Uploaded May 2026 | Updated September 2026, 10 hours ago
Same operation. 1,000x the energy. That's the gap between human cognition and today's GPU infrastructure.
Elad Raz puts the power problem in stark terms: your brain handles complex reasoning on 20 watts. A GPU rack doing the same computation burns 22 kilowatts. And that gap isn't going to be solved by shrinking transistors to 0.01 nanometers.
The efficiency revolution in AI compute isn't coming from manufacturing. It's coming from architecture.
đ Subscribe for the sharpest thinking on AI hardware, energy, and the future of compute.
Watch the full episode â bit.ly/3R4FAck
#AIEnergy #ComputeEfficiency #GPUPower #NextSilicon #BrainVsAI #EnergyEfficiency #AIInfrastructure #EladRaz #Semiconductors #Invested
Same operation. 1,000x the energy. That's the gap between human cognition and today's GPU infrastructure.
Elad Raz puts the power problem in stark terms: your brain handles complex reasoning on 20 watts. A GPU rack doing the same computation burns 22 kilowatts. And that gap isn't going to be solved by shrinking transistors to 0.01 nanometers.
The efficiency revolution in AI compute isn't coming from manufacturing. It's coming from architecture.
đ Subscribe for the sharpest thinking on AI hardware, energy, and the future of compute.
Watch the full episode â bit.ly/3R4FAck
#AIEnergy #ComputeEfficiency #GPUPower #NextSilicon #BrainVsAI #EnergyEfficiency #AIInfrastructure #EladRaz #Semiconductors #Invested
