We analysed dozens of Snowflake environments across enterprise customers to understand where warehouse spend was actually going. The biggest issue wasn't inefficient queries or runaway pipelines. It was idle compute.
On average, 34% of warehouse spend came from warehouses that were running without executing queries. Credits were still being consumed, even when nothing was happening.
Why Snowflake warehouses keep burning credits when nobody is using them
When a virtual warehouse finishes a query, it does not stop immediately. By default, Snowflake keeps it running until the auto-suspend setting kicks in. Out of the box, that is set to 600 seconds.
The default makes sense in theory. A warm warehouse responds faster to the next query. For teams running frequent, interactive workloads, that works well.
The problem is that most Snowflake environments run dozens of warehouses with very different workload patterns. An ETL job that fires once per hour does not need a 600-second warm window. A reporting warehouse that runs at 9am and again at 2pm is paying for idle warehouse costs in between. A pipeline that processes a batch and goes quiet for the rest of the day accumulates credits for nothing.
Most teams never revisit the default auto-suspend configuration after setup. Over time, that can translate into a substantial amount of idle compute spend.
What idle Snowflake warehouse costs look like in practice
At scale, a 34% idle rate can translate to over $1M per year in wasted compute credits. The waste is not concentrated in one place. It is spread across dozens of warehouses, each contributing a relatively small amount, but across the environment the costs add up fast.
Idle compute is just the beginning
Auto-suspend is the most visible setting in Snowflake. Every warehouse has it. But the number you see is seconds, not dollars. Snowflake tells you the configuration. It does not tell you what it costs. Decisions made early in setup quietly compound into costs that show up somewhere else entirely. The same is true across the rest of your environment.
Storage accumulating across tables nobody has queried in months. Clustering credits going to tables that do not benefit from them. Ingestion costs baked in before a single query runs.
Knowing where the waste is gets you halfway there. Having a clear, low-risk path to fix it, with the full context to do it confidently, is the other half.
From idle credits to verified savings, in days
PointFive is the Cloud & AI Efficiency Platform that goes beyond reporting waste — it eliminates it. Across AWS, GCP, Azure, Kubernetes, and data platforms like Snowflake, Databricks, and BigQuery, PointFive surfaces the architectural inefficiencies traditional tools miss, then routes each finding to the engineer who can fix it, with the context they need to act.
Connect Snowflake and that same engine goes to work on your warehouse environment. Idle compute, oversized clusters, auto-clustering on tables that see no benefit, storage on data nobody reads, ingestion overhead set on day one and never revisited. Every inefficiency comes with the query patterns and context an engineer needs to fix it confidently, without disrupting production workloads.
Findings land in Jira, Slack, or your tool of choice. Savings are verified against actual billing. Most teams see their first results within days of connecting.
See where your Snowflake credits are actually going
If you run Snowflake at scale, the fastest way to find out where your credits are actually going is to look. Connect PointFive and we will surface idle warehouse costs, storage waste, auto-clustering inefficiency, and ingestion spend automatically — each one turned into a ready-to-act ticket, pre-loaded with the query patterns, cost impact, and recommended fix.
Talk to us to see the full analysis on your environment.