AI Capex Flow MapHyperscaler → Supplier Attribution
LIVEFiled data through 2026Q1·Sourced from SEC EDGAR XBRL

AI capex flows beyond the headline names.

Microsoft, Meta, Alphabet, Amazon, and Oracle are committing more than $400 billion a year to data center build-out. This dashboard traces those dollars from filed cash flow statements through to the suppliers actually capturing the spend: power, networking, memory, foundry equipment, and cooling, not just the GPU vendor everyone owns.

Big 5 Capex · TTM
$482.1B+78.0%

Trailing 12 months across MSFT, META, GOOGL, AMZN, ORCL

Forward Guidance · Midpoint
$522.5B+8.4%

Range $485B – $560B per stated guidance

AI-Attributable (modeled)
$343.5B – $397.3B

Per-company AI share applied to forward capex

Concentration
3 names

MSFT + GOOGL + AMZN ≈ 70% of Big 5 capex

Quarterly capex by hyperscaler

Stacked, calendar-quarter, derived from quarterly 10-Q / 10-K cash flow statements.

Source: SEC EDGAR companyconcept API, us-gaap:PaymentsToAcquirePropertyPlantAndEquipment (AMZN: PaymentsToAcquireProductiveAssets). Q1–Q3 derived by differencing YTD entries.
Per-hyperscaler snapshot

Latest filed quarter, TTM run-rate, and management-stated forward guidance with source filings.

CompanyLatest QTTMYoYForward guidanceAI shareSources
MicrosoftMSFT
$30.9B
FY2026Q3 · 2026-03-31
$97.2B+58.5%
$110–$120BFY26 (Jul 2025 – Jun 2026)
75%
Meta PlatformsMETA
$19.0B
FY2026Q1 · 2026-03-31
$75.7B+73.0%
$100–$115BFY2026 (Jan – Dec 2026)
70%
AlphabetGOOGL
$35.7B
FY2026Q1 · 2026-03-31
$109.9B+90.4%
$95–$110BFY2026 (Jan – Dec 2026)
70%
AmazonAMZN
$44.2B
FY2026Q1 · 2026-03-31
$151.0B+62.2%
$150–$175BFY2026 (Jan – Dec 2026)
65%
OracleORCL
$18.6B
FY2026Q3 · 2026-02-28
$48.3B+223.1%
$30–$40BFY26 (Jun 2025 – May 2026)
90%
Latest management commentary

Structured-extraction output from live SEC 8-K filings (capex figures, forward guidance, named technologies) pulled via Claude CLI. 13 extractions in dataset (showing latest 6). Refresh with npm run extract:8k.

AI extracted
MSFT8-K filed 2026-04-29
constructive
Our AI business surpassed an annual revenue run rate of $37 billion, up 123% year-over-year.
OpenAIAzureMicrosoft 365LinkedInDynamics 365
META8-K filed 2026-04-29
constructive
We anticipate 2026 capital expenditures, including principal payments on finance leases, to be in the range of $125-145 billion, increased from our prior range of $115-135 billion.
Guidance: $125-145 billion, increased from our prior range of $115-135 billion
Meta Superintelligence Labs
GOOGL8-K filed 2026-04-29
constructive
Google Cloud revenues grew 63% with backlog nearly doubling quarter on quarter to over $460 billion.
GeminiGemini AppGemini EnterpriseGoogle Cloud Platform (GCP)Waymo
AMZN8-K filed 2026-04-29
constructive
Free cash flow decreased to $1.2 billion for the trailing twelve months, driven primarily by a year-over-year increase of $59.3 billion in purchases of property and equipment, net of proceeds from sales and incentives.
Guidance: Net sales are expected to be between $194.0 billion and $199.0 billion, or to grow between 16% and 19% compared with second quarter 2025
TrainiumGravitonNitroNVIDIA GPUsCerebrasGraviton4Trainium3AnthropicOpenAIMetaGlobalstar
AMZN8-K filed 2026-04-14
constructive
Amazon's multibillion-dollar investment in Amazon Leo is already creating broad economic opportunity across the U.S., Europe, and other regions around the world.
GlobalstarAppleMDA SpaceAmazon LeoXCOM RAN
AMZN8-K filed 2026-04-09
constructive
We're not investing approximately $200 billion in capex in 2026 on a hunch.
Guidance: approximately $200 billion in capex in 2026
NVIDIATrainiumTrainium2Trainium3Trainium4GravitonNitroBedrockSageMakerAgentCoreStrandsKiroMantleAmazon LeoAnthropicOpenAI
Where this dashboard goes next

The headline question for AM/IM allocators in 2026 is no longer is the AI capex cycle real. It is where the marginal dollar lands. The most consensus-overweight position (Mag 7 directly) captures only the operator side of the trade. The supplier side (power, cooling, memory, networking, specialized chips, and foundry capacity) is structurally underweighted and has earnings tied directly to the next four quarters of capex commitments.

The Capex Flow Map decomposes each hyperscaler's spend into infrastructure layers and routes it to specific suppliers using disclosed customer concentration. The Suppliers page surfaces backlog and disclosed exposure per name.

Data freshness
Latest filing2026-04-30
Dataset built2026-05-12
Hyperscalers5
Quarters40

Refresh: npm run fetch