Key takeaways:
- Most enterprise data turn cold within months, yet it still holds insight. Amazon Redshift Spectrum lets you query that data where it already sits, in Amazon Simple Storage Service (Amazon S3), without loading it into the warehouse first.
- Querying data in place, in open formats, is the practical on ramp to an Amazon Web Services (AWS) Lakehouse and to an artificial intelligence (AI) ready data foundation.
- The advantage is not a single feature. It is one governed view of hot and cold data, at a cost the business controls.
The Real Question Is Not How Big Your Warehouse Should Get
The instinct, when data volumes climb, is to scale the data warehouse to hold more of it. That works, and it also quietly raises the bill for data that few people ever query. There is a more economical path. Not all data needs to sit in warehouse storage to be useful; A large share of it can stay in low-cost object storage and still be queried the moment a question comes up. Redshift Spectrum is one of the clearest examples of this shift, and it is a good place for a business to see where the value, and the savings, actually come from.
Treat Cold Data as an Asset, not a Storage Line Item
Data ages fast. Research from Gartner estimates that well over half of the data enterprises store, by some measures 55 percent to more than 80 percent, is never analyzed. IDC's work on the global datasphere shows unstructured data alone making up roughly 78 percent of what organizations store and growing quickly year over year. For a Head of Data or a Chief Financial Officer (CFO), that combination is the whole point: The volume keeps rising, budgets do not rise with it, and much of the value is held in data that is expensive to keep hot and too important to discard. The opportunity is to make that cold data query able without paying warehouse rates to store all of it.
Query Your Data Where It Lives
So, what is Redshift Spectrum? It is a capability of Amazon Redshift that runs SQL queries directly against data held in Amazon S3, using external tables, without loading that data into the warehouse first. You pay for the data each query scans rather than for holding it in warehouse storage.
It helps to see where it sits alongside its neighbors. Amazon Athena runs serverless SQL directly on Amazon S3 for ad hoc questions. Redshift Serverless provides elastic warehouse compute without managing a cluster. Redshift Spectrum extends an existing Redshift environment out to the lake, so a single query can join hot warehouse tables with cold S3 data. The three are not rivals; They are three ways to reach the same data, chosen by how often and how fast you need to query it.
The cost logic is straightforward. Keep the hot, frequently queried data in the warehouse, where speed matters. Leave the cold, rarely queried data in Amazon S3 and query it on demand with Spectrum. Because Spectrum bills by the volume of data scanned, storing that data in columnar formats such as Apache Parquet, and partitioning it sensibly, means each query reads far less and costs far less. The result is not a cheaper warehouse alone; It is a deliberate split between what you pay to keep fast and what you pay only when you ask.
Build the Lakehouse, Not Just the Query
Querying in place is the first step. The destination most enterprises are moving toward is a Lakehouse on AWS, where a single, governed copy of data serves both analytics and AI. AWS has moved decisively in this direction: Amazon S3 Tables now carry built-in Apache Iceberg support, and the SageMaker Lakehouse architecture lets engines such as Redshift and Athena query one copy of data across Amazon S3 and the warehouse, with zero extract, transform, and load (ETL) integrations pulling in operational data without custom pipelines. Open formats keep that data portable and AI-ready rather than locked to one engine.
This is where KPI Partners works. As an AWS partner, we help enterprises move from scattered data and rising warehouse costs to a governed, cost-controlled Lakehouse, using the Data Platform Migration Accelerator to make the path faster and lower in risk. The same foundation feeds Enterprise AI, because a single governed copy of data is exactly what analytics and AI both need. For one example of an enterprise data warehouse built on AWS Redshift, see our enterprise data warehouse on AWS Redshift and Power BI work.
Already Mid-Migration? This Approach Still Applies
Many teams are already moving to a modern platform, whether that is Snowflake, Databricks, or Microsoft Fabric. Query-in-place and open table formats do not compete with that plan; They strengthen it. Storing cold data in open Apache Iceberg formats keeps it accessible across engines, so the work you invest now travels with you rather than being redone after the migration. The principle holds regardless of the destination: Do not move data twice, and do not pay to keep data hot that no one is querying.
Keep Data Once, Query It Anywhere
The advantage of Redshift Spectrum was never really about one feature. It is about a way of thinking: One governed view of hot and cold data, queried where it lives, at a cost the business decides. That is also the foundation a modern, AI-ready enterprise runs on.
