SemiAnalysis Weekly
The SemiAnalysis team on AI chips, data centers and power: numbers down to dollars, watts and named vendors
Deep reads here2 / 36 episodes
Cadence~1 every 3.0 days
Latest2026-10-05
TopicsAI & Tech
PriorityT3
1:02:09SemiAnalysis Weekly
0916
Safety Is Not a Speed Bump: Anthropic's Compliance Pledge Actually Eats More Compute
Dario's ‘pacing the frontier’ gets read as buying fewer chips, but OpenAI's chain-of-thought monitoring alone burns 20% of compute — safety is itself a compute-intensive industry, and demand keeps outrunning supply.
8 Points
7 Quotes
ComputeAI Safety
38:52SemiAnalysis Weekly
0914
Nvidia's Flagship GPU Memory Regresses for the First Time: Why 4-HI HBM Wins
Rubin Ultra's HBM was cut from a promised 1TB to 192GB — less than the 288GB shipping today. This isn't a technical step backward, it's capacity rationing under the DRAM shortage: 4-HI stacks yield better and can squeeze out more than twice as many HBM cubes.
8 Points
6 Quotes
ComputeHBM