Azure Cost Management Connector Refresh Timeouts and the $2–5M Spend Cap

Where do the dollar figures actually come from?

Three related numbers get quoted interchangeably; Microsoft documents them separately, and it's worth keeping them straight:

• The connector: ~$5M estimated maximum, not recommended above $1M. The connector documentation states it "supports up to an estimated maximum of $5 million of raw cost details," and the FinOps toolkit's report matrix marks the connector "not recommended when monitoring more than $1M in total cost." The same toolkit page notes "there are no plans to update the Cost Management connector."
• Cost Management exports: $2–5M/month depending on Power BI license. Microsoft's FinOps toolkit guidance: "We generally recommend starting with Cost Management exports, which support up to $2-5 million in monthly spend depending on your Power BI license. If you experience data refresh timeouts... use FinOps hubs."
• Above ~$2M/month: hubs with Data Explorer. Same page: "If you use more than $2 million in monthly spend, we generally recommend using FinOps hubs with Data Explorer for the best performance."

The mechanism behind all three is the same: connector and export reports run in import mode, so every refresh pulls the full (or incremental) dataset into Power BI within a bounded refresh window. Row volume scales with spend, refresh time scales with rows, and past a threshold the refresh window closes before the data arrives.

What are your options when refreshes fail?

1. Reduce scope or granularity (free, loses detail)

Shorter date ranges, fewer subscriptions per report, or summarized granularity all shrink the dataset. Legitimate triage, at the cost of exactly the detail — resource-level, tag-level, long-trend — that the report probably exists to show.

2. Configure incremental refresh (free, partial relief)

Partitioning the dataset so only recent periods refresh reduces per-run volume. Setup is fiddly (parameters, partition policies, first-load behavior) and the ceiling doesn't move — it just takes longer to hit.

3. Move to exports + FinOps hubs / Fabric (Microsoft's recommendation at scale)

This is the documented path past the connector: exports to storage, hubs for ingestion, Data Explorer or Fabric for query scale, toolkit reports on top. It works well above ~$2M/month — and it is a new reporting stack with infrastructure to size and operate. Our comparison page lays out when it's the right call.

4. Remove the import entirely with DirectQuery

MCA Continuity takes the other exit: Power BI connects to the CostManagement_Usage SQL view via DirectQuery, so there's no dataset import, no refresh window, and no spend ceiling — visuals query the database live, and the database is kept current by the application's daily collection. Per the product FAQ, "it works at any spend level on a $15/month SQL database." Column names match EA and MCA conventions simultaneously, so the swap from the connector doesn't force a report rebuild.