Uploaded May 2026 | Updated September 2026, 9 hours ago
Hopper to Blackwell: 2x performance, 2x power. Blackwell to Rubin: 2x performance, 2x power. Nvidia is scaling linearly — more transistors, proportionally more watts.
Elad Raz explains why that's actually good news for NextSilicon. Their architecture isn't chasing the same curve. With $100M chip budgets vs. Nvidia's $1B+, they're forced to innovate at the architecture level — and the efficiency gains compound differently. Between generations, NextSilicon's improvements outpace the linear scaling Nvidia is locked into.
The gap isn't closing. It's widening.
🔔 Subscribe for deep dives into AI hardware, chip architecture, and the companies competing at the frontier.
Watch the full episode → bit.ly/3R4FAck
#Nvidia #Blackwell #NextSilicon #ChipEfficiency #ComputePerWatt #AIHardware #EladRaz #Semiconductors #Invested #JensensLaw
Hopper to Blackwell: 2x performance, 2x power. Blackwell to Rubin: 2x performance, 2x power. Nvidia is scaling linearly — more transistors, proportionally more watts.
Elad Raz explains why that's actually good news for NextSilicon. Their architecture isn't chasing the same curve. With $100M chip budgets vs. Nvidia's $1B+, they're forced to innovate at the architecture level — and the efficiency gains compound differently. Between generations, NextSilicon's improvements outpace the linear scaling Nvidia is locked into.
The gap isn't closing. It's widening.
🔔 Subscribe for deep dives into AI hardware, chip architecture, and the companies competing at the frontier.
Watch the full episode → bit.ly/3R4FAck
#Nvidia #Blackwell #NextSilicon #ChipEfficiency #ComputePerWatt #AIHardware #EladRaz #Semiconductors #Invested #JensensLaw
