TL;DR
Get backup power and energy gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the July–September quarter, more than twice the year-earlier total, according to figures compiled from company filings. Their financing mix now includes cash, borrowing, leases, partnerships and share sales, while future lease payments and long-term power contracts create obligations that can last well beyond the life of the servers they support.
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment from July through September 2026, more than twice their reported total a year earlier, according to calculations from their filings. The spending surge is being financed through a mix of operating cash, borrowing, share sales, leases and partnerships, spreading the costs and risks of the AI data center build-out across more types of investors and counterparties.
The figures cover property and equipment, which include data centers, servers and chips but are not limited to AI. Amazon also builds warehouses. The source’s calculation puts combined spending at about $97 billion in the same quarter of 2025. Amazon reported $77.3 billion in the latest quarter, up from $35.1 billion a year earlier; the company said the increase primarily reflected AI investments. It expects about $220 billion in cash capital spending for all of 2026.
Alphabet reported $51.4 billion in property and equipment purchases, compared with $24 billion a year earlier. Its 2026 spending guidance remains between $195 billion and $205 billion, and it expects a “significant” increase in 2027. Microsoft spent $39.4 billion in its fiscal first quarter, covering July through September, while Meta spent $33.9 billion. The combined total and year-over-year comparison are the source author’s sums of company cash flow statements.
Recent financing moves show the range of funding sources. Meta said it sold $25 billion of common stock and $12.5 billion of preferred shares during the week described in the source. Microsoft issued $25 billion in bonds, including debt that matures in 2066. Amazon disclosed a loan facility of up to $80 billion secured by servers it plans to purchase. Separately, the source calculates that Amazon, Alphabet and Meta raised about $225 billion in long-term debt during the first nine months of 2026.
Capital flows · AI infrastructure · Q3 2026
AI Data Center Finance: Follow The Cash, Debt, Leases And Power Costs
Four technology giants accelerated property and equipment spending. Trace how cash, borrowing, equity, leases and partnerships distribute the cost of building AI infrastructure—and how long some obligations may last.
01 / The spending surge
Capital is moving into infrastructure
Reported property and equipment purchases include data centers, servers and chips, but are not exclusively AI spending. Amazon’s figure also includes warehouses.
Quarter ended September 30, 2026 · company filings
Company figures and combined totals are drawn from cash flow statements and filings. The combined amounts are the source author’s calculations.
02 / The financing mix
Follow the money—and the obligations
The build-out is supported by several channels at once. Each moves cash today while assigning repayment, ownership, or operating commitments to different parties.
Operating cash
Cash generated by advertising, software and online retail has historically funded infrastructure. Spending exceeded operating cash flow at three firms in the source’s comparison, though reported periods and measures differ.
Bonds & loans
Microsoft issued $25B in bonds, including debt maturing in 2066. Amazon disclosed a facility of up to $80B secured by servers it plans to purchase.
Share sales
Meta said it sold $25B of common stock and $12.5B of preferred shares during the week described in the source. Equity raises capital and can dilute existing shareholders.
Leases
Leases spread facility costs over time. The source calculates about $1.02T in future payments tied to signed leases not yet started, mostly for data centers.
Partnerships
Separate property-owning entities and strategic partnerships can distribute investment and returns among companies, investors and counterparties.
Power contracts
Long-term electricity agreements shape future operating commitments. The figures do not establish how much these deals may affect electricity prices or other customers.
“Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.”Meta · statement on October 26
03 / Time horizon
Infrastructure lives on different clocks
The source highlights a potential mismatch between the accounting lives of servers and the terms of the contracts and financing that support them.
Servers are typically written off over five to six years. Cited leases and power contracts can run much longer, while some bonds extend for decades.
If AI service revenue grows more slowly than costs and obligations, companies or financing partners would still need to cover the difference. The timing and size of any shortfall are unknown.
Newer GPU cloud providers also sit in the financing chain. The source says CoreWeave senior notes carry rates of 8.5% to 9.75%; it does not detail each note’s terms or performance.
Illustrative term comparison · maximum cited duration
04 / Exposure chain
Who carries the infrastructure costs?
Financing choices connect company spending to a wider network of capital providers, asset owners and operating counterparties.
Tech companies
Commit cash, issue shares and sign contracts.
Capital providers
Lenders, bondholders and investors supply funds.
Asset owners
Property partners and lessors own or finance sites.
Power system
Utilities and counterparties support long-term supply.
AI customers
Future service demand must support investment returns.
The source identifies lenders, investors, pension funds, building owners and electricity customers as potentially exposed. It does not quantify who would absorb losses or establish effects on customer prices.
05 / What filings can answer
Track the next spending cycle
Upcoming company filings and guidance can help readers compare infrastructure purchases with funding sources and expected returns.
Purchases vs. cash flow
Compare property and equipment purchases with operating cash flow, while checking that periods and definitions align.
Debt, equity & leases
Watch new debt issuance, share sales, lease terms and partnership disclosures for the commitments behind spending.
Forecasts vs. results
Amazon’s ~$220B 2026 cash spending outlook and Alphabet’s expectation of a “significant” 2027 increase are current markers, not final results.
Will revenue from AI services grow quickly enough to support infrastructure costs and obligations that can extend well beyond the life of the servers?
The cited filings document spending, financing and contractual commitments; they do not establish how much revenue the new infrastructure will generate, how fully it will be used, or what share of spending is specifically for AI.
Who Carries the Infrastructure Costs
The companies’ spending exceeded operating cash flow at three of the four firms in the source’s comparison. The source reports negative free cash flow of $42.4 billion for Amazon over the prior 12 months, $9.1 billion for Alphabet in the quarter and $2.1 billion for Meta in the quarter. Microsoft remained positive in the comparison.
These measures cover different periods and use different definitions: free cash flow is calculated differently from operating cash flow, and the Amazon figure covers a trailing 12-month period. Borrowing creates repayment obligations; equity sales dilute existing shareholders; leases and partnerships commit future payments or returns to other parties.
The source estimates that the four companies had signed leases not yet started, mostly for data centers, representing about $1.02 trillion in future payments as of September 30. That is a calculation from their quarterly filings, not a single reported company total. These commitments may not appear in headline debt figures, while still requiring substantial payments over time.
The source identifies lenders, investors, pension funds, building owners and electricity customers as parties potentially exposed to the economics of the expansion. Long-term power contracts can shape demand and costs for years. The figures provided do not establish the extent of any eventual effect on electricity prices or other customers.
Products Worth Considering
Anker SOLIX C1000 Gen 2 Portable Power Station
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
EcoFlow DELTA 3 Classic Portable Power Station, 1024Wh, 1800W
As an affiliate, we earn on qualifying purchases.
From Cash-Funded Growth to Mixed Finance
For years, large technology companies often funded data center construction from the cash generated by advertising, software and online retail. The source describes 2026 as a shift to several financing methods operating at once: company cash, debt, leases, separate property-owning entities, new shares and strategic partnerships. It also identifies long-term power contracts as another commitment connected to data center operations.
Meta stated its approach in an October 26 announcement, saying it intended to fund AI investments “through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.” This is the company’s stated financing plan.
The source identifies a potential mismatch between asset lives and payment periods. Servers are typically written off over five to six years, while cited power contracts can run 15 to 20 years, data center leases can extend up to 20 years, and some bonds reach 40 years. If revenue from AI services grows more slowly than costs and obligations, companies or their financing partners would still have to cover the difference. The timing and size of any such shortfall are unknown.
The source also identifies newer GPU cloud providers such as CoreWeave as part of the financing chain. It says CoreWeave senior notes carry rates of 8.5% to 9.75%. The provided material does not detail the terms or performance of each note.
““Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.””
— Meta, in a statement on October 26
Returns and Long-Term Exposure
The filings and figures cited here document spending, financing and contractual commitments; they do not establish how much revenue the new AI infrastructure will generate. The source’s totals for combined spending, long-term debt and future lease payments are its own calculations across company disclosures. They should be read with the underlying definitions and reporting periods in mind.
It is also unclear how much of each company’s property spending is specifically for AI, how much infrastructure will be used at full capacity, and whether demand will support the investments. The source notes that servers have shorter accounting lives than many leases, power contracts and bonds, but it does not quantify the cost of a mismatch or identify who would absorb any losses. The scale and timing of effects on electricity customers are also not specified.
Filings Will Track the Next Spending Cycle
The next evidence will come from company filings and spending guidance as the firms report subsequent quarters. Readers can compare reported property and equipment purchases with operating cash flow, debt issuance and updates to capital spending forecasts. Alphabet’s stated expectation of a significant 2027 increase and Amazon’s approximately $220 billion 2026 cash spending outlook are current markers, not final full-year results.
Further disclosures may clarify the terms and duration of leases, power contracts, loans and strategic partnerships, as well as how much infrastructure is being built for AI. Until those details and revenue results arrive, the key open question is whether earnings from AI services will grow quickly enough to support the long-lived commitments now being made.
Key Questions
How much did the four companies spend in the third quarter?
Amazon, Microsoft, Alphabet and Meta spent a combined about $202 billion on property and equipment from July through September 2026, based on the source author’s sum of company filings. The category includes assets beyond AI infrastructure.
What financing methods are they using?
The source identifies operating cash, debt, share sales, leases, strategic partnerships and separate entities that own buildings. Long-term power contracts add another set of future commitments.
Does the spending prove that AI data centers are profitable?
No. Spending and financing disclosures show the scale and funding of the build-out, but the material provided does not establish the revenue or returns those investments will produce.
Why do leases and power contracts matter?
They can require payments over periods longer than the five-to-six-year server write-off period cited in the source. That timing difference could leave companies and their partners carrying obligations even as equipment ages; the actual financial outcome remains uncertain.
Source: Thorsten Meyer AI
