콘텐츠로 건너뛰기
인사이트 > 기사

22 11월 2026

What Is AI Capex?

What Is AI Capex?

AI capex is the capital expenditure companies make on the physical infrastructure that runs artificial intelligence: data centers, servers and chips, networking equipment and the power systems that keep them running. Unlike operating costs, these purchases are recorded as long-lived assets, appear in the investing section of the cash flow statement, and reach the income statement gradually as depreciation. Investors follow AI capex because one company’s spending becomes another company’s revenue across the supply chain.

항목 AI capital expenditure (capex) Operating expense (opex)
What it buys Long-lived assets: data center buildings and land, servers, chips, networking gear, electrical and cooling equipment Costs used up in the period: salaries, electricity, maintenance, software subscriptions, rented cloud capacity
Balance sheet Recorded as property and equipment Not recorded as an asset
Cash flow statement Investing activities, usually “purchases of property and equipment” Operating activities
Income statement Reaches earnings over several years as depreciation Reaches earnings in full in the period incurred
Where investors read it Cash flow statement, capex guidance, depreciation notes Cost of revenue and operating expense lines

General accounting descriptions for comparison only. Companies apply their own policies, for example for leases and useful lives, and their filings describe them.

AI 설비투자(capex)란 무엇입니까?

AI capex is the part of a company’s capital spending that goes to computing infrastructure for training and running AI models. In practice it covers four groups of assets: buildings and land for data centers, servers with their accelerator chips, memory and storage, the networking that links thousands of servers together, and the electrical and cooling equipment that powers them.

Companies rarely report a separate “AI” line. They report total capital spending and describe what drives it. Alphabet’s quarterly report, for example, says its capital expenditures primarily reflected investments in technical infrastructure, which it defines as servers and network equipment, data center land, and building construction and improvements.1 The AI share is therefore an interpretation of company commentary, not a reported number.

How is capex different from operating expense?

Capex is different from operating expense because capex buys an asset that is used for years, while operating expense pays for something used up now. A server bought for a data center is capex. The electricity that runs it, the staff who maintain it and the software licenses it uses are operating expenses.

The same AI workload can be either, depending on who owns the hardware. A company that builds its own data center records capex. A company that rents computing time from a cloud provider records an operating expense, and the cloud provider records the capex. That is why AI capex is concentrated in the companies that own and rent out capacity.

How does AI capex show up in financial statements?

AI capex shows up first in the cash flow statement, under investing activities, usually labeled “purchases of property and equipment” or “additions to property and equipment.” The cash leaves when the company pays, but the cost does not hit earnings all at once. The asset sits on the balance sheet and is depreciated over its estimated useful life, which reduces reported earnings a little each year.

가상 예시: a company spends $10 billion on servers it expects to use for five years. Cash flow falls by $10 billion when it pays. Using straight-line depreciation, earnings carry about $2 billion of depreciation a year for five years. This example is hypothetical; actual useful lives and methods vary by company and asset type.

Microsoft reported depreciation expense of $11.1 billion for the three months ended September 30, 2026, up from $7.1 billion a year earlier.2 Rising capex today tends to mean rising depreciation in later periods.

Who spends on AI capex, and where is it reported?

The largest spenders are the big cloud providers, often called hyperscalers, along with large platform companies, AI developers and specialized data center operators. Two primary figures show the scale. Alphabet reported capital expenditures of $51.4 billion for the three months ended September 30, 2026.1 Microsoft reported additions to property and equipment of $39.4 billion for the same three months, its first quarter of fiscal year 2027.2

For U.S. public companies, reported capex appears in quarterly reports on Form 10-Q and annual reports on Form 10-K filed with the SEC, in the cash flow statement and in the liquidity discussion. Earnings releases and calls add commentary and guidance for future spending. Our article on 하이퍼스케일러 자본 지출(capex) compares the largest spenders and how they finance the buildout.

What do investors watch in AI capex?

Investors watching AI capex usually track five things.

Guidance: what companies say they plan to spend. Guidance is a plan, not a forecast of what happens, and companies revise it.

Depreciation: how quickly today’s spending turns into expense, and the useful lives companies assume for servers and buildings.

Free cash flow: operating cash flow minus capex. Heavy spending can shrink free cash flow even when earnings grow.

Return on investment: whether cloud and AI revenue grows enough to justify the spending. Microsoft’s own risk disclosure says these investments are made “in advance of fully developed revenue streams.”2

Financing: whether spending comes from operating cash flow or from debt, equity and leases.

Where does AI capex flow through the supply chain?

AI capex flows to the companies that supply each layer of a data center. Chip designers and manufacturers sell accelerators and processors; memory and storage makers sell the components around them; networking firms sell switches and optical links; construction and real estate companies build and lease the facilities; and electrical equipment makers and utilities supply transformers, switchgear, cooling and power. Our article on the AI supply chain from chips to power maps these layers, and 데이터센터의 작동 방식 explains what happens inside the building.

Thematic funds organize around parts of this chain. ETFIQ counted 12 U.S.-listed AI infrastructure ETFs as of November 17, 2026.3 A fund’s holdings show how much of it depends on a few buyers.

What are the risks of relying on AI capex?

The main risk is that capital spending moves in cycles. Companies can slow or pause spending if returns disappoint, if financing becomes more expensive, or if power and equipment shortages delay projects. Past technology buildouts have included periods of heavy investment followed by overcapacity. None of this is a forecast; it describes how spending cycles have behaved.

The second risk is concentration. A small number of buyers account for a large share of AI infrastructure demand, so a change in plans at one or two companies can affect many suppliers at once. Our 테마형 ETF 가이드 covers how to look at concentration in theme funds.

자주 묻는 질문

capex는 무엇의 약자입니까?

Capital expenditure: spending on long-lived assets such as buildings, equipment and servers, recorded as property and equipment rather than expensed immediately.

Is AI capex reported separately in company filings?

Usually not. Companies report total capital expenditures and describe in their filings and earnings calls how much of the spending relates to AI and data centers.

Where can I find a company’s capex?

In the investing section of the cash flow statement in its Form 10-Q or 10-K, usually labeled purchases or additions to property and equipment. Earnings releases often repeat the figure.

Why does capex affect earnings later?

Capex is recorded as an asset and depreciated over its useful life, so the cost reaches the income statement gradually over several years rather than in the year of purchase.

How much did Alphabet spend on capex last quarter?

Alphabet reported capital expenditures of $51.4 billion for the three months ended September 30, 2026, in its Form 10-Q filed October 29, 2026.

NOTES
  1. Alphabet Inc., Form 10-Q for the quarter ended September 30, 2026, filed October 29, 2026, capital expenditures of $51.4 billion for the three months ended September 30, 2026, primarily reflecting investments in technical infrastructure, sec.gov/Archives/edgar/data/1652044/000165204426000084/goog-20260930.htm. Read November 18, 2026.
  2. Microsoft Corporation, Form 10-Q for the quarter ended September 30, 2026, filed October 28, 2026, additions to property and equipment of $39.4 billion and depreciation expense of $11.1 billion (versus $7.1 billion a year earlier) for the three months ended September 30, 2026, and risk factor text, sec.gov/Archives/edgar/data/789019/000119312526436133/msft-20260930.htm. Read November 18, 2026.
  3. ETFIQ, AI infrastructure ETFs, 12 U.S.-listed funds as of November 17, 2026, etfiq.com/themes/ai-infra. Page read November 18, 2026.

VistaShares의 최고경영자(CEO)인 Adam Patti는 ETFIQ의 설립자이기도 합니다. ETFIQ는 ETF 발행사로부터 대가를 받지 않으며, ETFIQ의 수치는 공개 공시 자료와 시장 데이터를 기반으로 산출됩니다.

이 페이지 인용: Adam Patti, "What Is AI Capex?," VistaShares, updated 11월 19, 2026, https://www.vistashares.com/ko/what-is-ai-capex/. 일반 텍스트 버전.

IMPORTANT INFORMATION

본 자료는 교육 목적으로만 제공됩니다. 본 자료는 투자, 법률 또는 세무 자문이 아니며, 어떠한 증권의 매수 또는 매도에 대한 권유나 제안이 아닙니다. 정보는 발행일 현재 신뢰할 수 있다고 판단되는 출처에서 인용되었으며, 정확성이나 완전성이 보장되지 않고, 사전 통지 없이 변경될 수 있습니다. 제3자에게 귀속된 수치는 해당 제3자의 추정치입니다. 전망 및 예측은 미래 결과를 보장하지 않습니다. 투자자는 투자 결정을 내리기 전에 자신의 금융, 법률 및 세무 전문가와 상담해야 합니다. 투자에는 원금 손실 가능성을 포함한 위험이 수반됩니다. 옵션을 활용하는 전략에는 시장 상승에 대한 제한적 참여, 옵션이 의도한 대로 성과를 내지 못할 위험, 분배금의 변동 등 추가적인 위험이 수반됩니다. 세금 처리는 개인의 상황과 현행 법률에 따라 달라지며, 현행 법률은 변경될 수 있습니다.

투자자는 투자하기 전에 투자 목표, 위험, 수수료 및 비용을 신중히 고려해야 합니다. 이 정보 및 펀드에 관한 기타 정보가 포함된 투자설명서 또는 간이투자설명서는 (844) 875-2288로 전화하시거나 www.VistaShares.com을 방문하여 받으실 수 있습니다. 투자하기 전에 투자설명서 또는 간이투자설명서를 주의 깊게 읽으십시오.

Foreside Fund Services, LLC, 판매회사.