Skip to content

Republican calls out Trump on GPU sales to China

Technology
12 8 60
  • This post did not contain any content.
  • This post did not contain any content.

    Following the decision, Nvidia estimated the decision would cost the GPU giant nearly $10.5 billion in the first half of its 2026 fiscal year. AMD reported the export ban would cost the firm $1.5 billion in lost revenues in 2025.

    and they wouldn't be able to find customers elsewhere for their in-demand products to mitigate some or even all of that "lost" revenue?

  • This post did not contain any content.

    The irony is the chip ban is exactly what advanced Chinese AI. They were reliant on Nvidia like everyone, choked off, then:

    • Forced to get thrifty, collaborate and acutally innovate, while the US spins its wheels doing private (read: unshared and hoarded) research, skipping that to scale up instead or turning to focus on 'products' (to quote Zuckerberg): https://www.techrepublic.com/article/news-meta-llama-ai-mark-zuckerberg/
    • It forced them to grow their own training hardware, which they now have: Huawei NPUs. Along with pretty good models specifically designed for them, and power efficient on them, and open weights: https://huggingface.co/IntervitensInc/pangu-pro-moe-model
    • Being a 'step' behind saved money. And now they're racing ahead as US firms plateau and flounder as the AI Bro hype crashes into reality.

    So this is just Republicans (and Democrats) being paid off by Tech Bros.

    Trump ultimately did the right thing here (and was probably talked into it by the Nvidia CEO, TBH). But its too late anyway.

    There are other factors too (like Chinese companies seemingly sharing unspecified training data, maybe from the Chinese government, which I don't see evidence other countries are doing), but still.

  • This post did not contain any content.

    These imbeciles are funny.

    Did you know broadcom uses excess capacity in its semiconductor fabs to make the raspberry pi stuff? They sell it to the """ nonprofit """ Rπ foundation at cost to manufacture. This is not some charitable arrangement at all. Lower end hardware has expired patents and is capable of scaling into the computing space and growing from there. The path of least resistance created by the Rπ ecosystem suppresses grassroots adoption of any newcomers in the space. The unprofitable business structure for broadcom prevents scalable business investment by any competitors in low level compute. The actual Rπ chip is for TV tuners in particular. It is proprietary with only a partial datasheet for documentation. Three quarters of the actual die in the π is completely unused junk from the TV tuner stuff. In reality, if Rockchip could complete in a market without a monopoly and only compete on meritocratic value, broadcom would go out of business. The actual Rπ is barely good enough to suppress far newer and better spec hardware. All American businesses are anticompetitive crap of similar scope. The companies do not innovate and try to milk the lowest end ancient crap at a price point that makes large scale investments impossible, suppressing progress and innovation. Nvidia absolutely does this too. Buying a current GPU as a consumer is a joke of no value. They have produced the same tiers of VRAM for 3 generations. The 3090 series had firmware options all the way up to 32GB that only required the right chips and a configuration resistor to enable. Nvidia refused to let OEMs create models with more VRAM. If Nvidia was an honest business, a 5090 would likely be either 96 or 128 GB of VRAM and a notable value and progress. They do not do this because then their monopoly would be regulated. They are catfishing everyone, both consumers and competitors alike. Cutting them off from a market instantly makes domestic scalable competition possible.

    Either this halfwit red team is using spurious sophistry to criminally benefit from those that stand to gain massive market share, or they are so chronically incompetent we should tax the air they breathe to recoup losses suffered by the planet.

  • Following the decision, Nvidia estimated the decision would cost the GPU giant nearly $10.5 billion in the first half of its 2026 fiscal year. AMD reported the export ban would cost the firm $1.5 billion in lost revenues in 2025.

    and they wouldn't be able to find customers elsewhere for their in-demand products to mitigate some or even all of that "lost" revenue?

    I dunno how they figure that, yeah, as I thought Nvidia/AMD were supply limited by TSMC. Does it mean they’d raise prices? Get fewer barriers from China for other products, or maybe ramp up assembly there? Allocate more fab orders to the server GPUs?

  • This post did not contain any content.

    China will dominate everyone, and then Russia and India will join in.

  • The irony is the chip ban is exactly what advanced Chinese AI. They were reliant on Nvidia like everyone, choked off, then:

    • Forced to get thrifty, collaborate and acutally innovate, while the US spins its wheels doing private (read: unshared and hoarded) research, skipping that to scale up instead or turning to focus on 'products' (to quote Zuckerberg): https://www.techrepublic.com/article/news-meta-llama-ai-mark-zuckerberg/
    • It forced them to grow their own training hardware, which they now have: Huawei NPUs. Along with pretty good models specifically designed for them, and power efficient on them, and open weights: https://huggingface.co/IntervitensInc/pangu-pro-moe-model
    • Being a 'step' behind saved money. And now they're racing ahead as US firms plateau and flounder as the AI Bro hype crashes into reality.

    So this is just Republicans (and Democrats) being paid off by Tech Bros.

    Trump ultimately did the right thing here (and was probably talked into it by the Nvidia CEO, TBH). But its too late anyway.

    There are other factors too (like Chinese companies seemingly sharing unspecified training data, maybe from the Chinese government, which I don't see evidence other countries are doing), but still.

    I don't know, because according to Poe's data, American AI models are used more than Chinese ones. Although Poe is blocked in China, as are ChatGPT, Gemini, and Grok. So, Chinese AI models cover the Chinese market.

    American AI companies follow the competitive model, each developing their own AI models and occasionally publishing their research. With a few exceptions, they end up using a standard published by an American AI company.

    Chinese companies follow the collaborative model, although they also compete somewhat. They develop their own AI models, publish their AI code, publish their research, and use standards.

    The American government obtains AI models by awarding contracts to the companies that develop them. The Chinese government obtains AI models from its companies for free.

    But Nvidia still has a lot of market share in GPUs, only AMD has 4%, Huawei only 2% and the rest is shared by others, Intel and minority companies.

    In a race it is never too much if they are still running, it only applies when the race is already over.

  • I don't know, because according to Poe's data, American AI models are used more than Chinese ones. Although Poe is blocked in China, as are ChatGPT, Gemini, and Grok. So, Chinese AI models cover the Chinese market.

    American AI companies follow the competitive model, each developing their own AI models and occasionally publishing their research. With a few exceptions, they end up using a standard published by an American AI company.

    Chinese companies follow the collaborative model, although they also compete somewhat. They develop their own AI models, publish their AI code, publish their research, and use standards.

    The American government obtains AI models by awarding contracts to the companies that develop them. The Chinese government obtains AI models from its companies for free.

    But Nvidia still has a lot of market share in GPUs, only AMD has 4%, Huawei only 2% and the rest is shared by others, Intel and minority companies.

    In a race it is never too much if they are still running, it only applies when the race is already over.

    The Chinese models are locally hostable. This does not, and cannot count entities self hosting the models privately.

    The research posted by American AI companies (other than huggingface and a few startups) is pretty much a nothing burger.

    This is what I keep trying to tell everyone. It’s not US vs China nor AI vs no AI, the real battle is corporate APIs vs augmented, locally hosted, open weights and open research models.

    I hope the future is specialized models on smartphones, occasionally augmented by remote APIs. And that has a lot of gravity because, once set up, the calls are basically free.

    And AMD/Nvidia are still relevant in that future because they’ll likely be the one training models, at least.

  • These imbeciles are funny.

    Did you know broadcom uses excess capacity in its semiconductor fabs to make the raspberry pi stuff? They sell it to the """ nonprofit """ Rπ foundation at cost to manufacture. This is not some charitable arrangement at all. Lower end hardware has expired patents and is capable of scaling into the computing space and growing from there. The path of least resistance created by the Rπ ecosystem suppresses grassroots adoption of any newcomers in the space. The unprofitable business structure for broadcom prevents scalable business investment by any competitors in low level compute. The actual Rπ chip is for TV tuners in particular. It is proprietary with only a partial datasheet for documentation. Three quarters of the actual die in the π is completely unused junk from the TV tuner stuff. In reality, if Rockchip could complete in a market without a monopoly and only compete on meritocratic value, broadcom would go out of business. The actual Rπ is barely good enough to suppress far newer and better spec hardware. All American businesses are anticompetitive crap of similar scope. The companies do not innovate and try to milk the lowest end ancient crap at a price point that makes large scale investments impossible, suppressing progress and innovation. Nvidia absolutely does this too. Buying a current GPU as a consumer is a joke of no value. They have produced the same tiers of VRAM for 3 generations. The 3090 series had firmware options all the way up to 32GB that only required the right chips and a configuration resistor to enable. Nvidia refused to let OEMs create models with more VRAM. If Nvidia was an honest business, a 5090 would likely be either 96 or 128 GB of VRAM and a notable value and progress. They do not do this because then their monopoly would be regulated. They are catfishing everyone, both consumers and competitors alike. Cutting them off from a market instantly makes domestic scalable competition possible.

    Either this halfwit red team is using spurious sophistry to criminally benefit from those that stand to gain massive market share, or they are so chronically incompetent we should tax the air they breathe to recoup losses suffered by the planet.

    Always bet on incompetence

  • Always bet on incompetence

    Bet, sure. Without open skepticism for both, the worst of criminals get away with it. Pointing at the potential for corrupt gain is quite apropos for 47.

  • These imbeciles are funny.

    Did you know broadcom uses excess capacity in its semiconductor fabs to make the raspberry pi stuff? They sell it to the """ nonprofit """ Rπ foundation at cost to manufacture. This is not some charitable arrangement at all. Lower end hardware has expired patents and is capable of scaling into the computing space and growing from there. The path of least resistance created by the Rπ ecosystem suppresses grassroots adoption of any newcomers in the space. The unprofitable business structure for broadcom prevents scalable business investment by any competitors in low level compute. The actual Rπ chip is for TV tuners in particular. It is proprietary with only a partial datasheet for documentation. Three quarters of the actual die in the π is completely unused junk from the TV tuner stuff. In reality, if Rockchip could complete in a market without a monopoly and only compete on meritocratic value, broadcom would go out of business. The actual Rπ is barely good enough to suppress far newer and better spec hardware. All American businesses are anticompetitive crap of similar scope. The companies do not innovate and try to milk the lowest end ancient crap at a price point that makes large scale investments impossible, suppressing progress and innovation. Nvidia absolutely does this too. Buying a current GPU as a consumer is a joke of no value. They have produced the same tiers of VRAM for 3 generations. The 3090 series had firmware options all the way up to 32GB that only required the right chips and a configuration resistor to enable. Nvidia refused to let OEMs create models with more VRAM. If Nvidia was an honest business, a 5090 would likely be either 96 or 128 GB of VRAM and a notable value and progress. They do not do this because then their monopoly would be regulated. They are catfishing everyone, both consumers and competitors alike. Cutting them off from a market instantly makes domestic scalable competition possible.

    Either this halfwit red team is using spurious sophistry to criminally benefit from those that stand to gain massive market share, or they are so chronically incompetent we should tax the air they breathe to recoup losses suffered by the planet.

    Do you have any sources for this? Curious about the rpi/broadcom part and tv tuner chips specifically.

  • Do you have any sources for this? Curious about the rpi/broadcom part and tv tuner chips specifically.

    He has several others. This is just the first link I saw:
    https://www.youtube.com/watch?v=6477JFdpqew

  • Elon Musk Floats a New Source of Funding for xAI: Tesla

    Technology technology
    11
    93 Stimmen
    11 Beiträge
    94 Aufrufe
    S
    What do I call it, kif? Ugh..... Sex-lexia
  • 122 Stimmen
    3 Beiträge
    37 Aufrufe
    captainastronaut@seattlelunarsociety.orgC
    Anytime I get one as an Uber I try to play stupid like I can’t figure out the door handles. Slam the doors, pull the emergency door release (if there is one), push against the motorized door close mechanism. Ask if there’s a shade for the glass roof. Anything to remind the driver that it’s not a good car, especially as a taxi.
  • (LLM) A language model built for the public good

    Technology technology
    18
    1
    131 Stimmen
    18 Beiträge
    152 Aufrufe
    D
    Is the red cross involved? Because if not, using a red cross in the article is misleading and potentially a crime.
  • 300 Stimmen
    47 Beiträge
    311 Aufrufe
    T
    I worked in a bank for a bit. Literally any transaction that's large and unusual for the account will be flagged. Also people do bonkers things with their money for the stupidest reasons all the time so all that one has to do if they're making large transactions is be prepared to talk to the bank and explain what's going on. Unless of course you are handling money in relation to organized crime, in which case you were fucked the moment the money touched the banking system
  • 254 Stimmen
    42 Beiträge
    244 Aufrufe
    dojan@pawb.socialD
    Don’t assume evil when stupidity I didn't, though? I think that perhaps you missed the "I don’t think necessarily that people who perpetuate this problem are doing so out of malice" part. Scream racism all you want but you’re cheapening the meaning of the word and you’re not doing anyone a favor. I didn't invent this term. Darker patches on darker skin are harder to detect, just as facial features in the dark, on dark skin are garder to detect because there is literally less light to work with Computers don't see things the way we do. That's why steganography can be imperceptible to the human eye, and why adversarial examples work when the differences cannot be seen by humans. If a model is struggling at doing its job it's because the data is bad, be it the input data, or the training data. Historically one significant contributor has been that the datasets aren't particularly diverse, and white men end up as the default. It's why all the "AI" companies popped in "ethnically ambiguous" and other words into their prompts to coax their image generators into generating people that weren't white, and subsequently why these image generators gave us ethnically ambigaus memes and German nazi soldiers that were black.
  • 332 Stimmen
    35 Beiträge
    153 Aufrufe
    R
    We have batteries. But yeah, attacking the grid might be smart.
  • 38 Stimmen
    7 Beiträge
    48 Aufrufe
    D
    Not easy but not hard actually really simple if you had the right energy. Just ignore this so I don't scare you.
  • *deleted by creator*

    Technology technology
    1
    0 Stimmen
    1 Beiträge
    14 Aufrufe
    Niemand hat geantwortet