The real question with $CBRS isn't speed or scalability — those are table stakes for a wafer-scale architecture with on-chip memory.
What matters is economics and durability.
Do customers really need 59 wafers (roughly 20 compute racks) to run something like Kimi K3? What are the manufacturing yields on those wafer-scale chips? How many years can you count on them running without reliability issues creeping in?
Then there's the CS-6, which introduces 3D stacked DRAM. If that's feasible, you're stacking even more technical risk — more questions about yield, thermal management, and long-term stability.
If they can answer these questions convincingly — or if the market starts to believe they can — the stock could work. But until then, it's all about execution risk and whether the unit economics actually pencil out at scale.
What matters is economics and durability.
Do customers really need 59 wafers (roughly 20 compute racks) to run something like Kimi K3? What are the manufacturing yields on those wafer-scale chips? How many years can you count on them running without reliability issues creeping in?
Then there's the CS-6, which introduces 3D stacked DRAM. If that's feasible, you're stacking even more technical risk — more questions about yield, thermal management, and long-term stability.
If they can answer these questions convincingly — or if the market starts to believe they can — the stock could work. But until then, it's all about execution risk and whether the unit economics actually pencil out at scale.