Key Requirements
- ✓Reasonable $/GPU/hour pricing
- ✓Consumption model options (3-year, 1-year, 6-month, 3-month, 1-month)
- ✓Individual charges for storage, compute nodes, network vs. bundled pricing
- ✓Expansion and extension of existing contracts
- ✓Latest GPU availability and timeline
- ✓Kernel library availability for MFU boosting (e.g., Together AI kernel collection)