Serverless is often the preferred option for Power BI DirectQuery workloads because it starts in seconds and scales automatically. However, it’s not always the only option, and there are several ways to reduce costs.

A few high-impact optimizations:

• Set an aggressive auto-stop time (e.g., 5–10 minutes).
• Right-size the warehouse and avoid over-provisioning.
• Optimize queries and reduce unnecessary scans.
• Use aggregated/gold tables instead of querying large raw datasets.
• Run OPTIMIZE and clustering on frequently queried tables.
• Monitor usage through system.billing.usage and set budget alerts.

If your “real-time” requirement is actually a few minutes rather than seconds, consider:

• Power BI Import Mode with incremental refresh.
• Hybrid/Composite models (historical data imported, recent data via DirectQuery).
• Power BI Aggregations.

In many cases, moving part of the workload to Import/Hybrid mode significantly reduces warehouse costs while still providing near real-time reporting.

Could you share:

1. Your freshness requirement (seconds, minutes, hourly)?
2. Approximate concurrent users?
3. Whether you’re using DirectQuery, Import, or a Composite model?

That would help narrow down the most cost-effective approach.