🔍 Read the full analysis: AI Data Centers Need Capital And Power. Who Pays? on ThorstenMeyerAI.com
TL;DR
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the third quarter of 2026, more than twice the year-earlier amount, according to figures compiled from company filings. They are financing the build-out with operating cash, debt, leases, partnerships and share sales, while the costs of long-term leases and power contracts can place risk on investors and electricity customers.
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 totals compiled from company filings. As AI data center construction absorbs more cash, the companies are turning to debt, leases, partnerships and share sales as well as their own operating income—leaving investors and, potentially, electricity customers exposed to some of the costs.
The financing moves reported this week show the scale of the demand for AI data center power. Meta sold $25 billion of common stock and $12.5 billion of preferred shares, while Microsoft borrowed $25 billion, including bonds with maturities extending as far as 2066. Amazon disclosed Friday that it had arranged a loan of up to $80 billion, secured by servers it plans to purchase. The figures and descriptions are reported by ThorstenMeyerAI.com based on company disclosures.
Quarterly property and equipment spending was reported at $77.3 billion for Amazon, $51.4 billion for Alphabet, $39.4 billion in cash for Microsoft and $33.9 billion for Meta. Microsoft’s figure rises to $45.2 billion when new finance leases are included. The four-company total of about $202 billion compares with about $97 billion in the same quarter of 2025; the source says this is its calculation from company cash flow statements.
Company forecasts point to continued spending. Amazon expects about $220 billion in cash capital spending for 2026; Alphabet forecast $195 billion to $205 billion, Microsoft about $170 billion for the calendar year, and Meta $135 billion to $145 billion including finance-lease payments. These are company estimates, not final annual results. The source says Amazon characterized the spending increase as primarily reflecting AI investments.
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 financing choices matter because they decide who bears the risk if demand for AI services, or revenue from them, falls short of expectations. Borrowing puts repayment obligations on the companies and gives bondholders a claim on future cash. Leasing and special-purpose property companies can move ownership and some financing outside a technology company’s direct balance sheet, but they do not remove the underlying cost of using the facilities.
The source calculates that the four companies had about $1.02 trillion in future lease payments for leases not yet started as of September 30, 2026, based on their quarterly filings. It says this was more than twice their combined long-term debt. Those lease commitments, alongside power contracts, can create long-lived obligations even as the servers inside a facility become outdated.
That mismatch can extend beyond the technology companies. Investors, lenders, property owners and electricity customers may be affected if projects are delayed, facilities are underused, or power infrastructure must be paid for regardless of how much computing demand materializes. The scale of these commitments makes the question less about whether the companies can fund construction today and more about who absorbs losses or higher costs later.
Popular right now
Philips Café Aromis Automatic Espresso Machine with LatteGo Pro Milk System
De'Longhi Dinamica Plus Fully Automatic Espresso Machine
KitchenAid Artisan 5-Qt Tilt-Head Stand Mixer – Dried Rose (KSM150PS)
As an affiliate, we earn on qualifying purchases.
Spending Outruns Quarterly Cash Generation
Technology companies historically funded data centers largely from cash generated by advertising, software and online retail. The reported 2026 figures show a more capital-intensive phase. Amazon, Alphabet and Meta each spent more on property and equipment during the quarter than their businesses generated in operating cash, according to the source’s comparison. Microsoft generated $58 billion in operating cash against $39.4 billion of cash capital spending, leaving about $18.6 billion by the source’s arithmetic.
The source also reports negative free cash flow over the stated periods: Amazon’s was minus $42.4 billion over the previous 12 months, while Alphabet’s was minus $9.1 billion and Meta’s minus $2.1 billion for the quarter. These figures use different time windows and should not be treated as directly comparable. The source says Amazon, Alphabet and Meta raised about $225 billion in long-term debt during the first nine months of 2026.
Capital spending is only one part of the financing picture. The source describes five channels: operating cash, borrowing, leases, separate companies that own data center buildings, and issuing shares. Long-term electricity agreements add another commitment. It reports that data center leases can run up to 20 years and power contracts 15 to 20 years, while servers are typically written off over five to six years. The durations cited are from the source’s account, not a single standardized contract for every project.
„through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.“
— Meta, in a statement dated October 26, as quoted by ThorstenMeyerAI.com
Revenue and Long-Term Obligations
The reported spending and financing figures describe current plans and commitments, but they do not establish how much revenue the new AI capacity will produce. The source raises the possibility that AI revenue may grow more slowly than the bills; that is a risk scenario, not a confirmed outcome. It is also unclear how much of the planned capacity will be used, when projects will be completed, or whether companies will revise investment plans.
The reported $1.02 trillion in future lease payments is a sum attributed to the source’s reading of four company filings. The provided material does not detail each lease’s terms, counterparties, or how the amount may change as contracts begin or are modified. Nor does it quantify what share of power-system costs would ultimately fall on data center operators versus other electricity users.
Financing costs also vary. The source cites senior notes issued by newer GPU cloud provider CoreWeave at 8.5% to 9.75%, but does not give the issuance dates, maturities or comparison basis in the supplied material. Those rates should not be treated as representative of borrowing costs for the four larger companies.
Company Spending and Funding Updates
The next evidence will come from company filings and results as the 2026 fiscal year closes and firms report capital spending, operating cash flow and borrowing. Investors will be watching whether the companies maintain their current spending forecasts, particularly Alphabet’s expectation of a significant increase in 2027, and whether AI-related revenue develops alongside new capacity.
Further detail on loan terms, lease commitments, data center partnerships and power agreements would help show how much risk each company retains and what has been passed to lenders, property owners or other parties. Until those disclosures and later results are available, the reported commitments show the scale of financing needs, but not the eventual returns or the final distribution of costs.
Key Questions
How much did the four companies spend in the third quarter of 2026?
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment from July through September, according to a total compiled from their cash flow statements. The corresponding source figure for the same period in 2025 was about $97 billion.
How are the companies paying for the build-out?
The reported financing sources include operating cash, bonds and loans, leases, partnerships and share sales. Long-term power contracts also support data center operations while creating future payment commitments.
What is the reported $1.02 trillion figure?
It is the source’s calculation of future payments on leases that had not yet started as of September 30, 2026, based on filings from the four companies. The source says the leases are mostly for data centers; the figure is a reported sum, not an estimate of cash already paid.
Does the spending prove that AI data centers are a bubble?
No. The spending and financing figures show that the build-out is capital-intensive, but they do not establish whether future AI revenue will justify it. The source frames the concern as who would bear the costs if demand or revenue falls short.
Could electricity customers pay part of the cost?
That remains uncertain. The source identifies electricity customers as possible parties exposed to the build-out’s costs, but the provided material does not quantify how power-system expenses are allocated between data center operators and other users.
Source: ThorstenMeyerAI.com
