| Scenario | Training Demand | Inference Demand | Capital Availability | Optimal Data Center Utilization | Potential Impact on Bitcoin's Price |
|---|---|---|---|---|---|
| A | High | Moderate | Abundant | Training may occur, but returns are poor; inference + experimentation | Less constructive, miners are not training LLMs. Inference demand is intermittent during business hours. |
| B | Flat | Strong | Abundant | Inference dominates AI use; training limited | Least constructive, inference has better economics and sustainable demand could pull miners away from the network for non-intermittent time periods. Less volatility revenue streams and steadier operating cash flows reward the shift away from mining. |
| C | Flat | Strong | Constrained | Bitcoin baseload + inference overlay optimal | Neutral. Short-term disruption for long-term stability. Stabilizes mining industry long-term to account for volatile crypto revenue. |
| D | Weak | Moderate | Constrained | Mining preferred; limited inference | Constructive. Limited demand for training puts pressure on data centers competing for inference. Miners can compete for inference contracts. Helps smooth out revenue volatility but potential to still have pressure on operating cash flows. |
| E | Weak | Strong | Very constrained | Mining dominates; inference only if fully contracted | Most constructive. Constraints on capital mean limited training and data centers need to compete for inference contracts, resulting in more utilization for mining. Miners remain tied to volatile revenue streams which can put pressure on operating cash flows. |
Disclosure Title
Disclosures
Disclosure
Source: Schwab Center for Financial Research.
For illustrative purposes only.