AI Data Center Costs: Who Pays, And How?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: AI Data Center Costs: Who Pays, And How? on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment in the third quarter of 2026, more than twice their combined spending a year earlier, according to figures compiled from company filings. The financing mix increasingly includes debt, leases, partnerships and equity, but the scale of future obligations and the returns from AI infrastructure remain uncertain.

Amazon, Microsoft, Alphabet and Meta spent about $202 billion on property and equipment from July through September 2026, more than twice their spending in the same period a year earlier, according to a calculation based on their cash-flow statements. The scale of the AI infrastructure build-out is pushing the companies to use debt, leases, partnerships and equity alongside operating cash to fund data centers, servers and chips.

The companies’ filings show that Amazon spent $77.3 billion on property and equipment in the quarter, Alphabet $51.4 billion, Microsoft $39.4 billion in cash and Meta $33.9 billion. The combined total of about $202 billion compares with roughly $97 billion a year earlier, based on the source’s calculation. Microsoft’s figure rises to $45.2 billion when new finance leases are included, according to the company’s reporting.

The spending is large relative to cash generated in the same quarter. Amazon reported $42.4 billion in cash from operations, Alphabet $42.3 billion, Meta $32.8 billion and Microsoft $58 billion. A comparison of those figures with cash capital spending suggests only Microsoft had operating cash left after that spending among the four; the source calculates the difference at about $18.6 billion. This is a quarterly comparison, not a measure of each company’s overall financial health.

Funding methods vary. The source reports that Meta said on October 26 it plans to finance AI investment through a “balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital.” It also reports that Meta sold $25 billion of stock and $12.5 billion of preferred shares, Microsoft issued $25 billion in bonds, and Amazon disclosed a loan facility of up to $80 billion secured by servers it plans to purchase. The figures and descriptions are attributed to the source’s account of company disclosures.

At a glance
reportWhen: Reported October 30, 2026; figures cove…
The developmentA sharp rise in third-quarter capital spending by four major technology companies is being financed through a growing mix of cash, debt, leases, partnerships and equity.
Thorsten Meyer AI
Q3 2026 · filings

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.

$202Bspent on property and equipment by the four, July–September 2026 (my sum)
2.1×the same quarter a year earlier (about $97B)
3 of 4spent more than their business brought in
$225Bnew long-term debt at Amazon, Alphabet and Meta, Jan–Sep 2026
Spent on property and equipmentCash from operations

AmazonQ3 2025: $35.1B spent

Spent$77.3B
Cash in$42.4B

Spent $34.9B more than came in

AlphabetQ3 2025: $24.0B spent

Spent$51.4B
Cash in$42.3B

Free cash flow −$9.1B

MicrosoftQ3 2025: $19.4B spent

Spent$39.4B
Cash in$58.0B

$18.6B left over

MetaQ3 2025: $18.8B spent

Spent$33.9B
Cash in$32.8B

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 Infrastructure Bill

The financing choices affect more than the technology companies building the facilities. Debt shifts repayment obligations to future cash flows; leases and power contracts commit companies to payments over time; partnerships and property-owning entities can place some financing outside a technology company’s direct balance sheet. 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 its sum of their quarterly filings. That is a reported estimate of future commitments, not cash already spent.

The key financial question is whether revenue from AI services and products will grow enough to support the infrastructure costs. If returns arrive later or prove smaller than anticipated, investors and lenders may bear financial losses. The source also points to potential exposure for building owners, infrastructure firms and electricity customers, though the exact distribution of costs depends on contract terms and future demand. Who ultimately bears the risk is therefore as important as the headline spending total.

Products Worth Considering

Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera (Midnight)

Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera (Midnight)

Anker SOLIX C1000 Gen 2 Portable Power Station with 200W Solar Panel

Anker SOLIX C1000 Gen 2 Portable Power Station with 200W Solar Panel

As an affiliate, we earn on qualifying purchases.

From Cash-Funded Growth to Mixed Financing

The source describes a change from earlier years, when large technology companies could fund data centers from profits generated by advertising, software and online commerce. In 2026, the reported quarterly investment pace has outgrown the cash generated by some of those businesses over the same period. Companies are responding with several funding routes rather than relying on one source.

The commitments also have different time horizons. The source says servers are typically written off over five to six years, while data-center leases can extend up to 20 years and some power contracts run 15 to 20 years. Microsoft’s reported bond sale included debt extending to 2066. These differences can leave companies paying for facilities or electricity after particular generations of equipment have been replaced. The source cites high borrowing costs for newer GPU cloud provider CoreWeave, including senior notes carrying 8.5% to 9.75%, as an example of financing conditions that can differ beyond the largest technology companies.

Returns and Long-Term Exposure

The figures establish the scale of spending and disclosed commitments, but they do not show how much revenue the investments will generate. AI demand, infrastructure utilization and future pricing remain uncertain, as do the costs of operating data centers, including electricity. The source’s $1.02 trillion lease estimate is its sum of figures in four company filings; the precise composition and timing of each obligation may differ.

It is also unclear how much risk is held by the technology companies themselves versus lenders, investors, landlords, infrastructure providers and utilities. Some financing arrangements may involve partners or separate entities, and their terms are not fully detailed in the material provided. The discussion of a possible AI investment bubble is therefore not a confirmed finding about future losses: it concerns whether expected returns will match the scale and duration of commitments.

Company Spending Plans Through 2027

The companies’ published spending guidance will be a near-term measure of whether the build-out continues at its current pace. The source says Amazon expects about $220 billion in cash capital spending for 2026; Alphabet’s guidance is $195 billion to $205 billion and it expects a significant increase in 2027; Microsoft expects about $170 billion for calendar 2026; and Meta forecasts $135 billion to $145 billion for 2026, including finance-lease payments. These company estimates use different descriptions and should not be treated as perfectly comparable measures.

Future earnings reports and filings should provide updated spending totals, financing details and information about lease obligations. Investors will also look for evidence that AI-related services are producing revenue relative to infrastructure costs. Until those disclosures show how utilization and returns develop, the cost-sharing picture remains in motion.

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 calculation from their cash-flow statements. The figure is about $97 billion for the same quarter a year earlier by the same calculation.

Who pays for the AI data center build-out?

The companies use a mix of operating cash, borrowing, leases, partnerships and equity. The financing arrangements can also involve bondholders, lenders, landlords and other investors, but the exact allocation of risk depends on individual contracts.

What does the $1.02 trillion lease figure mean?

It is the source’s calculation of future payments for leases that had not yet started as of September 30, 2026, based on four companies’ filings. It represents future commitments, not spending already paid.

Does the spending prove there is an AI bubble?

No. The spending and financing figures do not establish that the investments will lose money. The unresolved issue is whether future AI revenue and demand will be sufficient to support the infrastructure and long-term payment commitments.

Source: ThorstenMeyerAI.com

You May Also Like

Running Gemma 4 26B At 5 Tokens/sec On A 13-Year-old Xeon With No GPU

A 13-year-old Xeon CPU successfully runs Gemma 4 26B at 5 tokens/sec without a GPU, demonstrating unexpected hardware efficiency.

Apertus. The architectural template.

Apertus, developed by Swiss institutions, introduces a new open, compliant, multilingual AI model supporting 1,811 languages, setting a European sovereign-AI benchmark.

Different Game, or Already Lost? Reading Mistral’s Sovereignty Bet

Analysis of Mistral’s strategic shift toward full-stack AI and on-prem solutions amid industry skepticism and emerging competition.

AI And No-Code: The Future Of Chrome Extension Development

New AI-powered no-code tools are emerging to help non-developers build Chrome extensions easily, transforming browser automation and productivity.