Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/nvidia-ceo-jensen-huang-says-we-dont-have-to-worry-about-the-chinese-military-using-us-chips-to-improve-their-capabilities-because-they-simply-can.23180/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Nvidia CEO Jensen Huang says 'we don't have to worry' about the Chinese military using US chips to improve their capabilities because 'they simply can

Daniel Nenni

Admin
Staff member
Nvidia CEO Jensen Huang sat down with CNN's Fareed Zakaria on Sunday to discuss a variety of issues, including the ongoing AI race between the US and China. Zakaria asked Huang about the previous bipartisan consensus regarding the restriction of high-end AI hardware to China, and to speak towards his previous comments that the sanctions had backfired against American companies.

"Depriving someone of technology is not a goal, it's a tactic. And that tactic was not in service of the goal", said Huang (via Bloomberg). "We would like the United States to be the world leader [in AI], there is nothing wrong with that aspiration, and we should definitely try to achieve that, and strive for that."

"Our mission, properly expressed... in order for America to have AI leadership", Huang continued, "is to make sure the American tech stack is available to markets all over the world, so that amazing developers, including the ones in China, are able to build on [the] American tech stack."

When asked by Zakaria whether this could potentially provide the Chinese military and intelligence services with "the capacity to supercharge their weapons with the very best American chips", Huang responded:

"We don't have to worry about that, because the Chinese military, no different [to] the American military, will not seek each other's technology out to be built on top of it. They simply can't rely on it. It could, of course, be limited at any time"

"Not to mention, there's plenty of computing capacity in China already. If you just think about the number of supercomputers in China, built by amazing Chinese engineers, that are already in operation."

"They don't need Nvidia's chips, certainly, or American tech stacks, in order to build their military."

Huang is scheduled to hold a media briefing in Beijing on July 16, his second visit this year after an earlier trip in April where he said he hoped to "continue to cooperate with China."

However, US republican senator Jim Banks and democratic senator Elizabeth Warren have sent a letter to Huang ahead of his trip, asking him to abstain from meeting with representatives of companies that are working with the People's Republic of China's military and intelligence bodies.

Jensen Huang

Credit: Nvidia
"We are worried that your trip to the PRC could legitimize companies that cooperate closely with the Chinese military or involve discussing exploitable gaps in US export controls", the letter warns.

The visit also comes in the wake of reports that China is currently constructing massive data centres to house over 115,000 Nvidia AI GPUs. This would appear be in direct contradiction of current US/China chip export restrictions surrounding high-end AI hardware, although it's unclear how the GPUs in question would be acquired.

The Trump administration's AI czar, David Stacks, has previously called for a relaxing of Biden-era regulations surrounding American-made AI chips, while an executive order regulating the developments of AI tools, software, and models was nixed early into Trump's current tenure.

Certainly, the Trump administration appears to look more favourably upon AI and AI hardware than the previous US government, so perhaps it's not unthinkable that the two countries could share AI developments (and chips) to their mutual benefit in years to come.

That being said, the US hit China with some of the largest trade tariffs of the lot at the start of the year, with little sign of let-up in recent months. So, whether Jensen's calming words might help lead to better technological relations between the two, or perhaps even a retraction of existing chip sanctions in the near future, is anyone's guess for now.

 
We need more strategic thinking like Jensen’s, when it comes to technology and the economy. Tactical tariffs to revive the wrong industries could harm us far more than help us.

Current policymakers are spending far too much time looking backward, fighting the last war. They should be spending much more time examining what’s emerging as a new China Shock.

We Warned About the First China Shock. The Next One Will Be Worse.​


 
We need more strategic thinking like Jensen’s, when it comes to technology and the economy. Tactical tariffs to revive the wrong industries could harm us far more than help us.

Current policymakers are spending far too much time looking backward, fighting the last war. They should be spending much more time examining what’s emerging as a new China Shock.

We Warned About the First China Shock. The Next One Will Be Worse.​



Wayne Gretzky: "I skate to where the puck is going to be, not where it has been."

In regards to semiconductors, US Politicians can't even skate. :cautious:
 

Chinese firms rush to buy Nvidia AI chips as sales set to resume​

  • - Nvidia expects US licences for H20 sales soon
  • - Chinese companies scramble to order chips
  • - Huang to brief media in Beijing during supply chain expo
  • - Nvidia also unveils new AI chip specifically for China
  • - AMD to resume shipments of MI308 AI chips as licenses approved
BEIJING/HONG KONG, July 15 (Reuters) - Chinese firms are scrambling to buy Nvidia's H20 artificial intelligence chips, two sources told Reuters, as the company said it planned to resume sales to the mainland days after its CEO met U.S. President Donald Trump.

 

Chinese firms rush to buy Nvidia AI chips as sales set to resume​

  • - Nvidia expects US licences for H20 sales soon
  • - Chinese companies scramble to order chips
  • - Huang to brief media in Beijing during supply chain expo
  • - Nvidia also unveils new AI chip specifically for China
  • - AMD to resume shipments of MI308 AI chips as licenses approved
BEIJING/HONG KONG, July 15 (Reuters) - Chinese firms are scrambling to buy Nvidia's H20 artificial intelligence chips, two sources told Reuters, as the company said it planned to resume sales to the mainland days after its CEO met U.S. President Donald Trump.

Isn't Chinese companies buy and use Huawei's ai chip?
 
Somehow I feel that in China , it’s hard to know it’s really the consumer companies buying it or they are acting as a front for the militaries….
 
Somehow I feel that in China , it’s hard to know it’s really the consumer companies buying it or they are acting as a front for the militaries….
absolutely. When any company/segment becomes large enough, it's no longer a private company's decision anymore.

As for NV CEO comments, let's keep in mind there is at least some conflict of interest.
 
We need more strategic thinking like Jensen’s, when it comes to technology and the economy. Tactical tariffs to revive the wrong industries could harm us far more than help us.

Current policymakers are spending far too much time looking backward, fighting the last war. They should be spending much more time examining what’s emerging as a new China Shock.

We Warned About the First China Shock. The Next One Will Be Worse.​


China Shock 2.0 is when Nvidia+OpenAI are replaced by Deepseek+Huawei. Cost and "good enough" performance disruption, Ala Christiansen.

Nvidia seem worried. They should be.
 
China Shock 2.0 is when Nvidia+OpenAI are replaced by Deepseek+Huawei. Cost and "good enough" performance disruption, Ala Christiansen.

Nvidia seem worried. They should be.

Have you tried ChatGPT and DeepSeek side by side? I have been using ChatGPT for the past 2+ years and it is improving dramatically. Not just speed but also quality of content. I don't think Nvidia is worried much about chip competitors.
 
China Shock 2.0 is when Nvidia+OpenAI are replaced by Deepseek+Huawei. Cost and "good enough" performance disruption, Ala Christiansen.

Nvidia seem worried. They should be.
There are definitely risks of something like that. But the risks are fare greater if NVIDIA isn't actively competing in the China market. Not sure if you read the article, but the greatest sins of current policy are not picking industry areas to focus on (tennis shoes as important as semiconductors), and tariffing off mature industries for protection instead of skating to where the puck is going.

It's probably a good idea to calibrate your thoughts about AI against what has happened since the the "big" Deepseek surprise. The article linked below points out that Deepseek, the company has hit a bit of a bust since achieving its newfound fame. The short version is that everybody learned from Deepseek the company and Deepseek models and infrastructure, and have actively reformulated their products to better meet market needs. Some have jumped on the Deepseek open model and done it one cheaper, better, faster. Others, like NVIDIA have learned how Deepseek squeezed out efficiency and generalized into AI OS features that allow everyone to get the same without all the laborious low-level coding.

Other big takeaway from this article are the 4-5 axes differentiation in LLM end products today:
1) Quality of query/reasoning results - all the subsequent comparisons / axes are based on two models delivering roughly equivalent quality of answers / actions. @Daniel A Nenni hinted that Deepseek is no longer a leader I that category, though at the start of the year, they were producing roughly equivalent results.
2) Cost per Million tokens - here's where all the inputs from OS efficiency, infrastructure cost & efficiency, chip cost & efficiency and energy costs all play a role.
3) Time to first token, which for reasoning, is a function of hardware architecture latency and token rate. This is the are that Groq and Cerebras seem most dialed in on.
4) Context window size - somewhat linked to quality. The context window size somewhat controls and limits what kind of applications the model can handle (how big the inputs and responses can be). This is one of the places that Deepseek shortchanged their "product" in order to keep the costs lower.
5) Open or closed models

I'll also point out that NVIDA is squarely taking aim at dramatically improving 2, 3 ,and 4 with their entire data center stack, to some degree based on some the same magic that gave Deepseek their efficiency, but usable by all LLMs on vLLM, SGLang, or TensorRT-LLM frameworks.

DeepSeek Debrief: >128 Days Later​

Traffic and User Zombification, GPU Rich Western Neoclouds, Token Economics (Tokenomics) Sets the Competitive Landscape

 
Back
Top