Our research methodology is the set of rules we follow to turn evidence into a published data center market figure: define the market, size it bottom-up and top-down, reconcile the two, forecast to 2035 around stated drivers and constraints, and have every number checked twice before release. It matters to a buyer because it tells you exactly what a figure means, where it came from and how far you can rely on it.
Written by Priyanka Mor, Senior Consultant and principal methodology reviewer. Reviewed by Deepti Agrawal, Senior Editor, Research.
Anyone who buys data center research is buying a set of judgements: which products count, which year the number describes, which currency applies. Two reports can describe the same market and disagree by a wide margin simply because they answer different questions without saying so.
This page sets out how we answer those questions, from definition and sizing to sources, checks and ownership. If a figure in one of our reports does not make sense to you, this page should explain why it was built that way. If it does not, ask us.
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.
How do we define a market before we size it?
We write a definition before we build a single number: what is in, what is out, which unit we count, which base year we describe and which currency we report in. Every figure we publish carries a stated base year, currency and definition, so a reader never has to guess what a number covers.
Our library of 519 published reports uses a 2025 base year, history from 2020 and a forecast to 2035. Market values are stated in US dollars. Each report gives market size, compound annual growth rate (CAGR), regional shares, segment splits, company profiles and recent developments.
Why a five-layer structure?
Data center spending is easy to count twice. A cooling unit can sit inside a modular data center, in a colocation operator’s capital budget and in a hyperscaler’s facility plan at the same time. To prevent that, every market we cover sits in one of five layers, and its boundary with the markets next to it is written down before sizing starts.
| Layer | What is in | What is out |
|---|---|---|
| DC Core | Data center markets by country and region, colocation, hyperscale, edge | The equipment and software sold into those facilities, which sit in Upstream and Downstream |
| Upstream | Power, cooling, racks, construction and equipment | Operator revenue from leasing the finished capacity |
| Downstream | Cloud, AI infrastructure, storage, networking, software | The physical building and plant that host them |
| Ecosystem | Telecom, interconnection, services | Core facility and equipment spend counted in other layers |
| Extended | Adjacent markets that depend on or feed data centers | Spend already counted in the four layers above |
How do we build a market size?
We build every market size twice, once from the bottom up and once from the top down, and only publish a figure once the two agree or once we can explain the gap between them. Averaging two numbers that disagree would hide an error; reconciling them is how we find it.
The bottom-up build
A bottom-up build counts the market from the supply side. Depending on the market, we work from installed capacity, unit shipments, price points, and the relevant revenue of operators and vendors. For an equipment market, that usually means units shipped multiplied by an average selling price, checked against what leading vendors report.
The top-down build
A top-down build starts from a larger, independently published total and works down to the market in question. Our inputs here are national statistics, trade data, company filings, and grid and regulator data. It is less detailed but harder to inflate, because it is anchored to totals someone else has measured.
Reconciliation
Once both builds exist, we compare them. The difference between the two estimates is examined and explained before a number is published. Typical causes include a definition mismatch, a missing segment, an outdated price or trade data that mixes data center equipment with other uses. We fix the cause and record what we changed.
A worked illustration: UK data center thermal management
Our UK Data Center Thermal Management Market report puts the market at USD 993.9 million in 2025, rising to USD 3,398.9 million by 2035. The table below shows the kinds of inputs that go into each side of a figure like that. It shows structure, not private model values.
| Build | Inputs used | What it tests |
|---|---|---|
| Bottom-up | Installed and planned data center capacity in the UK; cooling units shipped by type; average price points by cooling technology; vendor revenue for UK thermal products | Whether the count of real equipment and real prices supports the total |
| Top-down | National statistics; UK trade data for cooling equipment; company filings; grid and regulator data on data center connections | Whether the total fits inside what is measured at national level |
| Reconciliation | Comparison of the two builds by segment and year | Whether any gap comes from definitions, missing segments or outdated prices |
The same approach applies to fast-moving equipment markets. Our OCP Rack Market report puts that market at USD 1.63 billion in 2025, reaching USD 13.34 billion by 2035, a CAGR of 23.4%. A growth rate that high needs both builds behind it.
How do we forecast to 2035?
We forecast by modelling the drivers and constraints that move each market, testing them in scenarios, and publishing the base case. The published figure is always the base case, and the report explains what would push the market above or below it.
| Factor | Why it matters | How it enters the forecast |
|---|---|---|
| Power availability | A building cannot open, or fill, without power | Limits how fast capacity, and the equipment inside it, can grow |
| Grid connection lead times | Connection queues can delay projects by years | Shifts capacity, and spend, later in the forecast |
| Capital expenditure | Operator and hyperscaler budgets set the pace of building | Drives demand for construction, power, cooling and racks |
| AI demand | AI workloads change rack densities and cooling needs | Raises demand for high-density power, liquid cooling and AI infrastructure |
| Regulation | Energy, water, planning and data rules shape where and how sites are built | Can open or close markets, or change the technology mix |
What we do not do
- We do not publish a single number without a stated base year, currency and definition.
- We do not extend a past growth rate forward without testing it against the constraints above.
- We do not present an upside scenario as the headline figure.
What primary research goes into a report?
Primary research means interviews with operators, vendors, investors and advisers, and we use them to validate assumptions, pricing and market share. Interviews do not replace the model; they test it.
Who we interview, and why
- Operators, because they know what they are building, what it costs and what tenants pay.
- Vendors, because they know what they ship, at what price and to whom.
- Investors, because they see deal pipelines and the assumptions behind valuations.
- Advisers, because they see many projects across many markets and can tell us what is typical.
How interviews are used
We go into an interview with specific questions drawn from the model: a price point, a share estimate, a timing assumption. If what we hear contradicts the model, we look for the reason first: one voice is a data point, not a verdict. Where several independent sources agree, the model moves. Interviews are held in confidence, which is why we do not attribute views to named individuals or companies.
Which secondary sources do we rely on?
We rely on published, traceable sources, and we rank them by how close they are to the original measurement. A company’s own audited results outrank a commentary on those results; a grid operator’s connection data outranks a news report about it.
Source types, in rough order of weight
- Company annual reports and results releases.
- Regulatory filings.
- Government and grid operator publications.
- Trade data.
- Industry association data.
- Published studies from authoritative bodies.
What we never use
- Figures we cannot trace to an original source.
- Another research publisher’s market size presented as our own.
- Anonymous or unattributed numbers from forums, social media or marketing material.
The analytical tools behind this work are ones our wider firm uses on every project: quantitative forecasting, qualitative insight, PESTLE analysis, Porter’s Five Forces, expert interviews and data modelling.
How is a report checked before it is published?
Every report and every consulting deliverable is reviewed by a second analyst and then by our editor before it is released. No figure leaves our team on one person’s say-so.
- Author build. The lead analyst defines the market, builds both sides of the size and the forecast, and writes the report.
- Second analyst review. A second analyst checks the definition, the inputs, the reconciliation and the logic of the forecast.
- Editorial review. Deepti Agrawal, Senior Editor, Research, checks models, growth rates and company profiles against sources, and makes sure the report follows our house style.
- Release. The report is published with its base year, currency, definition and “last updated” date.
Our full rules on accuracy, corrections and independence are set out in our editorial standards.
How do we keep reports current?
We refresh a report when its base data changes materially, and each report shows its “last updated” date so you can see how recent it is. Examples include new company results, revised trade data or a regulatory decision.
We are also building three proprietary datasets, all currently in development: a Capacity Tracker, a Colocation Price Index and a Deal Tracker. They are not yet available, and we will not cite them in reports until they are.
How does the same method serve consulting work?
Our consulting work uses exactly the same method as our reports, applied to one client’s question rather than a whole market. The definitions, the two-way build, the reconciliation and the two-stage review are identical.
- Market sizing applies the bottom-up and top-down build to a client’s product or service, often down to TAM, SAM and SOM.
- Commercial due diligence uses the same models to test whether an asset or platform can earn the revenue a deal assumes.
- Competitive intelligence uses the same company profiles and source ranking to show where a client stands against its rivals.
Who is responsible for our methodology?
Four people own our methodology, each for a defined part of it. You can read more about them on our team page.
| Person | Role | Responsibility for the method |
|---|---|---|
| Priyanka Mor | Senior Consultant | Principal methodology reviewer; 15+ years in market research and consulting; leads cross-border commercial due diligence, market entry feasibility, TAM/SAM/SOM and M&A screening |
| Amit Jain | Senior Consultant, ICT & Emerging Technologies | Heads our ICT consulting and syndicated research practice; 20+ years in technology market research; data center infrastructure, cloud and telecom; market sizing, vendor benchmarking and commercial due diligence |
| Satyabrat Rajawat | Senior Research Analyst, Strategic Advisory | Power, cooling and rack equipment; works from trade data, filings and equipment teardowns; has contributed to 40+ syndicated market studies |
| Deepti Agrawal | Senior Editor, Research | Final quality gate; checks models, growth rates and company profiles; owns our house style |
Frequently asked questions
Can I see the model behind a report?
The report sets out the definition, base year, currency, segment splits and forecast drivers. The working model itself is not published. If you need to understand a specific figure in more depth, a senior consultant can walk you through how it was built.
Do you share interview notes?
No. Interviews are held in confidence so that operators, vendors, investors and advisers can speak freely. We use what we hear to test the model, and we report the conclusions, not the conversations.
How do you handle private companies?
Private companies do not publish full accounts, so we estimate their relevant revenue from what can be observed: capacity, shipments, price points, regulatory filings where they exist, and interviews. Each estimate is checked against the top-down total.
Why do your numbers differ from other publishers?
Most differences come from definitions: what is counted in the market, which base year is used, which currency and exchange rate apply, and whether adjacent products are included. Some come from forecast assumptions about power, grid timing, capital spending, AI demand and regulation. We state all of these in every report, so you can compare like with like. We never adjust our figures to match someone else’s.
What does “in development” mean for the datasets?
It means the Capacity Tracker, Colocation Price Index and Deal Tracker are being built and are not yet available to clients. We will not describe them as available, or rely on them in a report, until they are ready.
How do I get a figure checked?
Contact us through our consulting page or the contact page with the report and the figure in question. A senior consultant replies within one business day. If we find an error, we correct it in line with our editorial standards.