| Window | Theses | Resolved | Wins | Win rate | Avg return | Median return |
|---|---|---|---|---|---|---|
| 1d | — | — | — | — | — | — |
| 3d | — | — | — | — | — | — |
| 1w | — | — | — | — | — | — |
| 1m | — | — | — | — | — | — |
| 3m | — | — | — | — | — | — |
| 6m | — | — | — | — | — | — |
| 1y | — | — | — | — | — | — |
Past performance does not predict future results. Informational only, not investment advice.
GPUs should be viewed through the same structural, non-cyclical lens as memory because HBM is a legally contracted bottleneck for GPU deployment. As demand for AI and AGI scales aggressively, memory capacity locked through 2027 acts as a binding infrastructure constraint, supporting sustained pricing power for GPU makers.
GPUs should be viewed through the same structural, non-cyclical lens as memory because HBM is a legally contracted bottleneck for GPU deployment. As demand for AI and AGI scales aggressively, memory capacity locked through 2027 acts as a binding infrastructure constraint, supporting sustained pricing power for GPU makers.
Memory, particularly HBM3e/HBM4, has shifted from cyclical commodity to structurally constrained infrastructure locked into long-term contracts with hyperscalers through 2027. The custom 3D-stacked architecture and 3x wafer intensity vs. standard DRAM creates a durable supply bottleneck that should sustain pricing power, contrary to traditional cyclical valuation.
End of results.