# Data Center Market Sizing and Forecasting

> TAM, SAM and SOM for data center markets, built bottom-up and top-down and reconciled, with segment, country and scenario forecasts to 2035.

Publisher: Data Center Market Research & Consulting | DC Market Insights  
Published: 2026-10-05  
Last updated: 2026-10-05  
Page: https://www.dcmarketinsights.com/consulting/market-sizing

**Quick answer:**  Data center market sizing and forecasting gives a product, service or asset a defensible number: how big its market is today, how fast it will grow to 2035, and which segments, regions and countries will lead. DC Market Insights builds every figure bottom-up and top-down and reconciles the two, for vendors, investors, operators, IPO candidates and their advisers who need a number that survives a board, an investor or a regulator.

A market number is only as good as the range around it. The [International Energy Agency](https://www.iea.org/reports/energy-and-ai/executive-summary) expects data center electricity use to more than double, from 415 TWh in 2024 to around 945 TWh by 2030, and notes that global investment in data centers reached around half a trillion dollars in 2024. Yet the same agency's [2035 cases](https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai) run from around 700 TWh to around 1,700 TWh. A plan that rests on one point inside that range, without saying which, is a plan built on luck.

National forecasts show the same spread. The [US Department of Energy](https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers), reporting Lawrence Berkeley National Laboratory's study, puts US data center use at 176 TWh in 2023, about 4.4% of US electricity, rising to between 325 and 580 TWh by 2028. The top of that range is almost 80% above the bottom. Spend is moving faster still: global data center investment nearly doubled between 2022 and 2024, according to the same IEA report.

DC Market Insights is the data center practice of Credence Research, a research and consulting firm founded in 2015 with 200+ analysts and consultants and 450+ consulting projects a year. Our sizing work uses the same method and models as our 519 published data center reports.

## What is data center market sizing and forecasting?

It is the work of putting a value, in US dollars and in physical units such as megawatts, racks or systems, on a defined data center market, and projecting that value year by year to 2035. A good sizing answers four questions at once: how big, how fast, where, and how sure.

Clients usually come to us because their internal number does not match the figure an investor quotes, because their market is new and nobody has defined it cleanly, or because a board paper or offer document needs a market chapter that is traceable line by line. We solve all three the same way: define the market, build it two ways, reconcile, split and show the cases.

### TAM, SAM and SOM: which number do you actually need?

The total addressable market (TAM) is all the spend on a product or service, everywhere. The serviceable addressable market (SAM) is the part your product and channels can actually reach. The serviceable obtainable market (SOM) is the share you can realistically win in the plan period, given competitors, capacity and sales reach.

| Measure | What it answers | Typical data center example | Who relies on it |
| --- | --- | --- | --- |
| TAM | How big is the whole market? | Global spend on data center thermal management | Investors, boards, IPO market chapters |
| SAM | How much can we reach? | Spend on the cooling types we make, in the regions we sell | Strategy and product teams |
| SOM | How much can we win? | Our realistic share of SAM, account by account | Sales planning, budgets, lenders |

Confusing these three is the most common error we see in a client's own numbers. A deck that compares its revenue plan with a global TAM looks modest; the same plan set against a narrow SAM may need a share no competitor has ever held. We build all three and show the step from one to the next, so the reader can see exactly which products, regions and customers were excluded and why.

### How do you define a data center market cleanly?

We start every project with a written definition: what is in, what is out, where the boundary sits between two markets, and which unit we count. In data centers this matters more than in most sectors, because the same product can be counted twice. A cooling unit can sit inside a modular data center, a colocation operator's capex and a hyperscaler's server budget at the same time.

Our five-layer structure (DC Core, Upstream, Downstream, Ecosystem and Extended, described below) exists to stop that double counting. Each market sits in one layer, and its boundaries with the markets next to it are stated before any number is built.

## How do bottom-up and top-down sizing work, and why reconcile them?

Bottom-up sizing counts the market from the supply side, and top-down sizing works down from a larger total; a figure is only used once both builds agree, or once we can explain exactly why they do not. Averaging two numbers that disagree hides an error; reconciling them finds it.

### The bottom-up build

A bottom-up model counts units and prices, or adds up the relevant revenue of every company in the market. For a colocation market that means operator revenue by metro; for an equipment market it means units shipped, installed base and average selling price by product type; for a services market it means contracts and their value.

Company filings anchor the largest players. Equinix reported revenue of USD 8.748 billion for 2024 according to [Data Center Dynamics](https://datacenterdynamics.com/en/news/q4-2024-data-center-colocation-results-digital-realty-equinix-and-iron-mountain), and Digital Realty USD 5.55 billion according to [Zacks](https://www.zacks.com/stock/news/2415936/digital-realtys-q4-core-ffo-tops-estimates-revenues-grow-yy). Those two figures alone set a floor for any global colocation estimate. The smaller and private players are then built from capacity, pricing and interviews, market by market.

### The top-down build

A top-down model starts from a total that is measured independently, such as electricity use, construction spend, IT spend or total data center capex, and applies the share that belongs to the market being sized. For data center power and cooling, electricity use is a natural anchor, because every watt that reaches a server also has to be delivered, backed up and cooled.

The IEA's figure of around half a trillion dollars of data center investment in 2024 is the kind of ceiling a top-down build uses. No single equipment market inside it can be larger than its share of that spend allows.

### Reconciliation in practice

Reconciliation is where most of the value sits. When the two builds differ, we trace the gap to its cause: a company counted in the wrong layer, a price that applies to one region but was used for all, a share assumption that no longer holds, or a segment that one build covered and the other missed.

| Check | Bottom-up side | Top-down side | What a gap usually means |
| --- | --- | --- | --- |
| Value | Company revenue, units x price | Share of total capex or spend | A segment double counted or missed |
| Volume | Installed MW, racks or units | Electricity use, capacity built | Wrong utilisation or density assumption |
| Price | Contract and list prices | Spend divided by volume | Regional price applied globally |
| Growth | Company guidance and pipelines | Driver forecasts such as power demand | Optimism in guidance, or a lagging driver |

Capex totals are a good example. Alphabet alone reported capital expenditure of USD 91.4 billion in 2025 and guided to USD 175 to 185 billion for 2026, according to its [fourth-quarter 2025 results](https://www.sec.gov/Archives/edgar/data/1652044/000165204426000012/googexhibit991q42025.htm). A bottom-up build that adds up company capex has to separate data center spend from other spend, and timing of cash from timing of installation, before the sum can be compared with a top-down total. Skipping that step is how two careful teams arrive at very different market sizes.

## How do we split a market by segment, region and country?

We split every market along the lines the client actually sells, invests or reports, and every split must add back to the total. A model that is right at the global level but wrong in the client's three main countries is no use to a sales director.

### Segment splits

Segments follow the product or service, not the marketing language. For an equipment market that might be technology type, capacity rating, end user (hyperscale, colocation, enterprise, edge) and new build versus retrofit. For an operator market it might be wholesale versus retail colocation, or hyperscale versus enterprise demand. We agree the segment tree at the scoping call so the model matches your own reporting lines.

### Region and country splits

Geography is where data center markets are least even. The IEA reports that the United States accounted for 45% of global data center electricity use in 2024 and China for 25%. Cloudscene data published by Statista in March 2024 counted 5,381 data centers in the United States, against 521 in Germany and 514 in the United Kingdom ([Economy Middle East](https://economymiddleeast.com/10-countries-with-the-most-data-centers)). A global growth rate applied evenly across countries would misstate almost every one of them.

Our models come in global, regional and country editions, and city-level splits are added where a client's decision needs them. Our [UK Data Center Thermal Management Market](https://www.dcmarketinsights.com/report/uk-data-center-thermal-management-market) report, for example, shows London and Slough holding more than 65% of UK spend. A vendor planning a UK sales team needs that concentration, not a national average.

## How do scenario forecasts work, and how many cases do you need?

Every forecast we deliver carries at least three cases, base, upside and downside, and each case is driven by named assumptions rather than a percentage added to or taken from the base. The width between the cases is itself a finding: it tells the reader how much the decision depends on things nobody can yet know.

Public forecasts show why one line is never enough. The IEA's three 2035 cases for data center electricity run from around 700 TWh in its Headwinds case, through around 970 TWh in its High Efficiency case, to around 1,700 TWh in its Lift-Off case. The agency also expects electricity use in accelerated servers, driven mainly by AI, to grow by 30% a year in its Base Case.

| Driver we vary | Upside case | Downside case |
| --- | --- | --- |
| AI demand | Faster adoption of training and inference | Slower adoption or efficiency gains that cut compute need |
| Power availability | Grid connections arrive on time | Connection delays push capacity later |
| Equipment supply | Lead times shorten | Shortages of transformers, chillers or switchgear |
| Price | Tight vacancy supports pricing | New supply or competition pushes prices down |
| Policy | Incentives and fast permitting | Moratoria, efficiency or water rules |

Power is often the driver that moves a case most. The IEA estimates that around 20% of planned data center projects could be delayed by grid constraints, and that wait times for transformers and cables have doubled in the past three years. A forecast that assumes every announced megawatt arrives on schedule is an upside case, whether or not it is labelled as one.

We also show the history. Every model carries actual market size from 2020, so a reader can compare the forecast's growth with what the market has really done, and see where the curve bends and why.

## What are market sizing studies used for?

They are used wherever a decision or a document needs a market number that someone else will check. The four most common uses are board planning, investor materials, offer documents and M&A, and each puts different demands on the work.

### Board and strategy planning

Boards use a market model to set targets, choose where to invest and test whether a plan asks for more share than is realistic. For this use the SAM and SOM matter most, along with segment and country splits that line up with the business units. We build the model in Excel with open formulas so the strategy team can update it between reviews.

### Investor decks and fundraising

A company raising capital needs a market number that an investor's own analysts will accept. That means a clearly stated definition, a source trail behind every input and growth rates that match their endpoints. A TAM that cannot be traced is discounted, however large it is. Our [investor services](https://www.dcmarketinsights.com/consulting/investors/) page covers how funds use the same work from the other side of the table.

### IPO prospectus market chapters

Offer documents need an independent industry view that regulators, underwriters and investors will read closely. Our team prepared the India Beer and Malt Spirit Equipment Market 2018–2032 study, which sized the market at USD 204.08 million in 2024, rising to USD 302.28 million by 2032, and was cited in SpectraA Technology Solutions Limited's draft red herring prospectus filed on 21 March 2026 ([case study](https://www.credenceresearch.com/info/casestudy/spectraa-technology-solutions-limited-a-case-study-in-ipo-ready-market-validation)). The same standard applies when a data center operator, developer or equipment maker prepares to list: every number must be sourced, consistent and defensible under questioning.

### M&A and due diligence

Buyers and lenders use market sizing to test whether a target's growth plan fits the market it sells into. Data center M&A reached a record USD 73 billion in 2024, according to [Synergy Research Group](https://srgresearch.com/articles/its-official-data-center-ma-deals-broke-all-records-in-2024), and when valuations move that fast a wrong market assumption becomes expensive quickly. For transactions, sizing usually sits inside a wider [commercial due diligence](https://www.dcmarketinsights.com/consulting/due-diligence/) review that also tests power, pricing and tenants.

## Should you buy a published report or commission a custom model?

Buy a published report when your question is the one most buyers of that market ask; commission a custom model when your definition, segments, geography or cases differ from the standard. Many projects sit in between: they start from a published model and tailor it.

| Factor | Published report | Custom model |
| --- | --- | --- |
| Market definition | Standard, as most buyers define it | Your products, channels and boundaries |
| Segments and geography | Fixed set of splits | Your business units, regions and cities |
| Cases | Base forecast with drivers | Base, upside and downside on your drivers |
| Your own data | Not included | Combined with our model, kept confidential |
| Format | Report and data tables | Report, open Excel model, findings session |
| Best for | Fast context, benchmarking | Board papers, offer documents, deals |

## How does a market sizing engagement work?

Every engagement runs in five steps, and each step produces something you can use before the next one starts.

1. **Define the market and the decision.**  A scoping call to agree the decision the number will support, the market definition, the segment tree, the geography, the cases and the deadline. You receive a written scope and a quote.
2. **Start from the baseline.**  Our analysts pull the relevant DC Market Insights models and list what must change to fit your definition, and what new data the project needs.
3. **Build both ways.**  We build the bottom-up and top-down models in parallel, using filings, public datasets, trade data, pricing and your own data where you share it, and we test key assumptions in interviews with operators, suppliers, buyers and other people who know the market.
4. **Reconcile and forecast.**  We reconcile the two builds, explain every material gap, apply the segment, region and country splits, and run the scenario cases to 2035.
5. **Present and stand behind it.**  A findings session with your team, and with your board, investors or advisers if needed. Follow-up questions are answered after delivery.

### What you receive

| Deliverable | Format | Used for |
| --- | --- | --- |
| Market definition note | Short PDF | Agreeing scope before the build |
| Market sizing report | PDF with charts and tables | Board papers, investor decks, offer documents |
| Market model | Excel with open formulas | Your own planning and updates |
| Assumptions and source log | Excel | Audit trail for advisers and reviewers |
| Findings session | Video call or on site | Questions from your team or board |

## DC Exclusive: how does our 519-model library speed up a market sizing project?

It means most projects start from a tested model rather than a blank sheet: 519 published data center market models, each with history from 2020, a 2025 base year and a forecast to 2035, split by segment, region and country. A consulting team that starts there can spend its time on your definition and your cases.

The library is organised in five layers, which also act as our boundary rules against double counting:

| Layer | What it covers | Typical sizing question |
| --- | --- | --- |
| DC Core | Colocation, hyperscale, edge and modular facilities | How much capacity, and at what price, by metro? |
| Upstream | Power, cooling, construction and equipment | How big is the market for our product, and who buys it? |
| Downstream | Cloud, AI compute, interconnection and hosting | How fast is the demand that fills data centers growing? |
| Ecosystem | DCIM, security, maintenance and finance | How large is the service and software spend per MW? |
| Extended | Adjacent markets in their data center form | How much of a wider market is driven by data centers? |

Two published examples show the range. Our [OCP Rack Market](https://www.dcmarketinsights.com/report/ocp-rack-market) report values that market at USD 1.63 billion in 2025, rising to USD 13.34 billion by 2035, a compound annual growth rate of 23.4%. Our [UK Data Center Thermal Management Market](https://www.dcmarketinsights.com/report/uk-data-center-thermal-management-market) report puts UK spend at USD 993.9 million in 2025, rising to USD 3,398.9 million by 2035.

Three proprietary datasets are in development and will feed future sizing work: a Capacity Tracker (live and pipeline MW by country, city and operator), a Colocation Price Index (USD per kW per month by market) and a Deal Tracker (M&A, funding, land and power deals).

## How do we make sure a market number holds up?

We check every number against its own endpoints, against its neighbours and against the outside world before it leaves the firm. The checks are simple, and we run every one of them on every project.

- **Growth matches endpoints.**  Every stated growth rate is recalculated from the start and end values. For example, the IEA's move from 415 TWh to around 945 TWh between 2024 and 2030 implies growth of roughly 15% a year, and a model that quotes those endpoints with a different rate is wrong somewhere.
- **Parts add to the whole.**  Segments, regions and countries must sum to the total in every year, in value and in volume.
- **Neighbours agree.**  A cooling market cannot grow much faster than the capacity it cools for long without a reason, such as a shift to liquid cooling, that is stated and sourced.
- **Outside anchors.**  Company filings, government and regulator data, and trusted public datasets are checked against the model, and the source of each is logged.
- **Independent review.**  An editor who did not build the model checks figures, growth rates and sources before delivery.

## Who leads our market sizing work?

Market sizing engagements are led by senior people who build and review models every day, with our wider bench of 200+ analysts behind them.

- **Priyanka Mor, Senior Consultant.**  Priyanka leads TAM, SAM and SOM work, market entry feasibility, cross-border commercial due diligence and M&A screening, and is our principal methodology reviewer. She has more than 15 years in market research and consulting.
- **Amit Jain, Senior Consultant, ICT & Emerging Technologies.**  Amit heads our ICT consulting and syndicated research practice and has more than 20 years in technology market research. His work covers data center infrastructure, cloud and telecom, including market sizing and vendor benchmarking.
- **Deepti Agrawal, Senior Editor, Research.**  Deepti is the final quality gate for our reports and consulting work. She checks models, growth rates and company profiles, and owns our house style.

## Which DC Market Insights research supports market sizing?

Every sizing project links back to the published reports for the markets involved, so you can see our baseline before we tailor it. Useful starting points:

- [OCP Rack Market](https://www.dcmarketinsights.com/report/ocp-rack-market) for hyperscale rack and equipment demand
- [UK Data Center Thermal Management Market](https://www.dcmarketinsights.com/report/uk-data-center-thermal-management-market) for cooling spend and city concentration
- [Iberia Data Center Market](https://www.dcmarketinsights.com/report/lberia-data-center-market), valued at USD 11.53 billion in 2025, for demand in Spain and Portugal
- The full [data center research library](https://www.dcmarketinsights.com/reports/), organised in five layers

Related services: [competitive intelligence](https://www.dcmarketinsights.com/consulting/competitive-intelligence/) for share and competitor benchmarking, [market entry and power strategy](https://www.dcmarketinsights.com/consulting/market-entry-power/) for choosing the next country, and our work for [equipment vendors](https://www.dcmarketinsights.com/consulting/vendors/). See all services on the [consulting hub](https://www.dcmarketinsights.com/consulting/).

## Frequently asked questions

**What does a data center market sizing project cost?**

Fees are quoted per engagement and depend on the number of markets, segments and countries, whether a published model already covers the market, and the deadline; you receive a written quote after the scoping call.

**What is the difference between TAM, SAM and SOM?**

TAM is total spend on a product or service everywhere. SAM is the part your products and channels can reach. SOM is the share you can realistically win in the plan period. We build all three and show the step between each, so a reader can see what was excluded and why.

**Why do different sources give different sizes for the same market?**

Usually because they define the market differently, count different segments or regions, or use a different base year. Public energy forecasts show how wide honest ranges can be: the US Department of Energy puts US data center use in 2028 at anywhere from 325 to 580 TWh. We state our definition in writing and reconcile against other figures so you can see where any gap comes from.

**Can you work with our own sales and shipment data?**

Yes. Your data improves the model, especially for SOM and share. It is used only for your engagement, and confidentiality terms are agreed in writing before any data is shared.

**Can the study be cited in an offer document or investor deck?**

Studies are written to be read by boards, investors and regulators, with every figure traced to its source. Who may rely on the study and how it may be cited are agreed in the engagement letter before work starts.

**How do we start?**

Send us the market, the decision the number must support and your deadline through the form below. A senior consultant will reply within one business day to arrange a scoping call.

## Sources

- [International Energy Agency: Energy and AI, executive summary](https://www.iea.org/reports/energy-and-ai/executive-summary) (2025)
- [International Energy Agency: Energy and AI, energy demand from AI](https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai) (2025)
- [US Department of Energy: DOE releases new report evaluating increase in electricity demand from data centers (Lawrence Berkeley National Laboratory)](https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers) (20 December 2024)
- [Alphabet: fourth quarter and fiscal year 2025 results, SEC Form 8-K exhibit 99.1](https://www.sec.gov/Archives/edgar/data/1652044/000165204426000012/googexhibit991q42025.htm) (4 February 2026)
- [Data Center Dynamics: Q4 2024 colocation results, Equinix](https://datacenterdynamics.com/en/news/q4-2024-data-center-colocation-results-digital-realty-equinix-and-iron-mountain)
- [Zacks: Digital Realty full-year 2024 revenue](https://www.zacks.com/stock/news/2415936/digital-realtys-q4-core-ffo-tops-estimates-revenues-grow-yy)
- [Countries with the most data centers (Cloudscene via Statista, March 2024)](https://economymiddleeast.com/10-countries-with-the-most-data-centers)
- [Synergy Research Group: data center M&A deals broke all records in 2024](https://srgresearch.com/articles/its-official-data-center-ma-deals-broke-all-records-in-2024) (13 January 2025)
- [SpectraA case study: IPO-ready market validation](https://www.credenceresearch.com/info/casestudy/spectraa-technology-solutions-limited-a-case-study-in-ipo-ready-market-validation)