How much are hyperscalers spending on AI capex per quarter?

In Q1 2026, Microsoft, Alphabet, Amazon, and Meta reported a combined $129.8B of capital expenditure in a single quarter — Amazon $44.2B, Alphabet $35.7B, Microsoft $30.9B, and Meta $19.0B — versus a combined run rate of roughly $36–39B per quarter in 2022. Over the trailing four quarters (Q2 2025 through Q1 2026) the four companies spent $433.9B, all figures taken directly from 10-Q/10-K cash-flow statements filed with the SEC.

Is hyperscaler capex the same as AI spending?

Not exactly. These figures are total cash purchases of property and equipment from each company’s cash-flow statement, which include non-AI infrastructure (Amazon’s fulfillment network, office real estate, general cloud capacity). The companies themselves attribute the overwhelming majority of the increase since 2023 to AI and cloud infrastructure — data centers, servers, and networking — and the trajectory (roughly 3.4x the 2022 quarterly run rate by Q1 2026) tracks the AI buildout. Figures exclude leased capacity and off-balance-sheet vehicles such as Meta’s ~$27B Hyperion SPV.

How does hyperscaler capex compare to semiconductor supplier capex?

The four hyperscalers spent $433.9B in the trailing four quarters through Q1 2026 — roughly 4x the combined ~$110B annual capex of TSMC, Samsung, Micron, and SK Hynix tracked in the companion Semiconductor Capex dataset. The buyer side of the AI supply chain is outspending the chip-supplier side by a wide margin, because data center capex includes buildings, power, cooling, servers, and networking on top of the chips themselves.

Hyperscaler AI Capex by Company

Quarterly capital expenditure (cash purchases of property and equipment) by the four largest AI infrastructure buyers — Microsoft, Alphabet, Amazon, and Meta — from Q1 2022 through Q1 2026. The demand-side complement to the Semiconductor Capex (supplier-side) dataset: combined spend rose from ~$36B/quarter in 2022 to $129.8B in Q1 2026 ($433.9B over the trailing four quarters) as AI data center buildouts accelerated.

68 data points·Unit: $B·Last updated: Mar 2026

Historical data blurred — unlock Pro for exact values across 56 data points

56 data point values hidden. Unlock exact values, sources, and exports with Pro.

Unlock Pro

Data Table

Period	Series	Value	Range	Confidence	Source
Q1 2022	Microsoft	•••	•••	High	•••
Q1 2022	Alphabet	•••	•••	High	•••
Q1 2022	Amazon	•••	•••	High	•••
Q1 2022	Meta	•••	•••	High	•••
Q2 2022	Microsoft	•••	•••	High	•••
Q2 2022	Alphabet	•••	•••	High	•••
Q2 2022	Amazon	•••	•••	High	•••
Q2 2022	Meta	•••	•••	High	•••
Q3 2022	Microsoft	•••	•••	High	•••
Q3 2022	Alphabet	•••	•••	High	•••
Q3 2022	Amazon	•••	•••	High	•••
Q3 2022	Meta	•••	•••	High	•••
Q4 2022	Microsoft	•••	•••	High	•••
Q4 2022	Alphabet	•••	•••	High	•••
Q4 2022	Amazon	•••	•••	High	•••
Q4 2022	Meta	•••	•••	High	•••
Q1 2023	Microsoft	•••	•••	High	•••
Q1 2023	Alphabet	•••	•••	High	•••
Q1 2023	Amazon	•••	•••	High	•••
Q1 2023	Meta	•••	•••	High	•••
Q2 2023	Microsoft	•••	•••	High	•••
Q2 2023	Alphabet	•••	•••	High	•••
Q2 2023	Amazon	•••	•••	High	•••
Q2 2023	Meta	•••	•••	High	•••
Q3 2023	Microsoft	•••	•••	High	•••
Q3 2023	Alphabet	•••	•••	High	•••
Q3 2023	Amazon	•••	•••	High	•••
Q3 2023	Meta	•••	•••	High	•••
Q4 2023	Microsoft	•••	•••	High	•••
Q4 2023	Alphabet	•••	•••	High	•••
Q4 2023	Amazon	•••	•••	High	•••
Q4 2023	Meta	•••	•••	High	•••
Q1 2024	Microsoft	•••	•••	High	•••
Q1 2024	Alphabet	•••	•••	High	•••
Q1 2024	Amazon	•••	•••	High	•••
Q1 2024	Meta	•••	•••	High	•••
Q2 2024	Microsoft	•••	•••	High	•••
Q2 2024	Alphabet	•••	•••	High	•••
Q2 2024	Amazon	•••	•••	High	•••
Q2 2024	Meta	•••	•••	High	•••
Q3 2024	Meta	•••	•••	High	•••
Q3 2024	Microsoft	•••	•••	High	•••
Q3 2024	Alphabet	•••	•••	High	•••
Q3 2024	Amazon	•••	•••	High	•••
Q4 2024	Meta	•••	•••	High	•••
Q4 2024	Microsoft	•••	•••	High	•••
Q4 2024	Alphabet	•••	•••	High	•••
Q4 2024	Amazon	•••	•••	High	•••
Q1 2025	Meta	•••	•••	High	•••
Q1 2025	Microsoft	•••	•••	High	•••
Q1 2025	Alphabet	•••	•••	High	•••
Q1 2025	Amazon	•••	•••	High	•••
Q2 2025	Meta	•••	•••	High	•••
Q2 2025	Microsoft	•••	•••	High	•••
Q2 2025	Alphabet	•••	•••	High	•••
Q2 2025	Amazon	•••	•••	High	•••
Q3 2025	Meta	18.829	—	High	META Q3 2025 10-Q (SEC 0001628280-25-047240) (Oct 2025)
Q3 2025	Microsoft	19.394	—	High	MSFT FY26 Q1 10-Q (SEC 0001193125-25-256321) (Oct 2025)
Q3 2025	Alphabet	23.953	—	High	GOOGL Q3 2025 10-Q (SEC 0001652044-25-000091) (Oct 2025)
Q3 2025	Amazon	35.095	—	High	AMZN Q3 2025 10-Q (SEC 0001018724-25-000123) (Oct 2025)
Q4 2025	Meta	21.383	—	High	META 2025 10-K (SEC 0001628280-26-003942) (Jan 2026)
Q4 2025	Microsoft	29.876	—	High	MSFT FY26 Q2 10-Q (SEC 0001193125-26-027207) (Jan 2026)
Q4 2025	Alphabet	27.851	—	High	GOOGL 2025 10-K (SEC 0001652044-26-000018) (Feb 2026)
Q4 2025	Amazon	39.522	—	High	AMZN 2025 10-K (SEC 0001018724-26-000004) (Feb 2026)
Q1 2026	Meta	18.997	—	High	META Q1 2026 10-Q (SEC 0001628280-26-028526) (Apr 2026)
Q1 2026	Microsoft	30.876	—	High	MSFT FY26 Q3 10-Q (SEC 0001193125-26-191507) (Apr 2026)
Q1 2026	Alphabet	35.674	—	High	GOOGL Q1 2026 10-Q (SEC 0001652044-26-000048) (Apr 2026)
Q1 2026	Amazon	44.203	—	High	AMZN Q1 2026 10-Q (SEC 0001018724-26-000014) (Apr 2026)

Methodology & Sources

Quarterly purchases of property and equipment as reported in 10-Q/10-K cash-flow statements (SEC EDGAR); every point cites its filing accession number. Microsoft fiscal quarters (FY ends June 30) are mapped to calendar quarters. Where filings disclose only year-to-date cash flows, discrete quarters are derived by differencing (SA computation, noted per point). Meta 2022-2023 values reflect current (restated) presentation. Demand-side complement to the Semiconductor Capex (supplier-side) dataset.

Citations:

MSFT FY22 Q3 10-Q (SEC 0001564590-22-015675) (Apr 2022)
GOOGL Q1 2022 10-Q (SEC 0001652044-22-000029) (Apr 2022)
AMZN Q1 2022 10-Q (SEC 0001018724-22-000013) (Apr 2022)
META Q1 2022 10-Q (SEC 0001326801-22-000057) (Apr 2022)
MSFT FY22 10-K (SEC 0001564590-22-026876) (Jul 2022)
GOOGL Q2 2022 10-Q (SEC 0001652044-22-000071) (Jul 2022)
AMZN Q2 2022 10-Q (SEC 0001018724-22-000019) (Jul 2022)
META Q2 2022 10-Q (SEC 0001326801-22-000082) (Jul 2022)
MSFT FY23 Q1 10-Q (SEC 0001564590-22-035087) (Oct 2022)
GOOGL Q3 2022 10-Q (SEC 0001652044-22-000090) (Oct 2022)
AMZN Q3 2022 10-Q (SEC 0001018724-22-000023) (Oct 2022)
META Q3 2022 10-Q (SEC 0001326801-22-000108) (Oct 2022)
MSFT FY23 Q2 10-Q (SEC 0001564590-23-000733) (Jan 2023)
GOOGL 2022 10-K (SEC 0001652044-23-000016) (Feb 2023)
AMZN 2022 10-K (SEC 0001018724-23-000004) (Feb 2023)
META 2024 10-K (SEC 0001326801-25-000017) (Jan 2025)
MSFT FY23 Q3 10-Q (SEC 0000950170-23-014423) (Apr 2023)
GOOGL Q1 2023 10-Q (SEC 0001652044-23-000045) (Apr 2023)
AMZN Q1 2023 10-Q (SEC 0001018724-23-000008) (Apr 2023)
META Q1 2024 10-Q (SEC 0001326801-24-000049) (Apr 2024)
MSFT FY23 10-K (SEC 0000950170-23-035122) (Jul 2023)
GOOGL Q2 2023 10-Q (SEC 0001652044-23-000070) (Jul 2023)
AMZN Q2 2023 10-Q (SEC 0001018724-23-000012) (Aug 2023)
META Q2 2024 10-Q (SEC 0001326801-24-000069) (Aug 2024)
MSFT FY24 Q1 10-Q (SEC 0000950170-23-054855) (Oct 2023)
GOOGL Q3 2023 10-Q (SEC 0001652044-23-000094) (Oct 2023)
AMZN Q3 2023 10-Q (SEC 0001018724-23-000018) (Oct 2023)
META Q3 2024 10-Q (SEC 0001326801-24-000081) (Oct 2024)
MSFT FY24 Q2 10-Q (SEC 0000950170-24-008814) (Jan 2024)
GOOGL 2023 10-K (SEC 0001652044-24-000022) (Jan 2024)
AMZN 2023 10-K (SEC 0001018724-24-000008) (Feb 2024)
MSFT FY24 Q3 10-Q (SEC 0000950170-24-048288) (Apr 2024)
GOOGL Q1 2024 10-Q (SEC 0001652044-24-000053) (Apr 2024)
AMZN Q1 2024 10-Q (SEC 0001018724-24-000083) (May 2024)
MSFT FY24 10-K (SEC 0000950170-24-087843) (Jul 2024)
GOOGL Q2 2024 10-Q (SEC 0001652044-24-000079) (Jul 2024)
AMZN Q2 2024 10-Q (SEC 0001018724-24-000130) (Aug 2024)
MSFT FY25 Q1 10-Q (SEC 0000950170-24-118967) (Oct 2024)
GOOGL Q3 2024 10-Q (SEC 0001652044-24-000118) (Oct 2024)
AMZN Q3 2024 10-Q (SEC 0001018724-24-000161) (Nov 2024)
MSFT FY25 Q2 10-Q (SEC 0000950170-25-010491) (Jan 2025)
GOOGL 2024 10-K (SEC 0001652044-25-000014) (Feb 2025)
AMZN 2024 10-K (SEC 0001018724-25-000004) (Feb 2025)
META Q1 2025 10-Q (SEC 0001326801-25-000054) (May 2025)
MSFT FY25 Q3 10-Q (SEC 0000950170-25-061046) (Apr 2025)
GOOGL Q1 2025 10-Q (SEC 0001652044-25-000043) (Apr 2025)
AMZN Q1 2025 10-Q (SEC 0001018724-25-000036) (May 2025)
META Q2 2025 10-Q (SEC 0001628280-25-036791) (Jul 2025)
MSFT FY25 10-K (SEC 0000950170-25-100235) (Jul 2025)
GOOGL Q2 2025 10-Q (SEC 0001652044-25-000062) (Jul 2025)
AMZN Q2 2025 10-Q (SEC 0001018724-25-000086) (Aug 2025)
META Q3 2025 10-Q (SEC 0001628280-25-047240) (Oct 2025)
MSFT FY26 Q1 10-Q (SEC 0001193125-25-256321) (Oct 2025)
GOOGL Q3 2025 10-Q (SEC 0001652044-25-000091) (Oct 2025)
AMZN Q3 2025 10-Q (SEC 0001018724-25-000123) (Oct 2025)
META 2025 10-K (SEC 0001628280-26-003942) (Jan 2026)
MSFT FY26 Q2 10-Q (SEC 0001193125-26-027207) (Jan 2026)
GOOGL 2025 10-K (SEC 0001652044-26-000018) (Feb 2026)
AMZN 2025 10-K (SEC 0001018724-26-000004) (Feb 2026)
META Q1 2026 10-Q (SEC 0001628280-26-028526) (Apr 2026)
MSFT FY26 Q3 10-Q (SEC 0001193125-26-191507) (Apr 2026)
GOOGL Q1 2026 10-Q (SEC 0001652044-26-000048) (Apr 2026)
AMZN Q1 2026 10-Q (SEC 0001018724-26-000014) (Apr 2026)

Related Analysis

The Hyperscaler Capex Wall: $434B of Buyer-Side AI Spend, the Depreciation Lag, and Why Big Tech Borrows

Read analysis →

OpenAI & Google $84B AI Push Signals Custom Silicon War

Read analysis →

← All Market Data

Frequently Asked Questions

How much are hyperscalers spending on AI capex per quarter?: In Q1 2026, Microsoft, Alphabet, Amazon, and Meta reported a combined $129.8B of capital expenditure in a single quarter — Amazon $44.2B, Alphabet $35.7B, Microsoft $30.9B, and Meta $19.0B — versus a combined run rate of roughly $36–39B per quarter in 2022. Over the trailing four quarters (Q2 2025 through Q1 2026) the four companies spent $433.9B, all figures taken directly from 10-Q/10-K cash-flow statements filed with the SEC.
Is hyperscaler capex the same as AI spending?: Not exactly. These figures are total cash purchases of property and equipment from each company’s cash-flow statement, which include non-AI infrastructure (Amazon’s fulfillment network, office real estate, general cloud capacity). The companies themselves attribute the overwhelming majority of the increase since 2023 to AI and cloud infrastructure — data centers, servers, and networking — and the trajectory (roughly 3.4x the 2022 quarterly run rate by Q1 2026) tracks the AI buildout. Figures exclude leased capacity and off-balance-sheet vehicles such as Meta’s ~$27B Hyperion SPV.
How does hyperscaler capex compare to semiconductor supplier capex?: The four hyperscalers spent $433.9B in the trailing four quarters through Q1 2026 — roughly 4x the combined ~$110B annual capex of TSMC, Samsung, Micron, and SK Hynix tracked in the companion Semiconductor Capex dataset. The buyer side of the AI supply chain is outspending the chip-supplier side by a wide margin, because data center capex includes buildings, power, cooling, servers, and networking on top of the chips themselves.

Hyperscaler AI Capex by Company — Historical Time Series Data