Where the Trillion-Dollar AI Data Center Spend Is Going
Data center capital spending is expected to pass $1 trillion in 2026 and $3 trillion by 2030. Dell'Oro nearly doubled its 2030 outlook between January and August 2026, so every figure on this page carries its date. Here is where that money flows, and which part of it your product can compete for.[1][2]
Three ways to measure the market
Each figure on this page is tagged with what it measures. Mixing them is the most common mistake in AI infrastructure market sizing.
Operator spending
What hyperscalers, AI clouds, and operators invest in servers, networks, buildings, and equipment
Use it for
How much buying power an account has, and where it ranks
Supplier revenue
What vendors receive in a category such as optics, HBM, switches, liquid cooling, or power equipment
Use it for
The size of the market for a specific product
Cumulative investment
Modeled total investment in facilities, energy, and technology over several years
Use it for
The long-term opportunity, not a yearly market
Where the spending goes, layer by layer
These figures measure different things: yearly operator spending, supplier revenue, and multi-year investment. They show the size of each opportunity and overlap with one another, so they should never be added together.
All data center investment
Operator spending
What it covers
Servers, storage, networking, facilities, and supporting infrastructure, across clouds, AI clouds, telcos, and enterprises
What it means for you
The total customer wallet. It is not the market for any one product.
Server chips and components
Supplier revenue
Signal
$1.8 trillion by 2030[3]
What it covers
Accelerators, CPUs, memory, network cards, and storage drives
What it means for you
Compute sets the architecture and pulls demand through every layer around it.
High-bandwidth memory
Supplier revenue
Signal
$54.6 billion in 2026, up 58%[4]
What it covers
HBM stacks for GPUs and custom AI chips, not the wider DRAM or storage market
What it means for you
Memory is a major supplier market of its own, not only a GPU part.
AI back-end switches
Supplier revenue
Signal
About $1 trillion from 2026 to 2030 combined, with scale-up more than half by 2030[5]
What it covers
Scale-up and scale-out switches. This is not the same as optical module revenue.
What it means for you
A new networking cycle for switch silicon, systems, interconnect, and optics.
Datacom optical components
Supplier revenue
Signal
More than $19 billion in 2025, with over 42 million 400G and faster modules shipped[6]
What it covers
High-speed optical modules and components for data centers
What it means for you
A measurable photonics market that is growing now, with 1.6T modules ramping.
Power, cooling, and racks
Supplier revenue
What it covers
UPS, thermal management, power distribution, racks, and containment
What it means for you
A large market that sits next to photonics in the same build-out.
Liquid cooling
Supplier revenue
Signal
Close to $3 billion in 2025, approaching $7 billion by 2029[9]
What it covers
Direct-to-chip, two-phase, and immersion cooling
What it means for you
Designs are changing faster than incumbent equipment cycles, which opens the door to new suppliers.
AI data center buildings
Cumulative investment
Signal
About $800 billion through 2030 in McKinsey's 2025 model[10]
What it covers
Land, materials, construction, and site development
What it means for you
Equipment specified during facility design is sold to developers and builders, not server buyers.
Power, cooling, and connectivity investment
Cumulative investment
Signal
About $1.3 trillion through 2030 in McKinsey's 2025 model[10]
What it covers
Generation, transmission, electrical equipment, cooling, and connectivity. Overlaps the rows above.
What it means for you
Suppliers need utility, engineering firm, and developer relationships, not only hyperscaler IT contacts.
Electricity
Electricity use
Signal
About 485 TWh in 2025, roughly doubling to about 945 TWh by 2030[11]
What it covers
Electricity that data centers consume, not equipment revenue
What it means for you
Why access to power and time to energize now carry a price.
A few buyers hold most of the wallet
Five companies reported $586.3B of cash capital spending over their latest four quarters, through June 30, 2026 (Oracle: August 31, 2026), from SEC filings as of October 16, 2026. These are company-wide figures, broader than AI or data centers, and they leave out some leased capacity.[12]
See all 377 companies on the leaderboard| Company | Four quarters |
|---|---|
| Amazon | $173.0B |
| $132.4B | |
| Microsoft | $115.9B |
| Meta | $89.3B |
| Oracle | $75.7B |
What the largest buyers say about 2026
Alphabet
Guided to $195 billion to $205 billion of 2026 capital spending after spending $44.9 billion in the second quarter. About 60% of its technical infrastructure spend that quarter went to servers and 40% to data centers and networking.[13][14]
One of the clearest public views of how spending splits between IT equipment and buildings.
Meta
Expects $135 billion to $145 billion of 2026 capital spending, narrowed in October from $130 billion to $145 billion.[15]
Spending across compute, custom silicon, networking, optics, power, and facilities.
Microsoft
Spent $41 billion in its fiscal fourth quarter of 2026, including finance leases. Roughly two-thirds went to short-lived assets, mainly CPUs and GPUs.[16]
Shows how much hyperscaler spending goes straight to compute, with the rest in long-lived sites.
Amazon
Guided to about $220 billion of 2026 cash capital spending, raised from about $200 billion, citing memory costs.[17]
This is all of Amazon, not AWS AI spending alone. Size an AWS opportunity separately.
Oracle
About $75.7 billion of cash capital spending over its latest four reported quarters.[12]
A large buyer beyond the four biggest clouds.
Who matters in each layer
The ecosystem is not a list of AI companies. It is a network of companies that set requirements, supply parts, build systems, qualify products, and operate capacity. Supplier names come from the companies the Prospectory leaderboard tracks.
Compute
NVIDIA and AMD set the merchant accelerator roadmap, alongside custom chips from the hyperscalers, such as AWS Trainium.
Optics, cooling, networking, and power suppliers sell against requirements the accelerator roadmap creates, even when the chip maker never writes the purchase order.
Memory and packaging
SK hynix, Samsung, and Micron supply HBM. The wider market runs through DRAM, modules, flash, advanced packaging, substrates, and test.
A CXL or memory pooling product has to show better accelerator use, less stranded memory, or a lower cost per answer, at a latency the workload accepts.
Interconnect and photonics
Established optical and networking vendors, plus emerging optical I/O companies the leaderboard tracks, such as Lightmatter, Ayar Labs, Avicena, Point2, Eliyan, Quintessent, HyperLight, and AttoTude.
Excellent technology still fails if it approaches the operator at the wrong point in the supply chain.
Power
Bloom Energy, Delta, Fluence, ABB, Cummins, GE Vernova, Hitachi Energy, and newer generation companies such as Oklo, X-energy, and Kairos.
Time to power is now a constraint on how fast capacity can be built.
Cooling
JetCool, Accelsius, CoolIT, Motivair, Submer, Johnson Controls, Carrier, Modine, Trane, and Auras.
The win depends on fitting the next compute platform and the facility design, not only on cooling performance.
Facilities, developers, and AI clouds
CoreWeave, Crusoe, Fluidstack, Lambda, Aligned, CyrusOne, Equinix, Core Scientific, Cipher, and Galaxy.
A first customer does not have to be a hyperscaler. An AI cloud or developer can give you the reference deployment that opens hyperscaler doors.
Enablers
Cadence, Keysight, Viavi, Anritsu, Applied Materials, and Jabil, across design software, test, materials, and manufacturing.
Design tools, contract manufacturing, and test often decide whether a promising product becomes a qualified, volume product.
From hundreds of companies to the accounts worth pursuing
A photonics or cooling supplier can look at a universe of hundreds of targets. We narrow it with questions like these, then check which signals point to a real buying window.
- Which operators are increasing capital spending?
- Which switch and accelerator vendors are building the architectures your product needs?
- Which companies are hiring the engineers who would evaluate it?
- Which competitors just raised money?
- Which organizations are active in the standards that matter to you?
More from this practice
Business development and sales leaders
Winning Hyperscaler Design-Ins
The five gates between a first meeting and a design win, and who inside a hyperscaler decides at each one.
Optics and interconnect companies
Photonics Go-to-Market
Where optics create enough system value to replace copper or pluggables, and who has to approve the change.
Memory and CXL companies
AI Memory Wall
Why large models run out of GPU memory, which suppliers that creates room for, and what to prove.
Power, cooling, and facility suppliers
Power and Cooling Go-to-Market
How time to power and rack density became buying criteria, and how to sell into facility design.
Frequently Asked Questions
Dell'Oro Group expects worldwide data center capital spending to pass $1 trillion in 2026 and $3 trillion by 2030. That is what operators spend. Supplier markets inside it are smaller and measured differently, for example $54.6 billion for HBM in 2026 and more than $19 billion for datacom optical components in 2025.
No. Operator spending, supplier revenue, and multi-year investment overlap. A switch sold to a hyperscaler is counted in its capital spending and in the switch market. Use each figure for the question it answers.
No. Reported capital spending covers the whole company, including retail and logistics, and is broader than AI. Size an AWS opportunity from the programs and architectures your product serves.
References
Figures checked against these sources on October 22, 2026.
- [1] AI infrastructure buildouts and memory cost inflation drove data center capex higher in 1Q 2026 (Dell'Oro Group, June 2026). https://www.delloro.com/news/ai-infrastructure-buildouts-and-memory-cost-inflation-drove-data-center-capex-higher-in-1q-2026/
- [2] AI buildout maintains momentum as data center capex surpasses $3 trillion by 2030 (Dell'Oro Group, August 2026). https://www.delloro.com/news/ai-buildout-maintains-momentum-as-data-center-capex-surpasses-3-trillion-by-2030/
- [3] Data center IT semiconductors and components market to surpass $1.8 trillion over next five years (Dell'Oro Group, August 2026). https://www.delloro.com/news/data-center-it-semiconductors-and-components-market-to-surpass-1-8-trillion-over-next-five-years/
- [4] 2026 market outlook: focus on the HBM-led memory supercycle, citing a BofA estimate (SK hynix Newsroom, January 2026). https://news.skhynix.com/en/2026-market-outlook-focus-on-the-hbm-led-memory-supercycle/
- [5] AI back-end switch sales to approach $1 trillion over the next five years (Dell'Oro Group, July 2026). https://www.delloro.com/news/ai-back-end-switch-sales-to-approach-1-trillion-over-the-next-five-years/
- [6] Datacom optical component revenue surpasses $19B in 2025 (Cignal AI, April 2026). https://cignal.ai/2026/04/datacom-optical-component-revenue-surpasses-19b-in-2025/
- [7] Data center physical infrastructure market grows 27 percent to $14 billion in 2Q 2026 (Dell'Oro Group, October 2026). https://www.delloro.com/news/data-center-physical-infrastructure-market-grows-27-percent-y-y-to-14-billion-in-2q-2026/
- [8] Data center physical infrastructure market forecast to reach $120 billion by 2030 (Dell'Oro Group, August 2026). https://www.delloro.com/news/data-center-physical-infrastructure-market-forecast-to-reach-120-billion-by-2030/
- [9] Data center liquid cooling market to approach $7 billion by 2029 as AI deployments accelerate (Dell'Oro Group, January 2026). https://www.delloro.com/news/data-center-liquid-cooling-market-to-approach-7-billion-by-2029-as-ai-deployments-accelerate/
- [10] McKinsey estimate of $6.7 trillion in data center investment by 2030, as reported by DatacenterDynamics (April 2025). https://www.datacenterdynamics.com/en/news/ai-could-drive-67-trillion-investment-in-data-centers-maybe-claims-mckinsey
- [11] How much energy do data centers and AI use? Our World in Data, citing IEA Key Questions on Energy and AI (2026). https://ourworldindata.org/how-much-energy-do-data-centers-and-artificial-intelligence-use
- [12] AI Infrastructure Leaderboard (Prospectory, sources dated October 16, 2026). https://prospectory.ai/ai-infrastructure-leaderboard/
- [13] AI infrastructure demand pushes Alphabet's 2026 capex guidance to $205 billion (W.Media, July 2026). https://w.media/ai-infrastructure-demand-pushes-alphabets-2026-capex-guidance-to-us-205-billion/
- [14] Alphabet second quarter 2026 results, Form 8-K exhibit 99.1 (SEC filing, July 2026). https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000066/googexhibit991q22026.htm
- [15] Meta reports third quarter 2026 results (Meta Investor Relations, October 2026). https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Third-Quarter-2026-Results/default.aspx
- [16] Microsoft fiscal 2026 fourth quarter earnings call transcript (The Motley Fool, July 2026). https://www.fool.com/earnings/call-transcripts/2026/08/07/microsoft-msft-q4-2026-earnings-call-transcript/
- [17] Amazon second quarter 2026 earnings call transcript (Stock Analysis, July 2026). https://stockanalysis.com/stocks/amzn/transcripts/657334-q2-2026/
How large is your real market?
Bring one product. We size the part of the build-out it can win, by account and architecture generation.