TechnologyArtificial IntelligenceSemiconductors

How Big Is the AI Market When You Count Only the Money Companies Report?

Nvidia sold $89.0 billion of data center products in the quarter to July 2026, and the four largest cloud and internet companies spent $202.0 billion on infrastructure in the quarter to September 2026. This page sizes the AI market from what chip makers, hyperscalers and AI developers report in their own filings and announcements, layer by layer, and sets those numbers against the latest Gartner forecast.

Ali
Ali
Content Writer · Updated 6 November 2026
Methodology
Data: earnings releases, filings and earnings calls of Nvidia, TSMC, Broadcom, AMD, Amazon, Alphabet, Microsoft and Meta for quarters ending July to September 2026; company announcements by Anthropic and OpenAI.
Note: layers of the supply chain overlap and are not added together. Run rates are company statements. The only forecast is the Gartner May 2026 release, labelled as such. Calculations are by BusinessStats.
$89.0BNvidia Data Center, Q2 FY27
$202.0BBig Tech Capex, Q3 2026
$16.7BBroadcom AI, Q3 FY26
$47B+Anthropic Run Rate, May 2026
$37B+Microsoft AI Run Rate
$2.60TGartner 2026 Forecast
$89.0BNvidia DC Quarter
$202.0BCapex Quarter
$47B+Anthropic Run Rate
$2.60T2026 Forecast
Key Takeaways
  • Nvidia data center revenue reached $89.0 billion in the quarter to July 2026, up 117 percent, and $277.8 billion over four quarters.
  • Amazon, Alphabet, Microsoft and Meta spent $202.0 billion on property and equipment in the September 2026 quarter, more than double a year earlier, and guide to about $730 billion for 2026.
  • Outside Nvidia, Broadcom AI chip revenue tripled to $16.7 billion, AMD data center revenue almost doubled to $8.5 billion and high performance computing reached 68 percent of TSMC revenue.
  • AI developers report fast growing revenue: Anthropic passed a $47 billion run rate in May 2026, Microsoft AI business $37 billion and OpenAI $2 billion a month.
  • Gartner forecast in May 2026 that worldwide AI spending will reach $2.60 trillion in 2026; this is a forecast, and it is several times larger than any reported company total because it uses a much wider definition.

The AI market in money that has actually changed hands

Quick answer

Measured by reported money rather than forecasts, the AI market now moves well over $100 billion a quarter through its core suppliers. Nvidia alone booked $89.0 billion of data center revenue in the quarter to July 2026, up 117 percent in a year. Amazon, Alphabet, Microsoft and Meta spent $202.0 billion on property and equipment in the September 2026 quarter, mostly on AI data centers. On the revenue side, Anthropic said its run rate crossed $47 billion in May 2026 and Microsoft put its AI business above $37 billion in the quarter to March 2026. The broadest tracker, Gartner, forecast in May 2026 that worldwide AI spending will reach $2.60 trillion in 2026, a forecast rather than a count.

Most pages that answer the question of how big the AI market is quote a single number from a consultancy report, often a figure that cannot be checked and that differs from the next report by hundreds of billions of dollars. This page takes another route. It adds up what the companies at the centre of the AI economy report in their own filings: what chip makers sell, what the largest cloud and internet companies spend, and what AI developers say they earn. Every figure has a period and a primary source, and the only forecast on the page comes from the original publisher and is labelled as such.

The result is not one tidy total, because the money passes through several hands. A dollar of AI spending by a cloud company becomes revenue for Nvidia, part of which becomes revenue for TSMC, and the cloud company then tries to earn it back by renting the capacity to model developers and businesses. Counting each layer separately shows where the money sits today and how fast each layer is growing. For adoption and usage data, such as how many businesses and people use AI, see our broader artificial intelligence statistics hub.

$89.0B
Nvidia data center revenue
Quarter to July 2026, up 117 percent.
$202.0B
Big Tech capex in one quarter
Amazon, Alphabet, Microsoft and Meta, September 2026 quarter.
$47B+
Anthropic run-rate revenue
Crossed in May 2026, company statement.
$2.60T
Gartner AI spending forecast
2026 forecast, published 19 May 2026.

Why reported revenue and capex are a better yardstick than market size reports

Market size reports differ mainly because they define the market differently. Some count only AI software, some add chips and servers, and some include consulting and the AI share of products such as smartphones and personal computers. A report that includes AI-capable phones will produce a number several times larger than one that counts only software licences. None of these choices is wrong, but they make headline totals hard to compare and easy to inflate.

Company filings avoid most of that problem. Nvidia, Broadcom, AMD and TSMC report audited revenue every quarter, and three of them now break out AI or data center lines. The cloud and internet companies report how much cash they spend on property and equipment, and their managements say on earnings calls how much of it goes to AI. AI developers such as Anthropic and OpenAI are private, so their figures are limited to what they publish themselves, usually as an annualized run rate. These numbers are narrower than a full market estimate, but they are real, recent and comparable over time.

Layer 1
Foundries
TSMC makes most AI chips. HPC was 68 percent of its revenue in Q3 2026.
Layer 2
Chip and system makers
Nvidia, Broadcom and AMD sell accelerators, networking and processors.
Layer 3
Data center builders
Hyperscalers buy the chips. Their capex is the demand signal.
Layer 4
Clouds and model makers
AWS, Azure, Google Cloud, Anthropic and OpenAI sell compute and models.
Layer 5
Users
Businesses and consumers pay for AI services and subscriptions.
Read the numbers on this page this way
  • Do not add the layers together. TSMC revenue is part of what Nvidia pays, and Nvidia revenue is part of what the hyperscalers spend, so a sum would count the same dollar several times.
  • Capex is spending, not revenue. It shows how much money the largest buyers are committing, which is the closest thing to a reported measure of AI demand.
  • Run rates are annualized snapshots. A run rate of $47 billion means the latest month or period multiplied out to a year, not revenue already earned in a calendar year.
  • Fiscal years differ. Nvidia ends its year in late January, Broadcom in early November and Microsoft in June, so each figure carries its own quarter label.

Nvidia data center revenue: from $22.6 billion to $89.0 billion a quarter in two years

Nvidia is the single best gauge of AI hardware spending because its data center platform, which covers GPUs, complete rack systems and networking, is by far the largest AI accelerator business among the companies that report one. In the second quarter of fiscal 2027, which ended on 26 July 2026, data center revenue reached $89.0 billion, up 18 percent from the previous quarter and 117 percent from a year earlier. It made up 92.5 percent of total company revenue of $96.2 billion.

Two years earlier, in the quarter to April 2024, the same business brought in $22.6 billion. Quarterly data center revenue has therefore almost quadrupled in nine quarters, and the pace has picked up again since late 2025 as Blackwell systems ramped and the Vera Rubin platform, which Nvidia said in August 2026 was in full production, began to ramp. Over the four quarters to July 2026, data center revenue added up to $277.8 billion.

Nvidia Data Center Revenue by Fiscal Quarter, Q1 FY2025 to Q2 FY2027 (billion US dollars)
Quarterly revenue of the Data Center market platform, as recast by Nvidia in 2026. Month in brackets is when each quarter ended.
Switch to the table for exact figures.

Steeper again in 2026: data center revenue rose by $12.9 billion in the quarter to April 2026 and by $13.8 billion in the quarter to July 2026, the two largest quarterly gains in the series.

Nvidia revenue by market platform, fiscal 2025 to Q2 fiscal 2027 (US dollars)
Quarter (month ended)Data Centerof which Hyperscaleof which ACIEEdge ComputingTotal revenueData Center growth vs year earlier
Q1 FY25 (Apr 2024)$22.6B$10.9B$11.6B$3.5B$26.0Bn/a
Q2 FY25 (Jul 2024)$26.3B$13.4B$12.9B$3.8B$30.0Bn/a
Q3 FY25 (Oct 2024)$30.8B$15.6B$15.1B$4.3B$35.1Bn/a
Q4 FY25 (Jan 2025)$35.6B$19.9B$15.6B$3.8B$39.3Bn/a
Q1 FY26 (Apr 2025)$39.1B$22.3B$16.9B$5.0B$44.1B+73.3%
Q2 FY26 (Jul 2025)$41.1B$24.2B$16.9B$5.6B$46.7B+56.4%
Q3 FY26 (Oct 2025)$51.2B$35.1B$16.1B$5.8B$57.0B+66.4%
Q4 FY26 (Jan 2026)$62.3B$42.2B$20.1B$5.8B$68.1B+75.1%
Q1 FY27 (Apr 2026)$75.2B$43.0B$32.2B$6.4B$81.6B+92.4%
Q2 FY27 (Jul 2026)$89.0B$48.7B$40.3B$7.2B$96.2B+116.6%

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

ACIE stands for AI Clouds, Industrial and Enterprise, the Nvidia name for data center customers other than the largest cloud providers. Figures are rounded from millions as reported. Growth rates are BusinessStats calculations. Fiscal 2025 data center revenue added up to $115.2 billion and fiscal 2026 to $193.7 billion.

The fiscal year totals show the scale of the jump. Data center revenue was $115.2 billion in fiscal 2025 and $193.7 billion in fiscal 2026, a rise of 68 percent, and the first half of fiscal 2027 alone has already produced $164.3 billion. For the outlook, Nvidia guided to total revenue of $108.0 billion, plus or minus 2 percent, for the quarter ending October 2026, without assuming any data center compute sales to China. A company level profile with products, margins and share price history is on our Nvidia statistics page.

Hyperscalers still buy most Nvidia systems, but other AI clouds and enterprises are catching up

Since its 2026 recast, Nvidia splits data center revenue into two customer groups. Hyperscale covers the largest public cloud and internet companies. AI Clouds, Industrial and Enterprise, shortened to ACIE, covers specialised GPU cloud providers, sovereign and national AI projects, industrial companies and large enterprises buying their own systems. The split is a useful reality check on claims that AI demand rests on a handful of buyers.

Hyperscale revenue more than doubled to $48.7 billion in the quarter to July 2026, from $24.2 billion a year earlier. ACIE grew even faster, to $40.3 billion from $16.9 billion, an increase of 138 percent. As a result the hyperscale share of data center revenue fell from 67.7 percent in the quarter to January 2026 to 54.7 percent in the quarter to July 2026. Demand is broadening beyond the four or five largest spenders, which matters because it reduces the dependence of the whole AI supply chain on a few capital budgets.

Nvidia Data Center Revenue by Customer Group, Q1 FY2026 to Q2 FY2027 (billion US dollars)
Hyperscale versus AI Clouds, Industrial and Enterprise (ACIE), by fiscal quarter.
Switch to the table for exact figures.

ACIE doubled in two quarters: revenue from AI clouds, industrial and enterprise buyers rose from $20.1 billion in the quarter to January 2026 to $40.3 billion in the quarter to July 2026.

TSMC, Broadcom and AMD: AI revenue across the rest of the chip supply chain

Nvidia is the largest seller, but three other companies report numbers that show how much AI money flows through the semiconductor industry. TSMC, which manufactures the chips designed by Nvidia, AMD, Broadcom and most cloud companies, reported revenue of $46.72 billion in the third quarter of 2026, up 41.2 percent in US dollar terms. High performance computing, the platform that contains AI accelerators and server processors, made up 68 percent of that revenue, up from 57 percent a year earlier. Applied to total revenue, that share equals roughly $31.8 billion in a single quarter.

Broadcom, which designs custom AI accelerators for cloud companies and sells the Ethernet networking chips that link them, reported AI semiconductor revenue of $16.7 billion in its third fiscal quarter, which ended on 2 August 2026. That was 221 percent more than a year earlier and 56 percent of total company revenue of $29.6 billion. Broadcom guided to $21.7 billion of AI semiconductor revenue for the fourth fiscal quarter. AMD reported data center segment revenue of $8.5 billion in the quarter to 26 September 2026, up 97 percent, helped by the launch of its Instinct MI450 accelerators and demand for EPYC server processors.

AI-Related Chip Revenue in the Latest Reported Quarter, 2026 (billion US dollars)
Each company reports a different line and quarter; the bars are not additive, because TSMC revenue is partly paid by Nvidia, Broadcom and AMD.
Switch to the table for exact figures.

Nvidia is still several times larger: its data center revenue in one quarter was more than five times Broadcom AI revenue and ten times AMD data center revenue.

AI chip revenue lines reported by the four largest suppliers
CompanyLine reportedLatest quarterRevenueGrowth vs year earlierNext quarter guidance
NvidiaData Center platformQ2 FY2027, to 26 Jul 2026$89.0B+117%Total revenue $108.0B, plus or minus 2%
TSMCHPC platform (68% of revenue)Q3 2026, to 30 Sep 2026about $31.8BHPC share 57% a year earlierTotal revenue $52.3B to $54.3B
BroadcomAI semiconductor revenueQ3 FY2026, to 2 Aug 2026$16.7B+221%AI revenue $21.7B
AMDData Center segmentQ3 2026, to 26 Sep 2026$8.5B+97%Total revenue about $17B, plus or minus 3%

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

Guidance is company guidance for the following quarter, not a result. TSMC also expects full year 2026 revenue growth of around 44 percent in US dollar terms and kept its 2026 capital budget at $60 billion to $64 billion.

The Broadcom series is the clearest sign that custom chips built for individual cloud companies are becoming a second engine next to Nvidia GPUs. AI semiconductor revenue was $8.4 billion in the quarter to February 2026 and $10.8 billion in the quarter to May 2026 before jumping to $16.7 billion. TSMC management said on its October 2026 call that demand is still far above what it can supply, and that the mid to high 50 percent annual growth it had guided for AI accelerators between 2024 and 2029 will probably be exceeded. For the wider processor market behind these numbers, see the Intel and AMD x86 market share data.

Year-over-Year Revenue Growth in the Latest Reported Quarter, 2026 (%)
Growth as reported by each company for its AI or data center line; TSMC shown for total revenue in US dollars.
Switch to the table for exact figures.

Every supplier more than grew in line with demand: the smallest gain among the AI lines was still close to a doubling in twelve months.

Big Tech capex: $202.0 billion in one quarter, most of it for AI

Capital expenditure by the largest cloud and internet companies is the demand side of the chip numbers above. In the September 2026 quarter, Amazon, Alphabet, Microsoft and Meta together spent $202.0 billion in cash on property and equipment, compared with $97.3 billion in the same quarter of 2025, an increase of 107.6 percent. The companies tie most of it to AI: Amazon said the increase in its purchases of property and equipment primarily reflects investments in artificial intelligence, Microsoft said roughly three quarters of its quarterly capex went to short-lived assets such as CPUs and GPUs, and Alphabet said about 65 percent of its $51.4 billion went to servers and networking.

Share of Big Tech Capital Expenditure, Quarter Ended September 2026
Cash spent on property and equipment by the four largest AI infrastructure buyers in one quarter, $202.0 billion in total.
Switch to the table for exact figures.

Amazon alone spent $77.3 billion in a quarter, 38 percent of the total, and its spending over the twelve months to September 2026 reached $215.2 billion.

Capital expenditure of the four largest AI infrastructure buyers (US dollars)
CompanyQ3 2025Q3 2026ChangeCompany guidance for calendar 2026
Amazon$35.1B$77.3B+120%About $220B cash capex
Alphabet$24.0B$51.4B+115%$195B to $205B
Microsoft$19.4B$39.4B+103%About $170B, including finance leases
Meta Platforms$18.8B$33.9B+80%$135B to $145B, including finance lease principal
Total$97.3B$202.0B+107.6%About $730B at range midpoints

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

Quarterly figures are cash purchases or additions to property and equipment for the quarter ended 30 September. Microsoft fiscal year ends in June, so its September 2026 quarter is the first quarter of fiscal 2027. Guidance is as stated on the October 2026 results and is not measured on exactly the same basis for each company. Changes and the guidance total are BusinessStats calculations.

Put next to the chip numbers, the capex figures show where the money goes. The four companies spent $202.0 billion in the September quarter, while Nvidia booked $48.7 billion of hyperscale data center revenue in its quarter to July 2026. Chips are the largest single item, but the rest pays for buildings, power connections, cooling, memory, storage, networking from other suppliers and in-house chips such as Google TPUs and Amazon Trainium. A detailed comparison of revenue, margins, free cash flow and capex guidance for the same companies is on our Big Tech revenue comparison, and Alphabet figures over a longer period are in our Alphabet annual revenue data.

What the capex numbers say about the AI market
  • Capex roughly doubled in a year at all four companies. Microsoft trimmed its calendar 2026 expectation to about $170 billion because some third party capacity deliveries were delayed, while Meta narrowed its range to $135 billion to $145 billion from $130 billion to $145 billion.
  • Alphabet said it expects capex to increase significantly again in 2027, and Meta said it is still developing its 2027 budget while continuing to invest aggressively in AI.

Where the capacity is sold: cloud revenue and the backlog already signed

The hyperscalers earn back part of their spending by renting computing capacity, increasingly AI capacity, through their cloud units. In the September 2026 quarter, Amazon Web Services had net sales of $46.0 billion, up 39.3 percent, and an annualized revenue run rate of $184 billion. Google Cloud revenue rose 103 percent to $30.7 billion with operating income of $11.1 billion. Microsoft said Azure revenue grew 48 percent, within total Microsoft Cloud revenue of $63.9 billion. Market share data for these providers is in our analysis of the cloud infrastructure market.

A more forward looking measure is the value of contracts already signed but not yet delivered. Google Cloud ended September 2026 with a backlog of $561 billion, against $155 billion a year earlier, and management expects just over 55 percent of it to turn into revenue over the next 24 months. Microsoft reported commercial remaining performance obligation of $677 billion, up 73 percent. These are not AI-only figures, but large AI compute contracts with model developers are the main reason they have grown so fast. Revenue mix at Amazon is covered in our page on Amazon net revenue by segment.

Cloud revenue and contracted backlog, quarter ended September 2026
MeasureValueComparisonReported by
AWS net sales$46.0B$33.0B a year earlier (+39.3%)Amazon
AWS annualized revenue run rate$184Bn/aAmazon
Google Cloud revenue$30.7B+103% year over yearAlphabet
Google Cloud operating income$11.1B36.2% marginAlphabet
Google Cloud backlog$561B$155B at end of Q3 2025Alphabet
Microsoft Cloud revenue$63.9B+30% year over yearMicrosoft
Azure revenue growth+48%Guided 49% to 50% in constant currency for the next quarterMicrosoft
Microsoft commercial remaining performance obligation$677B+73% year over yearMicrosoft

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

What AI developers and software companies say they earn

Further down the chain, a few companies now disclose how much revenue their AI products generate. These are the closest reported figures to the size of the market for AI models and applications, but they come with a caveat: private companies publish run rates rather than audited annual revenue, and the definitions are their own.

Anthropic has given the most frequent updates. The company said its run-rate revenue was approximately $1 billion at the start of 2025, over $5 billion by August 2025 and about $9 billion at the end of 2025. It reached $14 billion in February 2026, surpassed $30 billion in April 2026 and crossed $47 billion in May 2026, when the company announced a $65 billion Series H funding round at a $965 billion post-money valuation. In April 2026 it said the number of business customers each spending at least $1 million a year exceeded 1,000, up from over 500 in February.

Anthropic Run-Rate Revenue as Stated by the Company, 2025 to 2026 (billion US dollars)
Annualized run-rate revenue at each date Anthropic disclosed it. Values are lower bounds where the company said over or surpassed.
Switch to the table for exact figures.

More than fivefold in five months: the stated run rate went from about $9 billion at the end of 2025 to over $47 billion in May 2026.

Microsoft said its AI business passed an annual revenue run rate of $37 billion in the quarter to March 2026, growing 123 percent year over year. It has not updated that figure since, but said in October 2026 that Foundry, its platform for building AI applications, had more than 120,000 customers with revenue more than tripling. OpenAI chief financial officer Sarah Friar wrote in January 2026 that annualized revenue grew from $2 billion in 2023 to $6 billion in 2024 and more than $20 billion in 2025. In March 2026, announcing $122 billion of committed capital at an $852 billion valuation, OpenAI said it was generating $2 billion of revenue per month, that ChatGPT had more than 900 million weekly active users and over 50 million subscribers, and that enterprise customers made up more than 40 percent of revenue.

AI Revenue Run Rates Disclosed by the Companies Themselves, 2026 (billion US dollars)
Latest annualized figure each company has published; dates differ.
Switch to the table for exact figures.

Disclosed AI revenue is now in the tens of billions per company, but it is still a fraction of the $202.0 billion the four largest buyers spent on infrastructure in a single quarter.

AI revenue and usage figures published by companies
CompanyFigureValueDate or period
AnthropicRun-rate revenueover $47BMay 2026
AnthropicBusiness customers spending $1M or more a yearover 1,000April 2026
AnthropicClaude Code run-rate revenueover $2.5BFebruary 2026
MicrosoftAI business annual revenue run rateover $37B (+123%)Quarter to March 2026
OpenAIRevenue per month$2BMarch 2026
OpenAIAnnualized revenueover $20B2025
OpenAIChatGPT weekly active usersover 900 millionMarch 2026
AlphabetGemini app users1 billionQ3 2026 earnings call
AlphabetFirst-party model API tokens per minutenearly 32 billionQ3 2026, over 4 times a year earlier

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

Google does not disclose a revenue figure for Gemini, but its usage numbers give a sense of scale: the Gemini app reached one billion users, and its first-party model APIs were processing nearly 32 billion tokens per minute in the third quarter of 2026, more than four times the level a year earlier. Meta does not report AI revenue separately; its return on AI spending shows up in advertising, where total company revenue, almost all of it advertising, rose 26 percent to $64.6 billion in the September 2026 quarter. For how these companies rank by stock market value, see the biggest companies in the world by market value.

The gap between what is spent on AI and what AI earns

Comparing the layers shows the central question of the AI market in 2026. The four largest buyers spent $202.0 billion on property and equipment in one quarter, an annual pace of roughly $800 billion. The largest disclosed AI revenue run rates, Anthropic at over $47 billion, Microsoft at over $37 billion and OpenAI at about $24 billion, add up to roughly $108 billion a year even before allowing for overlap between them. Cloud revenue is far larger, but most of it still comes from non-AI workloads.

That gap is not proof of a bubble or of its absence. Data centers last for many years while chips are depreciated over a few, contracted backlogs of $561 billion at Google Cloud and $677 billion of commercial obligations at Microsoft point to revenue that has been signed but not yet recognized, and AI developer run rates have been multiplying within months. But it does mean that the reported revenue of the AI market today is still much smaller than the reported investment, and that the investment is being paid for largely out of the cash flow of a few very profitable companies. Readers who want to set these sums against the size of national economies can compare them with our list of countries with the largest GDP.

$202.0B
Capex, four companies, one quarter
September 2026 quarter.
$108B
Sum of three disclosed AI run rates
Anthropic, Microsoft and OpenAI, latest figures; may overlap.
$1.24T
Signed cloud commitments
Google Cloud backlog plus Microsoft commercial RPO, September 2026.
$277.8B
Nvidia data center revenue
Four quarters to July 2026.

Power is the other physical limit on how fast this spending can turn into capacity, which is why AI companies increasingly describe their compute in gigawatts of electricity rather than in numbers of chips; OpenAI, for example, said its available compute rose from 0.6 gigawatts in 2024 to about 1.9 gigawatts in 2025. Electricity market data is in our page on the global electricity market.

Forecast: Gartner expects $2.60 trillion of AI spending in 2026

For readers who need a single all-in figure, the most widely used tracker is the Gartner forecast of worldwide AI spending. It counts what buyers spend across eight segments: AI infrastructure, services, software, models, data science and application development platforms, cybersecurity and data. This is a forecast, not a measurement. In its release of 19 May 2026, Gartner put worldwide AI spending at $1.76 trillion in 2025 and forecast $2.60 trillion for 2026, growth of 47 percent, and $3.49 trillion for 2027. In January 2026 the same tracker had forecast $2.53 trillion for 2026, so the estimate was raised by about $68 billion within four months.

Forecast: Worldwide AI Spending by Segment, 2026 (billion US dollars)
Gartner forecast published 19 May 2026. Forecast, not reported revenue.
Switch to the table for exact figures.

Infrastructure dominates the forecast: 55 percent of the 2026 total is AI infrastructure, which is consistent with what chip makers and hyperscalers report.

Gartner forecast of worldwide AI spending by segment (US dollars, published May 2026)
Segment20252026 forecast2027 forecastGrowth 2026
AI infrastructure$975.6B$1431.5B$1890.3B+47%
AI services$436.4B$585.5B$759.4B+34%
AI software$282.9B$453.2B$638.4B+60%
AI cybersecurity$25.9B$51.3B$86.0B+98%
AI models$15.5B$32.6B$59.2B+110%
AI platforms for data science and machine learning$21.3B$29.9B$42.6B+41%
AI application development platforms$6.6B$8.4B$10.9B+28%
AI data$0.8B$3.1B$6.5B+278%
Total$1,764.9B$2,595.7B$3,493.4B+47%

Download this table (CSV) · Free to reuse under CC BY 4.0 with credit to Business Stats

Source: Gartner, 19 May 2026. 2025 values are Gartner estimates; 2026 and 2027 are forecasts. Growth rates are BusinessStats calculations from the table. Earlier BusinessStats editions used older and secondary estimates that have been replaced.

The Gartner numbers and the company figures above are measuring different things, and the difference is instructive. Gartner AI infrastructure spending of $1.43 trillion in 2026 is much larger than the roughly $730 billion of 2026 capex guided by the four largest buyers, because it covers every buyer worldwide, including other cloud providers, enterprises and governments, rather than four companies. The software and services segments, $1.04 trillion together in 2026, include work such as consulting and AI features inside existing software that no company reports as a separate AI line. Gartner itself issued a further update in September 2026; the May release is the most recent edition that BusinessStats could read in full at the source.

The next reported numbers that will move the picture

Upcoming data points
  • Nvidia, quarter ending October 2026: company guidance is $108.0 billion of total revenue, plus or minus 2 percent, up from $96.2 billion in the quarter to July 2026.
  • Broadcom, quarter ending early November 2026: guidance of $21.7 billion of AI semiconductor revenue, 236 percent more than a year earlier.
  • AMD, December 2026 quarter: guidance of about $17 billion of total revenue, about 66 percent growth at the midpoint, and chief executive Lisa Su expects growth to accelerate further.
  • TSMC, December 2026 quarter: guidance of $52.3 billion to $54.3 billion of revenue, and full year 2026 growth of around 44 percent.
  • 2027 capex plans: Alphabet expects capex to rise significantly in 2027, Meta is still developing its 2027 budget, and Microsoft expects more than $50 billion of capex in the December 2026 quarter alone.

For background on the other technology markets that feed and depend on AI, our technology and telecom statistics section collects the related pages, and the software topic overview covers the application layer.

Methodology and data notes

All revenue and spending figures come from company earnings releases, filings, investor presentations and earnings call transcripts published on the investor relations websites of the companies or with the US Securities and Exchange Commission. Run-rate figures for Anthropic and OpenAI come from announcements on their own websites. The forecast section uses the Gartner press release of 19 May 2026 and is labelled as a forecast throughout. Figures were checked against these sources on 6 November 2026.

Data notes
  • Nvidia quarters: fiscal years end in late January. Q2 FY2027 is the quarter to 26 July 2026. All ten quarters use the 2026 recast into Data Center and Edge Computing platforms.
  • Capex: cash purchases or additions to property and equipment. Amazon figures are gross of proceeds from sales and incentives. Microsoft and Meta also report totals including finance leases, which are higher.
  • TSMC HPC: the dollar value is a BusinessStats calculation from a rounded share and includes non-AI chips.
  • Run rates: company definitions, not audited annual revenue. The OpenAI annualized value of $24 billion is calculated from its statement of $2 billion per month.
  • Calculations: sums, shares, growth rates and four-quarter totals are BusinessStats calculations from reported figures and are labelled where used.
  • Not used: market size estimates from consultancies and data aggregators that do not publish their method, and figures that could not be confirmed at a primary source.

The previous edition of this page reported a blended consensus market size of $390 billion for 2025 and $1.81 trillion for 2031, an unsourced 37 percent growth rate, regional shares and a Nvidia data center figure for fiscal 2025 given as both over $80 billion and over $90 billion. None of these could be traced to a primary source or matched the company filings. They have been replaced with the reported figures above.

The reported AI market in eight numbers

$89.0B
Nvidia data center revenue
Quarter to July 2026, 92.5 percent of company revenue.
$193.7B
Nvidia data center, fiscal 2026
Up from $115.2 billion in fiscal 2025.
$16.7B
Broadcom AI semiconductors
Quarter to 2 August 2026, up 221 percent.
68%
HPC share of TSMC revenue
Q3 2026, up from 57 percent a year earlier.
$202.0B
Big Tech capex
Four companies, September 2026 quarter.
$561B
Google Cloud backlog
End of September 2026, from $155 billion.
$47B+
Anthropic run rate
May 2026, from about $9 billion at end of 2025.
$2.60T
Gartner 2026 forecast
Worldwide AI spending, published May 2026.

Questions about the size of the AI market

It depends on what is counted. Gartner forecast in May 2026 that worldwide AI spending will reach $2.60 trillion in 2026, including hardware, software and services. Measured by reported money, Nvidia alone booked $89.0 billion of data center revenue in the quarter to July 2026, and Amazon, Alphabet, Microsoft and Meta spent $202.0 billion on capex in the September 2026 quarter.

Nvidia reported data center revenue of $89.0 billion in the quarter to 26 July 2026, up 117 percent from a year earlier. Data center revenue was $193.7 billion in fiscal 2026, which ended in January 2026, and $115.2 billion in fiscal 2025.

Amazon, Alphabet, Microsoft and Meta spent $202.0 billion in cash on property and equipment in the quarter ended September 2026, more than double the $97.3 billion of a year earlier. Their combined guidance for calendar 2026 is about $730 billion at the midpoints of their ranges.

Anthropic said its run-rate revenue crossed $47 billion in May 2026. It had been about $9 billion at the end of 2025, $14 billion in February 2026 and over $30 billion in April 2026, according to company announcements.

OpenAI said in March 2026 that it was generating $2 billion of revenue per month, equal to about $24 billion a year. Its chief financial officer wrote in January 2026 that annualized revenue was more than $20 billion in 2025, up from $6 billion in 2024.

Microsoft said its AI business surpassed an annual revenue run rate of $37 billion in the quarter to March 2026, growing 123 percent year over year. Azure revenue grew 48 percent in the quarter to September 2026.

Broadcom reported $16.7 billion of AI semiconductor revenue in the quarter to 2 August 2026. AMD reported $8.5 billion of data center revenue in the quarter to 26 September 2026. TSMC, which manufactures most AI chips, earned 68 percent of its $46.72 billion third quarter 2026 revenue from high performance computing.

Estimates use different definitions. Some count only AI software, others add chips, servers, consulting and AI-capable phones and PCs, and many are not based on published methods. Company filings are narrower but can be checked, and layers of the supply chain should not be added together because the same dollar passes through several companies.

Sources

NVIDIA: NVIDIA Announces Financial Results for First Quarter Fiscal 2027, 20 May 2026 (data center, compute and networking, new market platforms).

NVIDIA, Form 8-K Exhibit 99.1: results for the second quarter of fiscal 2027, 26 August 2026 (revenue, data center, Q3 outlook).

NVIDIA Investor Relations: Revenue by Market Platform, Q1 FY2025 to Q2 FY2027 (Data Center, Hyperscale, ACIE, Edge Computing).

TSMC: TSMC Reports Third Quarter EPS of NT$29.43, 15 October 2026, with the 3Q26 management report (revenue by platform) and the 3Q26 earnings call transcript.

Broadcom, Form 8-K Exhibit 99: third quarter fiscal 2026 results, 2 September 2026, with the second quarter (3 June 2026) and first quarter (4 March 2026) releases.

AMD: AMD Reports Third Quarter 2026 Financial Results, 3 November 2026.

Amazon: third quarter 2026 earnings release (purchases of property and equipment, AWS, 2026 capex expectation).

Alphabet: 2026 Q3 earnings call, 28 October 2026 (capex, guidance, Google Cloud, backlog, Gemini), and the Q3 2025 Form 8-K (prior year capex and backlog).

Microsoft Investor Relations: fiscal 2027 first quarter results and the earnings call, 28 October 2026, and the fiscal 2026 third quarter call (AI business run rate).

Meta Platforms, Form 8-K Exhibit 99.1: third quarter 2026 results (capital expenditures, 2026 outlook).

Anthropic: Series H announcement, 28 May 2026, with Google and Broadcom compute deal (6 April 2026), Series G (12 February 2026) and Series F (2 September 2025).

OpenAI: OpenAI raises $122 billion to accelerate the next phase of AI, 31 March 2026, and A business that scales with the value of intelligence by Sarah Friar, 18 January 2026.

Gartner: Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026, 19 May 2026, and January 2026 edition.

Revenue, capex and backlog figures are as reported by each company for the period stated. Run rates are company statements, not audited revenue. Shares, sums, growth rates, four-quarter totals, the TSMC HPC dollar value and the OpenAI annualized value are BusinessStats calculations from reported figures. The Gartner figures are forecasts and are labelled as such. The supply chain layers on this page overlap and must not be added together.
Newsletter

Get new statistics in your inbox

One short email when we publish a new report: the key numbers, the chart and the sources. Free, no spam, unsubscribe in one click.

We never share your email. You can unsubscribe at any time.