Skip to content

CIRCO Research · AI capex

$1.5 Trillion Has Been Spent On AI Build-Out

Catalyst Intelligence Research ·

Companies listed on US exchanges have put $1.5 trillion of capital spending behind AI since the start of 2022, by their own account: $1,515.3 billion through their latest reported quarters, as of October 29, 2026.

The pace keeps rising. In 2026 alone the figure is $632.2 billion so far, already 39% more than all of 2025, with a quarter still to report. By year-end, US-listed companies would have spent $1,789.7 billion to $1,823.7 billion on AI capital expenditure if they meet their full-year guidance.

The total is built strictly from what companies disclose to investors. It adds up each company’s reported capex in the periods when its filings, earnings releases or calls explicitly tie that spending to AI.

Why the count starts in 2022

2022 is when the AI boom started, so the count starts there. OpenAI released ChatGPT on November 30, 2022, and the race to build AI capacity followed.

Year by year

2026 has already passed all of 2025

Capex companies tie to AI, $ billions by year. $1,515.3B since 2022.
2026 to date is 39% above 2025; full-year guidance is 99–106% above 2025.

2022: $32.8B$32.8B20222023: $133.4B$133.4B20232024: $255.9B$255.9B20242025: $455.9B$455.9B20252026 to date: $632.2B$632.2B2026 to dateFull-year guidance: $906.6–940.6B; $274.4–308.4B above 2026 to date$906.6–940.6B guided

Dotted: the rest of the companies’ newest full-year capex guides, $906.6–940.6B in total, each for its own fiscal year.
Source: Catalyst Intelligence Research, https://cir.co/reports

The total has grown 78–92% a year since 2023, partly because more companies began tying their capex to AI.

  • 2019 - 2022: Non-capex spending, investments in AI.
  • 2022: $32.8 billion, companies begin attributing AI to their capital expenditure.
  • 2023: $133.4 billion, as Amazon, Alphabet and Microsoft joined. Microsoft counts from its September 2023 quarter, when it first tied capex to scaling its AI infrastructure.
  • 2024: $255.9 billion, up 92%.
  • 2025: $455.9 billion, up 78%. Oracle enters the hyperscalers.
  • 2026 so far: $632.2 billion, 39% above all of 2025 with a quarter to go. Full-year guidance is $906.6–940.6 billion, leaving $274.4–308.4 billion still to spend.

The four largest spenders, Amazon, Microsoft, Alphabet and Meta, went from $244.5 billion in 2024 to $408.4 billion in 2025, up 67%. In 2026 they have spent $508.1 billion so far, already 24% more than all of 2025, and their full-year guidance of $740–760 billion would be 81–86% above 2025.

With Oracle, the five hyperscalers went from $244.5 billion in 2024 to $428.9 billion in 2025, up 75%, with Oracle counted from the second half of 2025. They have spent $571.7 billion in 2026 so far, 33% more than all of 2025, and their full-year guidance of $830–855 billion would be 94–99% above 2025. Microsoft’s and Oracle’s guides cover their fiscal years ending in June and May 2027.

Who spent it

The five hyperscalers carry 93% of the total: Amazon, Microsoft, Alphabet, Meta and Oracle.

Hyperscalers carry 93% of AI capex since 2022

Share of the $1,515.3B companies tie to AI, 2022 through each company’s latest reported quarter.

Amazon: 28.2%, $426.6B since 2022Microsoft: 22.1%, $334.4B since 2022Alphabet: 20.3%, $307.6B since 2022Meta: 16.9%, $255.9B since 2022Oracle: 5.6%, $84.1B since 2022All others: 7.0%, $106.8B since 2022$1,515.3Bsince 2022
CompanyShareSince 2022
Hyperscalers93.0%$1,408.5B

Click a slice or a company to see it; click it again to reset.

Hyperscalers: Amazon, Microsoft, Alphabet, Meta and Oracle.
Source: Catalyst Intelligence Research, https://cir.co/reports

From spending to returns

CIRCO will investigate the broader AI investment ecosystem, including hyperscalers, neoclouds, frontier labs, suppliers and capital providers. This analysis report focuses on Amazon, Microsoft, Alphabet, Meta and Oracle, while the upcoming 2027 AI Infrastructure Outlook report will examine where the capital is going, how it is possibly financed, who the suppliers are and which customers are expected to pay for the resulting capacity.

The spending total is just one factor for that assessment. The next questions concern execution and returns: whether facilities receive power and become operational on schedule, whether customers use the capacity, and whether the revenue generated can justify the investment.

With the cumulative figure now past $1.5 trillion, the focus turns to who will survive the build-out.

How the figure is built

Expand

Every dollar is a number a company published. Nothing is estimated, modelled or annualised.

  • The company’s own figure. Capex as stated in a 10-K, 10-Q, earnings release or call. For a quarter a company did not state, the capex line in its own 10-Q or 10-K is used, but only where its stated quarters match its filings.
  • Only periods tied to AI. A period counts when the company’s own statement names AI, AI capacity, GPUs or compute as a driver of that fiscal year’s capex. Analysts’ questions and remarks about customers’ spending do not count.
  • Gross, not net. A figure the company labels net of prepayments or government incentives is replaced by its gross figure. Amazon’s figures are its cash capital expenditures, which Amazon itself reports net of proceeds from sales and incentives.
  • Spend, not plans. Guidance and forecasts are never added to the total. The companies’ newest full-year guides sum to $906.6–940.6 billion, each for its own fiscal year. The dotted part of the 2026 bar is guidance, shown apart from the total.
  • Each period once. A full-year figure covers its year; otherwise the year’s quarters are summed. Tesla’s stated AI capex to date ($5.0 billion at the end of 2024) replaces the periods before it.
  • For most companies this is capital spending that management says AI is driving, not funds set aside for AI alone. The figure is a floor: a company that has never tied its capex to AI is not counted.

The result is a floor. It covers only companies listed on US exchanges, only those that put their own number on AI-related spending, and only the periods they tied to AI.

Sources

All figures come from CIRCO’s Cluster Platform (public summary: AI CapEx Tracker), as of October 29, 2026. Each figure in the tracker carries the sentence or filing line it came from.

Filings and earnings releases (SEC EDGAR)

Earnings calls

  • Meta: 14 calls, April 2022 to October 2025
  • Microsoft: 13 calls, October 2023 to October 2026
  • Micron: 11 calls, December 2023 to September 2026
  • Alphabet: 6 calls, October 2023 to February 2025
  • CoreWeave: 4 calls, November 2025 to August 2026
  • Oracle: 2 calls, September and December 2025
  • Amazon: Q1 2024 call, April 30, 2024 (full-year 2023 capital investments)
  • Tesla: Q4 2024 call, January 29, 2025 (AI capex to date)
  • Nebius: Q2 2026 call, August 12, 2026

Information and analysis, not investment advice.

Receive the free Catalyst Intelligence Briefing.

The key findings from the CIRCO data: the risks spreading across companies, which issuers are most exposed, and what the latest filings and earnings calls reveal.

Free. Unsubscribe anytime.

$1.5 Trillion Has Been Spent On AI Build-Out — CIRCO