TL;DR
Get smart everyday buys 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 the year-earlier amount, according to figures compiled from company filings. The companies are funding the build-out with cash, borrowing, leases, partnerships and share sales; how much risk ultimately reaches investors, building owners and electricity customers depends on future AI revenue and contract terms.
Amazon has lined up a loan of up to $80 billion, secured by servers it will buy with the proceeds, as major technology companies scale up data center construction and turn increasingly to outside financing. Filings and company statements cited by Thorsten Meyer AI show Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the third quarter of 2026, raising questions about how much of the build-out’s cost and risk falls on investors, lenders, landlords and electricity customers.
The four companies’ combined spending from July through September was about 2.1 times the roughly $97 billion they spent in the same period a year earlier, according to a total calculated from their cash flow statements. The category includes buildings, servers and chips; it is not limited to AI. Amazon, for example, also spends on warehouses. Amazon said the increase in its spending primarily reflected investment in artificial intelligence.
Amazon reported $77.3 billion in property and equipment purchases for the quarter, up from $35.1 billion a year earlier. Alphabet reported $51.4 billion, compared with $24 billion. Microsoft reported $39.4 billion for its fiscal first quarter, which ran from July through September, and Meta reported $33.9 billion. The source material says Amazon expects about $220 billion in cash capital spending for all of 2026; Alphabet held its 2026 spending outlook at $195 billion to $205 billion and expects a significant increase in 2027.
The financing is coming from several channels. Meta said it plans to use operating cash flow, investment-grade debt, strategic partnerships and equity capital. The source material reports that Meta sold $25 billion of its own stock and $12.5 billion of preferred shares during the week, Microsoft borrowed $25 billion, and Amazon disclosed its server-backed loan facility. Amazon, Alphabet and Meta together raised about $225 billion in long-term debt during the first nine months of 2026, according to a calculation from their cash flow statements.
Capital • Contracts • Power
Are Investors Or Utility Customers Paying For AI Data Centers?
A record build-out is drawing on cash, borrowing, leases, partnerships and share sales. Who ultimately carries the risk depends on future AI income and the terms behind each commitment.
Amazon, Microsoft, Alphabet and Meta spent about 2.1× their year-earlier total on property and equipment.
Property and equipment
Compared with July–September 2025
Loan secured by servers it will buy
Amazon, Alphabet and Meta, first 9 months
The build-out, by company
Reported property and equipment purchases for July–September 2026. These totals include non-AI projects as well as data centers.
Up from $35.1B a year earlier
AI cited as primary driverUp from $24B a year earlier
2026 outlook: $195–205BFiscal first quarter, July–September
Operating cash remained positiveQuarterly property and equipment
Multiple funding channels*Combined lease commitments whose payments had not yet begun as of September 30; mostly tied to data centers. This is a future obligation, not a single bill due now.
Many funding routes, different obligations
Companies are drawing on operating cash and capital markets while using leases and partnerships to finance infrastructure.
Debt comes due
Microsoft borrowed $25B; reported bonds extend as far as 2066. Amazon’s server-backed facility could provide up to $80B.
More capital sources
Meta cited operating cash flow, investment-grade debt, partnerships and equity. The source reports $25B in stock and $12.5B in preferred shares sold during the week.
Payments stretch out
Some data center leases run up to 20 years; power contracts may last 15–20 years. These terms spread financing and operating commitments across parties.
3 of 4
companies spent more on property and equipment in the quarter than their businesses generated in operating cash, according to the source. Reported figures include Amazon trailing 12-month free cash flow of −$42.4B, Alphabet quarterly free cash flow of −$9.1B and Meta’s of −$2.1B. Microsoft remained clearly positive.
Commitments can outlast equipment
Long contract periods can leave companies managing obligations after particular servers have aged or been replaced.
Cash needs are still rising
Alphabet also expects a significant increase in 2027. Capital spending includes property and equipment beyond AI.
What companies say
Statements describe the financing mix and spending rationale; they do not settle who bears every downstream cost.
“Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.”
Meta • Funding approach
Capital spending “primarily reflects investments in artificial intelligence.”
Amazon • Spending increase
The company expects a “significant” increase in capital spending in 2027.
Alphabet • Outlook
The reported cash flow measures are snapshots. Operating cash flow and free cash flow are different: free cash flow generally reflects operating cash after capital spending. Neither measure forecasts cash generation over a data center’s full life.
Where could the exposure land?
Financing creates obligations across companies and counterparties. The available figures do not quantify the eventual share borne by each group.
Borrowers
Repay debt and meet lease, equipment and power commitments.
Investors & lenders
Hold debt, equity or other interests tied to the build-out.
Landlords & partners
May share financing, construction and operating risks.
Utilities & customers
Could be affected by the cost of supplying sustained power demand.
The figures do not show how much data center construction or power cost has entered utility rates or customer bills.
Lease totals are not due all at once, and contract terms determine how obligations are shared among counterparties.
The open question: can AI income carry the commitments?
Future revenue growth will shape whether today’s financing remains manageable or begins to strain borrowers and partners.
Revenue may support the build
Strong AI income could help cover construction, equipment, financing and power costs over time.
Pressure can spread
Borrowers, landlords, infrastructure partners and investors could face stress if revenue falls short of commitments.
GPU cloud financing
The source flags CoreWeave senior notes at 8.5%–9.75% as a potential weak point, without forecasting losses.
What to watch next
New filings and public decisions can clarify both the scale of commitments and who bears their costs.
Cash, spending & AI revenue
Compare operating cash flow and updated capital spending outlooks with disclosed AI-related revenue, remembering total equipment spending includes non-AI projects.
Quarterly filingsDebt, leases & partnerships
Track borrowing, payment schedules, lease commitments and which parties carry performance or operating risk.
Terms matterUtility rates & power deals
Utility filings, rate decisions and power agreements may show whether data center demand changes costs for other electricity users.
Customer impactThe spending surge matters because the cost of building and operating data centers can extend well beyond a company’s immediate cash outlay. Debt puts repayment obligations on the borrower. Leases commit a company to future payments, while partnerships and separate building-owning companies can distribute financing and operating risks among several parties. Long-term power contracts also bind buyers and sellers to future costs and demand.
According to Meyer’s sum of the four companies’ filings, they had about $1.02 trillion in leases for which payments had not yet begun as of September 30, mostly tied to data centers. That figure is more than twice their combined long-term debt, but it represents future contractual payments, not a bill due all at once. The scale is relevant to investors and lenders assessing how the companies might meet obligations if AI-related income grows more slowly than expected.
The potential reach goes beyond technology company shareholders. Investors and pension funds may hold debt or other interests tied to the build-out, while utilities and customers may be affected by the costs of supplying large, sustained power demand. The source material identifies these groups as possible holders of exposure; it does not quantify how much cost has been passed through to electricity bills or establish that customers are already paying a specific share.
Products Worth Considering
Samsung 34-Inch ViewFinity S50GC Ultra-WQHD Monitor
Samsung Essential S3 27-Inch Curved FHD Monitor
Dell P2725H 27-Inch Full HD Monitor
As an affiliate, we earn on qualifying purchases.
From Cash Funding to Multiple Sources
For years, large technology companies could use cash generated by advertising, software and online shopping to fund new infrastructure. The figures presented for 2026 show a sharper funding challenge: three of the four companies spent more on property and equipment in the quarter than their businesses generated in operating cash, based on the source’s comparison. It reports trailing twelve-month free cash flow of negative $42.4 billion for Amazon, and quarterly free cash flow of negative $9.1 billion for Alphabet and negative $2.1 billion for Meta. Microsoft remained clearly positive in the comparison.
These measures have limits. Capital spending includes property and equipment beyond AI, and company cash flows do not identify every dollar’s purpose. Operating cash flow and free cash flow also measure different things: the latter generally reflects operating cash after capital spending. The reported quarter figures are snapshots, not forecasts of how much cash each company will generate over the life of a data center.
Financing terms vary across the sector. Microsoft’s $25 billion borrowing reportedly included bonds extending to 2066, while the source says some data center leases run for up to 20 years and power contracts for 15 to 20 years. Servers, by comparison, are typically written off over five to six years, according to the source. Those different time frames can leave companies managing long commitments after particular equipment has aged or been replaced.
““Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.””
— Meta
Where the Financial Exposure Lands
The filings and figures presented here do not show exactly how much of the spending is specifically for AI, how the $1.02 trillion in future lease payments is distributed over time, or which parties ultimately bear each contract’s risk. They also do not establish how much of data center construction or power costs will be reflected in utility rates or customer bills.
Future AI revenue is another open question. If demand and income grow enough to cover construction, equipment, financing and power commitments, the companies may be able to carry the costs. If revenue grows more slowly, borrowers, landlords, infrastructure partners and investors could face pressure. That is a risk scenario, not a confirmed outcome. The source material also flags newer GPU cloud companies such as CoreWeave, whose senior notes it says carry rates of 8.5% to 9.75%, as a potential weak point, but does not detail their exposure or forecast losses.
Spending and Contracts to Watch
The next evidence will come from company results, updated capital spending outlooks and disclosures about borrowing, leases and partnerships. Investors can compare those commitments with operating cash flow and reported AI-related revenue, while keeping in mind that the companies’ total property and equipment figures include non-AI projects.
Utility filings, rate decisions and power agreements may clarify whether data center demand changes costs for other electricity users. Until those details are available, the current figures show the size and range of financing commitments, but not a final accounting of who pays for them. The central question is whether future AI income can support obligations that, in some cases, extend well beyond the useful life of the servers being purchased.
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 a sum of their company filings cited by Thorsten Meyer AI. The category includes more than AI infrastructure.
Are utility customers already paying for these data centers?
The source material identifies electricity customers as a group that could bear some costs, but it does not show how much has reached customer bills or establish that customers are paying a particular share now. Utility rates and power contract details would help answer that question.
What financing methods are the companies using?
The source describes a mix of company cash, debt, leases, partnerships and share sales, along with long-term power contracts. Amazon’s disclosed loan facility of up to $80 billion is secured by servers it plans to buy.
Why do long-term commitments matter?
Some data center leases and power contracts can last much longer than the five-to-six-year period over which servers are written off, according to the source. If revenue grows more slowly than expected, companies may still owe payments after particular equipment has aged or been replaced.
Source: Thorsten Meyer AI
