interest rates of 8.5% to 9.75%, as an example of a borrower facing higher financing costs. It also identifies building-owning companies and other financing partners as part of the system. Their exposure can differ from that of the technology tenants, and the source does not provide a complete accounting of every party’s commitments.
““Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.””
— Meta, in a statement on October 26, 2026
Revenue and Long-Term Commitments
The disclosed spending and financing show the scale of investment, but they do not establish how quickly AI products will produce enough revenue to cover it. The source does not provide a common forecast for returns across the four companies, or a full breakdown of which data centers, chips and other assets serve AI workloads rather than existing businesses.
It is also unclear how much of the $1.02 trillion in future lease payments could be changed, renegotiated or offset if projects are delayed or demand shifts. The total is the source author’s sum of four filings, not a single company-reported figure. Long-term power commitments and financing through separate property owners add exposures that may not be captured by looking at corporate debt alone.
The source frames the “bubble” debate as a question of who holds the risk, rather than presenting evidence that the investments will fail. Large technology companies have substantial businesses and access to capital, but the financial position of landlords, lenders, infrastructure providers and customers can differ. The eventual effects on electricity prices also cannot be determined from the spending figures alone.
Company Guidance and Financing Costs
The next near-term indicators are whether companies meet their 2026 spending forecasts and how they update plans for 2027. Investors will also be able to track new bond and loan disclosures, changes in lease commitments, and whether capital spending continues to outpace operating cash flow. Alphabet’s expected increase in 2027 spending is one stated signal to watch, though the company has not supplied a specific amount in the material provided.
For the broader financing picture, attention will also fall on the terms demanded by lenders and investors, including borrowing costs for companies that specialize in GPU cloud capacity. Future filings can clarify how much infrastructure is owned directly, leased, or held through partners. Until those details and revenue outcomes become clearer, the scale of commitments is confirmed more clearly than the returns they will generate.
Key Questions
How much did the four companies spend in the third quarter of 2026?
Amazon, Microsoft, Alphabet and Meta spent about $202 billion combined on property and equipment from July through September, according to a sum of their company filings in the source report. The figure was about $97 billion in the same period a year earlier.
Their infrastructure spending has grown so large that cash from operations did not cover quarterly property-and-equipment spending at Amazon, Alphabet and Meta, using the source report’s comparison. They are combining cash, debt, leases, partnerships and equity to finance the build-out.
What does the $1.02 trillion lease figure represent?
It is the source author’s total of future payments on leases that had not started at the four companies as of September 30, 2026. It is a calculated total from four quarterly filings, not a payment due immediately or a single figure reported by one company.
Who could be exposed if AI demand falls short?
Potentially exposed parties include the technology companies, lenders and bondholders, data-center landlords, infrastructure providers and electricity customers. The size of any losses or effects is not established by the spending figures and depends on contract terms, demand and future revenue.
Source: ThorstenMeyerAI.com
