Will AI accelerators improve in FLOPs/watt by 100x of an NVidia H100 by 2033?
Basic
17
Ṁ33632033
90%
chance
1D
1W
1M
ALL
Compared to an H100, will tensor TFLOPs/Watt improve by 100x by 2033? AI accelerators in scope for this question must be deployed significantly - with at least 100k units or $100M (in 2024 dollars) in production, and have published perf/watt numbers.
This market will count peak FLOPs/watt at k bits of precision, adjusted by a factor of 2^(1 - 32/k). That is, 16 bit precision counts 1/4 as much as 32 bit, which counts 1/4 as much as 64 bit precision.
This question is managed and resolved by Manifold.
Get
1,000
and3.00
Sort by:
Related questions
Related questions
When will a US government AI run overtake private AI compute by FLOP?
Will 2024 be the year when AI capabilities progress from AI hardware scaling hits a wall?
26% chance
What will be the maximum achievable flop utilization on the next generation of Nvidia server chips?
100GW AI training run before 2031?
37% chance
Will AGI be powered by Nvidia GPUs?
61% chance
Will OpenAI inference costs fall by 100x over the next 18 months?
32% chance
Will AMD release a product that is competitive with NVIDIA in the AI hardware accelerator space before 2028?
76% chance
Will a machine learning training run exceed 10^27 FLOP in China before 2028?
44% chance
Will the Groq chip inspire Nvidia/AMD to produce radically new AI chips before 2026?
45% chance
At least one of the most powerful neural nets at end of 2026 will be trained using 10^27 FLOPs
81% chance