"Since ChatGPT kicked off the AI arms race in late 2022, Nvidia has shipped millions of GPUs predominantly for use in AI training and inference. That's a lot of chips that are going to be left idle when the music stops and the finance bros come to the sickening realization that using a fast-depreciating asset as collateral for multi-billion dollar loans wasn't such a great idea after all."
"These chips were originally designed to speed up video game rendering, which, by the late '90s, was quickly becoming too computationally intensive for the single-threaded CPUs of the time. As it turns out, the same thing that made GPUs great at pushing pixels also made them particularly well suited for other parallel workloads - you know, like simulating the physics of a hydrogen bomb going critical."
Nvidia's revenues are currently dominated by hardware sales driven by massive GPU shipments for AI training and inference since late 2022. Many GPUs risk becoming idle if AI investment collapses, but the chips retain value due to broad versatility. GPUs originated as graphics processing units for video game rendering and excel at parallel workloads, making them suitable for HPC and physics simulations; modern accelerators focus on vector and matrix math for AI. Broad programmability requires software, so Nvidia developed CUDA in 2007 and subsequently built hundreds of libraries, frameworks, and microservices to accelerate diverse parallel workloads.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]