TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Amazon, Microsoft, Alphabet and Meta reported about $202 billion in property and equipment spending from July through September 2026, more than twice the amount a year earlier. Company filings and disclosures cited in the source show that the build-out is funded through a mix of cash, debt, leases, partnerships and equity, with long-term obligations that can outlast the servers they support.
Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the three months through September 2026, according to a sum of their company filings cited by Thorsten Meyer AI. The total was about 2.1 times the roughly $97 billion reported for the same period a year earlier, as the companies expanded data center, server and chip capacity. Their financing announcements and filings show that the bill is being met through several channels, including borrowing and leases, spreading exposure beyond the companies’ operating cash.
The financing activity has accelerated alongside the spending. The source reports that Amazon, Alphabet and Meta raised about $225 billion in long-term debt during the first nine months of 2026. Microsoft announced a $25 billion bond sale during the week covered by the source, with some bonds due in 2066. Meta said on October 26 that it intended to fund AI investment through operating cash flow, investment-grade debt, strategic partnerships and equity capital.
The source also cites an Amazon disclosure on Friday that the company had arranged a loan of up to $80 billion, secured by servers it will buy with the proceeds. It says Meta sold $25 billion of its own stock and $12.5 billion of preferred shares during the week. These arrangements differ in who supplies capital and what backs repayment, but together they show that the companies are drawing on more than cash generated by their businesses.
Cash flow figures in the source underline that pressure, though they do not mean every dollar of spending went to AI. Three of the four companies spent more on property and equipment in the quarter than their businesses brought in as operating cash, according to the article’s calculations. Amazon reported $77.3 billion in property and equipment purchases, Alphabet $51.4 billion, Microsoft $39.4 billion and Meta $33.9 billion. Amazon’s spending also includes warehouses; its company statement said the increase primarily reflected AI investments.
Capital, Capacity & Commitments · 2026
Who Bears the Cost of Building AI Data Centers?
Big Tech’s data center build-out is funded through cash, debt, leases, partnerships and equity. The obligations can last long after the servers they support are replaced.
Q3 2025 baseline
$97B Combined property and equipment spendLong-term debt raised
$225B Amazon, Alphabet and Meta · first 9 monthsUnstarted lease payments
$1.02T Four-company filing total as of September 30Microsoft bond sale
$25B Some bonds due in 2066Four companies, one quarter, $202 billion
These filing figures cover property and equipment. They include spending beyond AI infrastructure.
$77.3B
Purchases in the quarter; also includes warehouses. Company cited AI investment as a primary driver of the increase.
$51.4B
Up from $24.0B in the same quarter a year earlier.
$39.4B
July–September was its fiscal first quarter of 2027.
$33.9B
Property and equipment purchases reported for the quarter.
The source’s combined total is calculated from company filings. Three of the four spent more on property and equipment than their businesses generated in operating cash during the quarter.
Capital comes from more than operating cash
Each channel shifts repayment duties, ownership or risk in a different way.
Operating cash
Cash generated by advertising, software, shopping and other business lines can fund construction directly.
Bonds & loans
Debt raises capital now and requires repayment. Amazon arranged a loan of up to $80B secured by servers bought with the proceeds.
Leases & power
Facility leases and electricity contracts commit companies to future payments, often over many years.
Partnerships
Partners can divide investment and returns, while adding another party to the project’s economics.
Shares & preferred stock
Meta said it sold $25B of common stock and $12.5B of preferred shares during the reported week.
GPU cloud providers
CoreWeave senior note rates cited in the source ranged from 8.5% to 9.75%. Rates show borrowing cost, not default certainty.
“Through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.”
Meta statement cited by Thorsten Meyer AI · October 26, 2026
Some commitments outlast the hardware
Equipment may be replaced while leases, power agreements and bonds remain in force.
The source describes a gap between the useful life of computing equipment and the length of the commitments used to build supporting infrastructure. If demand or revenue grows more slowly than expected, companies could still owe money for capacity that is less productive or valuable than planned.
Reported estimate of future payments under leases not yet started as of September 30, summed across four companies’ filings. Most were described as data center leases. This is a long-term contractual commitment, not an amount due immediately.
Who could bear a shortfall?
The source identifies possible exposure channels, but does not estimate what each group could ultimately bear.
Companies
Spend cash, borrow, lease facilities or share project investment.
Capital providers
Lenders, bondholders and equity investors supply funds and assume different risks.
Owners & utilities
Building owners and power providers support the facilities and may face long commitments.
Customers & markets
Electricity customers may feel effects of major new demand; AI revenue remains to be proven at scale.
Revenue must catch up with commitments
Spending totals show scale. They do not, by themselves, answer whether the build-out is a bubble.
Follow spending guidance
Amazon expected about $220B in 2026 cash capital spending. Alphabet maintained guidance of $195B–$205B and expected a significant increase in 2027. These are company outlooks, not final results.
Track funding and utilization
Future filings can show how much spending comes from operating cash, how debt and lease commitments change, and whether new capacity is being used to generate AI-related revenue.
The source does not provide project-level revenue, utilization, full financing terms, or a breakdown of the $202B quarterly total devoted specifically to AI. Companies could absorb losses, refinance, change plans or earn enough to meet obligations.
Long Contracts Spread the Exposure
The central financial issue is the gap between the useful life of computing equipment and the length of the commitments used to build the facilities around it. The source says servers are typically written off over five to six years, while power contracts can run 15 to 20 years, data center leases up to 20 years, and some bonds 40 years. If demand or revenue from AI grows more slowly than expected, companies may still owe money for infrastructure that is less productive or less valuable than planned.
The source’s sum of four companies’ filings puts future payments under leases that had not yet started at about $1.02 trillion as of September 30, mostly for data centers. That is a reported estimate of contractual commitments, not a bill due immediately. The distinction matters: the headline amount signals obligations over time, while the actual cost and risk depend on lease terms, project use, future revenue and the companies’ ability to meet payments.
Large technology firms can absorb some setbacks through their existing businesses and access to capital. But financing structures can place exposure with other parties too, including lenders, bondholders, investors in companies that own data center buildings, and utilities or electricity customers affected by major new demand. The source identifies newer GPU cloud providers such as CoreWeave as another pressure point, citing senior note rates of 8.5% to 9.75%. Those rates indicate borrowing costs; they do not by themselves establish that a company will default or that its projects will fail.
Products Worth Considering
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
EcoFlow DELTA 3 Classic Portable Power Station, 1024Wh, 1800W
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
As an affiliate, we earn on qualifying purchases.
From Cash-Funded Growth to Borrowing
The source describes a change from a period when large technology companies could fund data centers from profits generated by advertising, software and online shopping. In 2026, spending on property and equipment climbed sharply: the four companies’ combined third-quarter total was about $202 billion, compared with about $97 billion in the same quarter of 2025. The figures come from company cash flow statements as summarized by the source; the category includes equipment and property beyond AI infrastructure.
Company disclosures provide examples of the scale. Amazon reported $77.3 billion in third-quarter property and equipment purchases, compared with $35.1 billion a year earlier, and expected about $220 billion in cash capital spending for 2026. Alphabet reported $51.4 billion, up from $24.0 billion, and maintained its 2026 capital spending guidance of $195 billion to $205 billion. It expected a significant increase in 2027. Microsoft’s July–September period was its fiscal first quarter of 2027.
The financing methods are not interchangeable. Operating cash comes from the business; bonds and loans require repayment; leases commit companies to future payments for facilities; share sales raise equity capital; and partnerships can divide investment and returns. The source groups these as five ways to finance data centers, with long-term power contracts adding another major commitment. Its figures are based on company filings and announcements, but some totals are the source author’s calculations rather than a single company-reported number.
““through a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.””
— Meta, in a statement cited by Thorsten Meyer AI on October 26, 2026
Revenue Must Catch Up With Commitments
The filings and announcements establish spending and financing activity, but they do not establish how quickly AI services will generate enough revenue to cover the cost of the new capacity. The source frames a slower-than-expected revenue ramp as a risk scenario, not a confirmed outcome. It also does not provide project-by-project revenue, utilization rates or the full terms of the financing arrangements.
The $1.02 trillion lease figure is the source author’s sum of four companies’ filings for commitments that had not started by September 30. The source describes most of these as data center leases, but does not break the total down by company, project, payment schedule or cancellation terms. Nor does it quantify how much of the $202 billion quarterly spending was specifically for AI, rather than other property and equipment.
Who ultimately bears any shortfall is also unsettled. Companies could absorb losses, refinance, change construction plans or earn enough from AI and other services to meet their obligations. Investors, lenders, building owners and electricity customers may face different forms of exposure, but the source does not estimate the amount each group could bear. The “bubble” question therefore cannot be answered from spending totals alone.
Watch Filings and Financing Terms
The next evidence will come from company results and filings showing whether spending continues at the current pace, how much is funded from operating cash, and how debt and lease commitments change. Amazon’s stated expectation of about $220 billion in cash capital spending for 2026 and Alphabet’s 2026 guidance of $195 billion to $205 billion provide company-specific figures to track. Alphabet’s expectation of a significant spending increase in 2027 is a stated outlook, not a final spending result.
Investors and customers will also need clearer evidence on how new capacity is being used and whether AI-related revenue grows alongside it. Future disclosures may clarify the terms and counterparties behind loans, leases, partnerships and long-term power contracts. Until then, the confirmed development is a sharp rise in reported spending and a broader financing mix; the eventual returns, and how costs would be divided if growth falls short, remain open questions.
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 2026, according to the source’s sum of their company filings. The category includes more than AI data centers and equipment.
Does the $1.02 trillion lease figure mean that amount is due now?
No. The source describes it as about $1.02 trillion in future lease payments under agreements that had not started as of September 30, 2026. The source does not give a complete payment schedule or project-by-project breakdown.
Who could bear the cost if AI revenue grows more slowly?
The source identifies technology companies, lenders, bondholders, investors in data center building owners and electricity customers as possible points of exposure. It does not quantify what each group could bear, and a slower revenue ramp is a risk scenario rather than an established outcome.
Are all of the reported property and equipment purchases for AI?
No. The filings report property and equipment purchases, not an AI-only total. The source notes that Amazon also builds warehouses, while Amazon said the increase in its spending primarily reflected AI investments.
Source: Thorsten Meyer AI
