TL;DR
Get everyday essentials 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 July–September 2026, more than twice their spending a year earlier, according to figures compiled from company filings. The source material says the companies are combining cash, debt, leases, partnerships and equity to finance data center expansion, while the scale and terms of some future commitments leave questions about who would bear costs if AI revenue grows more slowly than expected.
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment from July through September 2026, more than double the roughly $97 billion reported for the same period a year earlier, according to figures compiled from their filings. The scale of the outlay, much of it associated with data centers, servers and chips, is being met through a mix of company cash, borrowing, leases, partnerships and stock sales, shifting some of the financial exposure beyond the companies themselves.
The source compilation says Amazon reported $77.3 billion in property and equipment purchases for the quarter, compared with $35.1 billion a year earlier. The company expects about $220 billion in cash capital spending for all of 2026. Amazon said the increase primarily reflects investments in artificial intelligence, though its spending also includes warehouses and other assets.
Alphabet reported $51.4 billion in quarterly purchases, up from $24 billion a year earlier. Its 2026 capital-spending guidance remained at $195 billion to $205 billion, and the company expects a significant increase in 2027. Microsoft reported $39.4 billion and Meta $33.9 billion. The four-company total and year-over-year comparison are the source author’s sums of company filings; the spending categories include property and equipment and are not limited to AI.
The source says Amazon, Alphabet and Meta raised about $225 billion in long-term debt during the first nine months of 2026. It also reports that Microsoft sold $25 billion in bonds during the week, with some debt extending to 2066; Meta sold $25 billion of common stock and $12.5 billion of preferred shares; and Amazon disclosed a loan arrangement of up to $80 billion secured by servers it plans to purchase. These financing figures and descriptions are attributed to the source material and company disclosures cited there.
AI Data Center Boom: A Look at Cash, Bonds, Leases and Power Bills
Four technology giants are scaling data center infrastructure at record pace. Their funding mix spreads the cost across cash, debt, leases, partners and equity—while leaving the eventual returns an open question.
Up from $35.1B a year earlier
Up from $24.0B a year earlier
Fiscal first quarter of 2027
Property and equipment purchases
Capital is moving fast
The figures cover property and equipment, including data centers, servers and chips. They are not a precise measure of AI spending alone.
Amazon
$77.3BQuarterly purchases; +$42.2B YoY
Alphabet
$51.4BQuarterly purchases; +$27.4B YoY
Microsoft
$39.4BQuarterly purchases; fiscal Q1 2027
Meta
$33.9BQuarterly purchases of property and equipment
The $202B combined total and the year-over-year comparison are calculations from the source compilation of company filings. Spending categories include non-AI assets.
Who carries the infrastructure bill?
Different funding routes place repayment obligations and project exposure with different parties. Contract details determine the exact allocation.
Reported financing moves
*Amazon, Alphabet and Meta reportedly raised about $225B in long-term debt in the first nine months of 2026. Microsoft’s bonds include maturities extending to 2066. Figures are attributed to the supplied source and company disclosures.
Cash & earnings
Funds investment directly from business cash generation.
Bonds & loans
Companies take on borrowing and future repayment.
Leases & partners
Future facility costs may be shared through contracts.
Power agreements
Long-term electricity contracts tie costs to facilities.
Commitments outlast the build
AI demand, facility utilization and revenue may change faster than the contracts that support infrastructure.
Power contracts
The source describes electricity agreements that can run 15 to 20 years, tying buyers and suppliers to long-lived costs.
Leases & buildings
Data center leases may extend up to 20 years. Property owners and financing partners may share exposure under specific contracts.
Long-dated bonds
Some bonds extend decades into the future. The supplied source also cites 8.5%–9.75% senior note rates at GPU cloud provider CoreWeave.
Investors, pension funds and electricity customers could be affected, according to the source’s analysis; it does not quantify potential costs. The cited financing rates do not establish that any borrower will default.
Spending outgrew quarterly cash
The comparison describes a quarter’s cash flows. It is not proof that a company cannot meet its obligations.
The source says three of the four companies spent more on property and equipment in the quarter than their businesses generated in operating cash. Amazon’s trailing 12-month free cash flow was reported at negative $42.4B.
It also cites quarterly free cash flow of negative $9.1B for Alphabet and negative $2.1B for Meta; Microsoft remained positive in the comparison. These figures are not identical measures: some comparisons are source calculations from reported figures, while others are company-reported measures.
Amazon’s reported property and equipment spending also includes warehouses and other non-AI construction. Meta described its approach as a balanced mix of operating cash flow, investment-grade debt, strategic partnerships and equity capital.
Returns remain an open question
The supplied figures show spending and obligations. They do not settle whether future AI revenue will cover the full cost of new infrastructure.
Company guidance
Amazon expects about $220B in cash capital spending for 2026. Alphabet maintained guidance of $195B–$205B and expects a significant increase in 2027.
Operating cash flow
Future filings can show whether spending plans change and how investment compares with cash generated by each business.
Facilities & utilization
Construction schedules, electricity supply and actual data center utilization will shape the economics of projects still being built.
Lease and partner terms
The $1.02T lease estimate lacks company-by-company payment timing and detailed terms in the supplied material. Exposure for partners is not quantified.
“AI bubble” is presented in the source as a debate about who holds risk, not evidence that a collapse is underway. No supplied data establishes whether current spending will produce adequate returns.
Who Carries the Infrastructure Bill
The financing mix matters because it determines who bears the cost if data center projects earn less than expected. Cash spending and debt leave the companies responsible for funding and repayment. Leases create future payment obligations, while separate property-owning companies or strategic partners may place some costs and risks with other investors. The source describes these arrangements as ways to distribute exposure; the specific allocation depends on each contract.
The article estimates the four companies have about $1.02 trillion in signed leases that had not started as of September 30, 2026, based on its sum of their quarterly filings. That figure represents future payments, not an immediate cash bill. It is larger than their combined long-term debt in the source’s calculation, but the comparison does not by itself show how likely the payments are to strain any company’s finances.
Power costs add another long-term commitment. The source says electricity contracts can run 15 to 20 years, while data center leases may extend up to 20 years and some bonds for 40 years. If demand or AI revenue falls short, companies and their financing partners may still face obligations tied to facilities and power. The source identifies investors, pension funds and electricity customers among parties that could be affected, though it does not quantify potential costs to them.
Products Worth Considering
Jackery Explorer 1000 v2 Portable Power Station
Jackery Explorer 300 Portable Power Station, 292Wh
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
As an affiliate, we earn on qualifying purchases.
How Spending Outgrew Quarterly Cash
The source frames 2026 as a shift from a period when major technology companies could largely fund infrastructure from operating profits. In the third quarter, it says three of the four companies spent more on property and equipment than their businesses generated in operating cash. That comparison is a measure of the quarter’s cash flows, not proof that those companies cannot meet their obligations. The source reports Amazon’s trailing 12-month free cash flow at negative $42.4 billion, and quarterly free cash flow at negative $9.1 billion for Alphabet and negative $2.1 billion for Meta. Microsoft remained positive in the comparison provided.
The reported cash-flow figures are not identical measures: Amazon and Microsoft spending-versus-cash gaps in the source are arithmetic based on their reported figures, while Alphabet and Meta free cash flow figures are those companies’ reported measures. Microsoft’s July–September period was its fiscal first quarter of 2027. Amazon’s property and equipment spending also includes non-AI construction, so the total should not be treated as a precise measure of AI investment alone.
Meta described its funding approach on October 26 as a “balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.” That statement illustrates the broader financing approach described in the source. The article also points to newer GPU cloud providers such as CoreWeave, whose senior notes it says carry rates of 8.5% to 9.75%, as potentially more exposed to financing costs. Those rates and risks are presented by the source; they do not establish that any borrower will default.
Open Questions on Future Returns
The filings and figures described in the source show spending and commitments, but they do not settle whether AI services will generate enough revenue to cover the full cost of new infrastructure. Future demand, pricing and utilization remain uncertain, as do the returns on individual data centers and the extent to which planned facilities will be completed on schedule.
The roughly $1.02 trillion lease estimate is the source author’s sum of commitments described as not yet started as of September 30. The material does not provide a company-by-company breakdown here, the timing of each payment, or the detailed terms that could affect actual costs. It also does not quantify the exposure held by building owners, lenders, power providers or other partners. The source’s discussion of who might ultimately bear losses is analysis, not a confirmed forecast.
The term “AI bubble” is presented in the source as a debate about who holds risk, rather than as evidence that a collapse is underway. No data in the supplied material establishes that current spending will or will not produce adequate returns. The financing arrangements, project outcomes and future AI revenues will determine how that question develops.
Spending and Commitments to Track
The next indicators are company disclosures on capital spending, operating cash flow, borrowing and lease obligations. Amazon has forecast about $220 billion in cash capital spending for 2026; Alphabet has held its 2026 guidance at $195 billion to $205 billion and expects spending to rise significantly in 2027. Future filings can show whether those plans change and how spending compares with cash generated by each business.
Investors and customers will also be watching for details on facility construction, electricity supply agreements and the financing of data centers through outside partners. The source material does not identify a single next announcement or milestone, so the timing of further disclosures is unclear. The next company results and filings should provide updated figures, but they cannot by themselves establish the long-term returns on infrastructure still being built.
Key Questions
How much did the four companies spend in the third quarter?
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment from July through September 2026, according to the source author’s sum of their filings. The figure includes spending beyond AI, including Amazon warehouse construction.
Which financing sources are being used?
The source describes a mix of operating cash, bonds and other debt, leases, partnerships and stock sales. Companies’ individual arrangements vary, and the supplied material does not give full terms for every financing deal.
What does the $1.02 trillion lease estimate represent?
It is the source author’s estimate of future payments under signed leases that had not started as of September 30, 2026, based on four company filings. It is a commitment over time, not a bill due all at once.
Does the spending prove there is an AI bubble?
No. The figures show a sharp rise in infrastructure spending and financing commitments, but they do not determine whether AI revenues will justify those costs. The source presents the bubble debate as a question about future returns and who bears the financial risk.
Source: Thorsten Meyer AI
