4 ms·

Selling the _entire_ stake sounds really aggressive though? Is that normal? Even if you're all-in on OpenAI, does it not make financial sense to have _some_ st

Selling the _entire_ stake sounds really aggressive though? Is that normal? Even if you're all-in on OpenAI, does it not make financial sense to have _some_ stake in Nvidia considering they are the only ones with an actual moat? Unless there are CUDA alternative breakthroughs we will hear about in the next few days.

l5870uoo9y 1y ago
Isn't it merely a matter of time before China has an alternative?
- viraptor 1y ago
  They don't even need to invent anything. Just put in some work to polish what already exists: https://github.com/vosen/ZLUDA https://github.com/vosen/ZLUDA
  - actionfromafar 1y ago
    IMHO AMD should have done this a decade ago.
    - viraptor 1y ago
      This was sponsored by AMD for quite a while and dropped. There had to be some talks behind closed doors that resulted in the current situation. I mean, we all know they should've and they're not dumb either.
- chemmail 1y ago
  All you have to look at is ASIC miners. Once they had them, they were 10x faster than GPUs easily and made GPUs useless for those algos. Something very similar can happen soon.
  - adastra22 1y ago
    The fundamentals are different. Bitcoin mining is not intrinsically suited to acceleration on a GPU. It is a not-very-wide serial integer operation. AI inference on the other hand is basically just very large floating point tensor matrix multiplication. What does an ASIC for matmul look like? A GPU.
    - skylurk 1y ago
      Sorta? If that was the full story, TPU would not be a thing.
      - lukeschlather 1y ago
        I'm not an expert in chip design by any means but I think it's fair to say that TPU is a marketing term and it's not substantially different from a GPU like an H100. H100's cores are also called "Tensor Cores."
        jasonwatkinspdx 1y ago
        You are entirely mistaken. The TPU and GPU are organized very differently particularly with how the memory subsystem works. In the big picture, TPUs are systolic arrays. They don't have threading, divergence, or similar. GPUs in the big picture are SIMT, a hybrid of SIMD and multithreading where individual data streams in SIMD are relaxed to allow them to diverge somewhat. Memory Wise the TPU can keep the partial products right in the array. Parameters and weights are held in large on die scratch memories, and backing those are streams coming from the HBM. The TPU acts as a single giant CISC coprocessor, and has much more predictable memory and communication patterns vs a GPU, which its design exploits for higher efficiency at inference vs GPUs. So even if they use the word "tensor" and both have HBM based memory systems, how those are actually architected is very different.
      - adastra22 1y ago
        TPUs are not fundamentally different or more efficient than NVIDIA hardware. They are just cutting out the middleman.
vintermann 1y ago
For researchers and academics, cuda is painful to avoid, but I'm not sure that it is for large companies, once the time comes to train and deploy large models.
delaminator 1y ago
Amazon’s Idaho 2Gw data center it’s building for Anthropic has 0 Nvidia Cuda cores. https://finance.yahoo.com/news/amazon-says-anthropic-will-use-1-million-of-its-custom-ai-chips-141829311.html https://finance.yahoo.com/news/amazon-says-anthropic-will-us... > Anthropic is using 500,000 of Amazon’s Trainium2 chips for its Claude AI models as part of the tech giant's Project Rainier
creer 1y ago
> Selling the _entire_ stake sounds really aggressive though? Is that normal? They did not. They sold MOST of their stake in Nvidia long ago before anyone paid attention to Nvidia vs AI. It's one decision as part of a complex cloud of decisions. Also this "stake" was around one thousandth of Nvidia. Minor. Most companies have lots of owners with much larger stakes.
greycol 1y ago
If you believe the AI bubble will pop in the same way as the dotcom bubble and feel you can pick winners it makes sense. It would be like selling cisco stock for google. You can still invest in cisco later if you believe they'll grow but you might expect a stagnant period and definitely expect a drop as they don't have as much demand. NVidia is a safe bet to continue exisitng and being profitable but by being so central to the market you could almost guarantee a large drop in stock price while other companies are a gamble.

l5870uoo9y 1y ago

Isn't it merely a matter of time before China has an alternative?

viraptor 1y ago
They don't even need to invent anything. Just put in some work to polish what already exists: https://github.com/vosen/ZLUDA https://github.com/vosen/ZLUDA
- actionfromafar 1y ago
  IMHO AMD should have done this a decade ago.
  - viraptor 1y ago
    This was sponsored by AMD for quite a while and dropped. There had to be some talks behind closed doors that resulted in the current situation. I mean, we all know they should've and they're not dumb either.
chemmail 1y ago
All you have to look at is ASIC miners. Once they had them, they were 10x faster than GPUs easily and made GPUs useless for those algos. Something very similar can happen soon.
- adastra22 1y ago
  The fundamentals are different. Bitcoin mining is not intrinsically suited to acceleration on a GPU. It is a not-very-wide serial integer operation. AI inference on the other hand is basically just very large floating point tensor matrix multiplication. What does an ASIC for matmul look like? A GPU.
  - skylurk 1y ago
    Sorta? If that was the full story, TPU would not be a thing.
    - lukeschlather 1y ago
      I'm not an expert in chip design by any means but I think it's fair to say that TPU is a marketing term and it's not substantially different from a GPU like an H100. H100's cores are also called "Tensor Cores."
      - jasonwatkinspdx 1y ago
        You are entirely mistaken. The TPU and GPU are organized very differently particularly with how the memory subsystem works. In the big picture, TPUs are systolic arrays. They don't have threading, divergence, or similar. GPUs in the big picture are SIMT, a hybrid of SIMD and multithreading where individual data streams in SIMD are relaxed to allow them to diverge somewhat. Memory Wise the TPU can keep the partial products right in the array. Parameters and weights are held in large on die scratch memories, and backing those are streams coming from the HBM. The TPU acts as a single giant CISC coprocessor, and has much more predictable memory and communication patterns vs a GPU, which its design exploits for higher efficiency at inference vs GPUs. So even if they use the word "tensor" and both have HBM based memory systems, how those are actually architected is very different.
    - adastra22 1y ago
      TPUs are not fundamentally different or more efficient than NVIDIA hardware. They are just cutting out the middleman.

vintermann 1y ago

For researchers and academics, cuda is painful to avoid, but I'm not sure that it is for large companies, once the time comes to train and deploy large models.

delaminator 1y ago

Amazon’s Idaho 2Gw data center it’s building for Anthropic has 0 Nvidia Cuda cores. https://finance.yahoo.com/news/amazon-says-anthropic-will-use-1-million-of-its-custom-ai-chips-141829311.html https://finance.yahoo.com/news/amazon-says-anthropic-will-us... > Anthropic is using 500,000 of Amazon’s Trainium2 chips for its Claude AI models as part of the tech giant's Project Rainier

creer 1y ago

> Selling the _entire_ stake sounds really aggressive though? Is that normal? They did not. They sold MOST of their stake in Nvidia long ago before anyone paid attention to Nvidia vs AI. It's one decision as part of a complex cloud of decisions. Also this "stake" was around one thousandth of Nvidia. Minor. Most companies have lots of owners with much larger stakes.

greycol 1y ago

If you believe the AI bubble will pop in the same way as the dotcom bubble and feel you can pick winners it makes sense. It would be like selling cisco stock for google. You can still invest in cisco later if you believe they'll grow but you might expect a stagnant period and definitely expect a drop as they don't have as much demand. NVidia is a safe bet to continue exisitng and being profitable but by being so central to the market you could almost guarantee a large drop in stock price while other companies are a gamble.