AI glossary · Basic · Data engineering · AI platforms & solutions
Data warehouse
A data warehouse (DW) is a store that gathers data from across a company in one place, organized in a form that is good for analysis. Unlike databases that handle day-to-day operations, it is designed for aggregating and querying large amounts of data at once.
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Data, ops & platforms
- Added
- 2026-10-08
How it works
A data pipeline regularly pulls data from multiple systems such as orders, members and inventory and loads it into the warehouse. Because the data is organized into a predefined table structure (schema) on the way in, users can pull the numbers they want right away with a query language called SQL. Internally, many warehouses store data by column, so they quickly handle questions that sum a few fields across hundreds of millions of rows. These days they are often used as cloud services such as Snowflake, Google BigQuery and Amazon Redshift.
How it differs from an operational database
An online store's order database is built to quickly handle small, frequent operations such as 'save one order' or 'change delivery status'. Throwing large aggregation queries at it is not only slow but can also slow down the live service. A data warehouse provides a separate space just for analysis, so you can accumulate and analyze historical data without burdening the service.
Operational DB vs. data warehouse
Operational databaseWhere the service runs
- Small, frequent reads and writes
- Saving one order, changing status
- Mostly current data
Data warehouseWhere analysis happens
- Optimized for large aggregation queries
- Gathers data from multiple systems
- Keeps historical data too
공통 Both store data as tables and are handled with SQL
Why it matters
Most BI (business intelligence) tools, such as management reports and sales dashboards, run on top of a warehouse. It also serves as a single point of reference, 'the official numbers come from here', to reduce the problem of each department getting different numbers. Recently, many teams have also been trying approaches where you ask an LLM a question in natural language and it writes SQL to find the answer in the warehouse.
Things to know
Because data must fit a fixed table structure, it is not well suited to unstructured data such as images, audio and long documents. Changing the structure requires fixing pipelines and reports as well, so it is somewhat slow to adapt to change. To make up for these limits, data lakes or data lakehouses are used alongside it.
Example
A retail company gathers data from its store POS systems, online store and logistics system into a cloud data warehouse every night. The next morning, executives check sales by branch and product on a BI dashboard, and analysts use SQL to pull 'products that sold well on rainy days'. It can also be connected so that when you ask the in-house AI assistant 'Tell me last month's sales in the Seoul region', the AI queries the warehouse and brings back the answer.