Databricks
AI & ML · 37% complete
Stage Checklist
- ✓ Research (.md file)
- ✓ POC App
- ✓ POC UI (templates)
- ✗ Microservice
- ✗ Containerized (Dockerfile)
- ✗ Helm Chart
- ✗ CI/CD Pipeline
- ✗ Frontend Page
AI Research
Refund Rules
Databricks uses consumption-based pricing with monthly billing in arrears, so traditional refunds for unused seats don't apply. However, customers can receive credits for billing errors or service outages that prevent compute usage. Prepaid Databricks Units (DBUs) purchases may have limited refund options - typically require enterprise sales team negotiation. Unused credits generally don't roll over unless specified in enterprise agreements.
Credit Policies
Databricks provides Service Level Agreement (SLA) credits for downtime based on availability guarantees (typically 99.9% uptime). Credits are issued as DBU credits for future use. Billing adjustments may be provided for provisioning errors or documented service issues. Enterprise customers may negotiate additional credit terms for extended outages or performance issues. No automatic credits for idle/unused compute - customers are responsible for managing auto-termination.
Contractual Terms
Standard terms include monthly billing cycles with 30-day payment terms. Enterprise contracts typically have 12-month minimum commitments with auto-renewal unless 30-60 days written notice is provided. DBU pricing is generally locked for contract term, but infrastructure costs (AWS/Azure/GCP) fluctuate with cloud provider pricing. Price increases for renewals are common but may be capped in enterprise agreements. Cancellation requires written notice, and consumption is billed through the end of the notice period.
Integration Complexity
Technical Notes
Authentication via Personal Access Tokens (PAT) or Service Principal OAuth. Rate limits are generous (1000+ requests/hour) but undocumented. Key APIs: Workspace APIs for clusters/jobs, Account APIs for multi-workspace usage, System Tables (Unity Catalog) for detailed billing/usage via SQL. Multi-cloud complexity - different base URLs for AWS/Azure/GCP. System tables require Unity Catalog enabled. Two-part billing complexity: Databricks DBUs + separate cloud infrastructure costs require integration with cloud provider billing APIs for complete picture.