Are there any object-store-based analytics platforms that handle concurrent queries without blowing up costs?

Yes — but the key is the query engine and caching layer, not just the object store.

Object stores like S3, GCS, and Azure Blob are cheap for storage, but they can get expensive under heavy concurrent analytics if every query repeatedly scans lots of data. The platforms that handle concurrency well usually combine:

• columnar file formats like Parquet or Iceberg/Delta
• metadata/pruning to avoid scanning irrelevant data
• compute isolation / autoscaling
• result caching / data caching
• sometimes materialized aggregates / indexes

Good options

1. Databricks Lakehouse

• Uses object storage plus Delta Lake
• Strong concurrency handling with Photon, caching, and autoscaling
• Good if you want managed Spark + SQL + ML in one platform
• Cost control is decent if tables are well-partitioned and optimized

2. Snowflake on external tables / Iceberg

• Not purely object-store-native in the classic sense, but it’s very strong for concurrent analytics
• Handles many concurrent users well
• Costs can still rise with concurrency because compute is warehouse-based, but autoscaling and multi-cluster warehouses help

3. Trino / Starburst

• Query engine that sits on top of object storage
• Can handle high concurrency if tuned properly
• Often paired with:
  • caching
  • small result sets
  • pre-aggregations
  • workload isolation
• Starburst Enterprise adds more management and caching features than vanilla Trino

4. ClickHouse with object storage

• ClickHouse can query data in object storage and is extremely fast for analytics
• Works well for concurrent BI-style workloads
• Best when data is loaded into ClickHouse or managed with tiered storage; pure object-store querying is possible but not always ideal for every workload

5. Dremio

• Designed around object storage and query acceleration
• Uses reflections/materializations to reduce scan costs
• Good for concurrent SQL analytics over data lakes
• Often chosen specifically to avoid “scan S3 every time” problems

6. Apache Doris / StarRocks

• More of a lakehouse/MPP analytic database style
• Can integrate with object storage and support concurrency well
• Strong for BI workloads, often with cost controls via caching and compaction

What actually prevents cost blowups

If you’re evaluating a platform, look for these features:

• Predicate pushdown and column pruning
• Partitioning / clustering / sorting
• Metadata indexes or file statistics
• Result caching
• Data caching / warm storage
• Materialized views / reflections
• Workload isolation
• Concurrency scaling
• Query limits / quotas
• Automatic compaction and file sizing

Practical rule of thumb

If your workload is:

• many BI users running similar dashboard queries → Dremio, ClickHouse, Starburst, Snowflake
• mixed ETL + SQL + ML on lake data → Databricks
• wanting open-ish lakehouse with control → Trino/Starburst or Iceberg-based stack
• very high-speed analytics with some loading step acceptable → ClickHouse / StarRocks / Doris

Bottom line

Yes, there are platforms that can do this well — especially Dremio, Starburst/Trino, Databricks, Snowflake, ClickHouse, and StarRocks/Doris — but the “no cost blowup” part depends heavily on caching, data layout, and workload isolation.

If you want, I can also give you:

1. a short comparison table, or
2. a recommendation by use case (BI dashboards, ad hoc SQL, multi-tenant SaaS, log analytics, etc.).