内容来源：https://aibusiness.com/nvidia/china-ban-on-nvidia-chips-sparks-geopolitical-strife-market-shifts

内容总结：

中美AI芯片竞争近日再度升级。9月17日，中国网信办要求国内科技企业停止采购英伟达为AI市场设计的特供芯片RTX Pro 6000D，阿里巴巴、字节跳动等科技巨头已收到禁令。此举被视为中美科技博弈的最新动向。

作为回应，英伟达宣布向美国芯片制造商英特尔投资50亿美元共同设计AI芯片，并获得美国政府90亿美元资金支持，这一联盟将对AMD和英国Arm公司构成竞争压力。受禁令影响，英伟达股价下跌近3%，其CEO黄仁勋公开表示失望。分析师指出，中国业务约占英伟达总收入的15%，若禁令持续将对其造成显著冲击。

业内专家认为，中国目前尚未拥有能完全替代英伟达的国产AI芯片，但长期来看不应低估中国的发展潜力。HyperFrame Research分析师预测，中国AI芯片性能有望在五年内达到英伟达产品的80%。与此同时，这场科技竞争可能推动全球形成中美主导的两个独立技术生态体系，包括芯片、软件和企业平台的全面分化，这将增加全球企业的运营成本与技术合规风险。

深层次来看，中国此举可能为寻求摆脱美国技术依赖的国家提供范例。分析显示，澳大利亚、印度、瑞士等国已开始构建本土大语言模型，欧洲企业也在探索替代美国技术方案的路径。虽然多数国家难以完全脱钩，但全球科技体系正加速重构已成不争事实。

（注：本文根据公开信息综合整理，不代表任何机构立场）

中文翻译：

由谷歌云赞助
如何选择首个生成式AI应用场景
开展生成式AI项目时，应首先关注能够提升人类信息交互体验的领域。

芯片限令标志着中美人工智能技术竞争出现重大升级。中国网信办要求科技企业停用英伟达为AI市场设计的芯片，使得两国AI竞赛进入新阶段。尽管已下达订单，阿里巴巴、字节跳动等中国科技巨头目前被禁止采购或测试英伟达RTX Pro 6000D芯片。

9月17日实施的AI芯片禁令是中美AI争夺战的最新交锋。本周四英伟达向陷入困境的美国芯片制造商英特尔投资50亿美元，成为芯片领域的又一重大进展。这家AI芯片巨头将与英特尔共同设计用于数据中心和个人电脑的AI芯片，而后者已获得美国政府90亿美元的资金支持。此次联盟给英伟达的竞争对手AMD和英国安谋公司带来压力。

针对中国的芯片禁令，英伟达创始人兼CEO黄仁勋在英国记者会上表示对此感到失望。该消息导致公司股价下跌近3%。有观点认为中国监管机构的举措兼具战略考量与谈判筹码的双重属性。

"这可能是中美贸易谈判中的新杠杆，"RPA2AI研究公司首席执行官卡夏普·康佩拉指出，"若禁令持续，英伟达将损失约占总营收15%的中国市场业务。"他补充道："更重要的是，英伟达从AI芯片主导供应商转变为地缘政治工具，折射出全球科技生态体系正在经历深度重构。"

但中国短期内难以找到替代方案。"业界普遍认为中国尚未具备完美替代品，"Futurum集团分析师大卫·尼科尔森表示，"若认为已有替代方案，目前看来只是一厢情愿。"他同时指出美国科技经济也扮演着重要角色：若存在全球开放市场，中国企业本可自由采购所需芯片。

"众多美国企业正与中国公司合作满足其需求。不断变化的监管环境对各方均无益处，"尼科尔森强调，"某些拥有深厚合作关系的企业难免会受到冲击。"

HyperFrame研究公司CEO史蒂文·迪肯斯认为，紧张局势升级可能导致科技企业面临中美两套技术体系。他同时表示，虽然中国目前尚未拥有媲美英伟达的AI芯片或基础设施，但中国AI产业不容小觑。

"我不会低估中国，"迪肯斯说，"这是一个充满创新精神的国家。"他预测正如中国电动车行业已与美国并驾齐驱，中国AI芯片的工艺水平有望在五年内达到英伟达的80%。

康佩拉补充道，中美科技竞争预示着两个独立AI生态系统的形成："双方将各自发展芯片、软件和企业平台体系。这种分化将增加全球企业的成本与复杂性，带来技术绑定和合规风险。"

Deep Analysis创始人艾伦·佩尔兹-夏普则指出，中国可能为寻求摆脱美国技术依赖的国家提供范本，这不仅涉及硬件基础设施，还包括AI技术的其他要素。

"现实情况是，全球各国正在积极规划减少对美国技术的依赖，"佩尔兹-夏普表示。他注意到澳大利亚、印度、瑞士等国已开始构建本土大语言模型，欧洲企业也在探索脱离美国技术体系的途径。

"公平地说，各国与美国技术脱钩已持续多年，但去年和明年将是加速推进的关键时期。"尽管大多数国家无法完全脱钩，但中国是个例外。

您可能还喜欢

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The chip crackdown marks a significant escalation in the ongoing tension between the U.S. and China over AI technology.
The AI race between China and the U.S., has taken another turn after China's Cyberspace Administration told tech companies to stop using Nvidia chips manufactured for the AI market.
Big Chinese tech vendors, including Alibaba and ByteDance, are now not allowed to buy or test Nvidia's RTX Pro 6000D chips, even though companies have ordered them.
China's AI chip ban, imposed on Sept. 17, is the latest salvo in the China-U.S. AI war. Another big development in the AI chip arena came Thursday with Nvidia's investment of $5 billion in struggling U.S. chipmaker Intel.
The AI chip giant will co-design AI chips for data centers and personal computers with Intel, which the U.S. government has propped up with its own $9 billion investment. The new alliance puts pressure on Nvidia and Intel competitors AMD and U.K.-based Arm.
Meanwhile, in response to China's chip ban, Nvidia's CEO and founder, Jensen Huang, said during a press conference in the U.K. that he was disappointed by the development. The news caused the vendor's share price to drop by nearly 3%.
Some see the move by China's internal regulatory agency as either strategic or a bargaining tool.
"It could be another leverage point in the ongoing trade talks between the U.S. and China," said Kashyap Kompella, CEO of RPA2AI Research. "If the ban is not reversed, Nvidia loses significant Chinese business -- estimated to constitute 15% of its total revenue."
"Beyond that, Nvidia's shift from the leading supplier of AI chips to a geopolitical instrument underscores the broader reordering of the global tech ecosystem," he added.
However, China is unlikely to immediately come up with a substitute for the chips it is banning.
"It is generally accepted that China does not have perfect substitutes ... for what they are going to be denying their domestic companies," David Nicholson, an analyst at Futurum Group, said. "Thinking that they do, most people believe, is a bit of wishful thinking at this point."
Meanwhile, the U.S. tech economy also has a part to play because if it were an open market globally, Chinese vendors could buy any Nvidia chips they wanted, Nicholson said.
"There are a whole bunch of U.S. companies that are working with Chinese companies to deliver what these Chinese companies need. The ever-changing sand beneath them in terms of regulation doesn't help anybody," he continued. "It's impossible for them not to experience some penalties because there are companies that have extensive relationships that will be made more difficult."
Steven Dickens, CEO and analyst at HyperFrame Research, said the newly exacerbated tensions also could lead to a bifurcated system for tech companies working with both China and the U.S.
He added that while China does not yet have AI chips or infrastructure that rivals Nvidia's, the Chinese AI industry shouldn't be counted out.
"I wouldn't bet against China," he said. "China is an innovative nation."
And just as China's electric vehicle sector is now as competitive as that in the U.S., China's AI chip production and quality could become about 80% as good as Nvidia's in the next five years, Dickens said.
Kompella added that the AI and tech war between China and the U.S. also indicates a shift into two AI ecosystems: one led by China and another led by the U.S.
"Each ecosystem will develop a discrete stack of chips, software and enterprise platforms," he said. "This divide raises costs and complexity for global enterprises and can introduce technology lock-in and compliance risks."
However, Alan Pelz-Sharpe, founder of Deep Analysis, said it is also possible that China will set an example for other countries looking to break away from the U.S., not only with hardware infrastructure but also with other elements of AI technology.
"The reality is that worldwide countries are looking and actively planning to be less dependent or indeed independent of U.S.-based tech," Pelz-Sharpe said. He noted that countries like Australia, India and Switzerland have started building nation-specific large language models. European enterprises are also exploring ways to move away from U.S. tech stacks.
"To be fair, the decoupling of countries from U.S. tech dependence has been going on for years, but this past year and likely the next year are seeing a sharp acceleration," Pelz-Sharpe said.
While most countries can’t fully decouple, China is an exception.
You May Also Like