{"id":2376,"date":"2024-10-02T23:27:14","date_gmt":"2024-10-02T15:27:14","guid":{"rendered":"https:\/\/www.kafeizha.com\/?p=2376"},"modified":"2024-10-02T23:27:14","modified_gmt":"2024-10-02T15:27:14","slug":"%e4%b8%ad%e5%9b%bd%e6%94%bf%e5%ba%9c%e5%bb%ba%e8%ae%ae%e5%9c%b0%e6%96%b9%e4%bc%81%e4%b8%9a%e8%b0%a8%e6%85%8e%e4%bd%bf%e7%94%a8nvidia%e4%ba%ba%e5%b7%a5%e6%99%ba%e8%83%bd%e8%8a%af%e7%89%87","status":"publish","type":"post","link":"https:\/\/news.tomjun.com\/?p=2376","title":{"rendered":"\u4e2d\u56fd\u653f\u5e9c\u5efa\u8bae\u5730\u65b9\u4f01\u4e1a\u8c28\u614e\u4f7f\u7528Nvidia\u4eba\u5de5\u667a\u80fd\u82af\u7247"},"content":{"rendered":"

\u5317\u4eac\u52a0\u5927\u5bf9\u4e2d\u4f01\u8d2d\u4e70\u672c\u571fAI\u82af\u7247\u7684\u5f15\u5bfc\u529b\u5ea6\uff0c\u610f\u56fe\u501f\u6b64\u6269\u5927\u534a\u5bfc\u4f53\u884c\u4e1a\u89c4\u6a21\u5e76\u5bf9\u6297\u7f8e\u56fd\u7684\u5236\u88c1\u884c\u52a8\u3002\u76d1\u7ba1\u90e8\u95e8\u5df2\u529d\u8bf4\u4f01\u4e1a\u51cf\u5c11\u91c7\u8d2dNvidia\u65d7\u4e0b\u7528\u4e8e\u7814\u53d1\u4e0e\u8fd0\u884cAI\u6a21\u578b\u7684H20\u7cfb\u5217\u82af\u7247\uff0c\u63aa\u65bd\u4ee5\u6307\u5bfc\u6027\u6587\u4ef6\u5f62\u5f0f\u5448\u73b0\u800c\u975e\u53d1\u5e03\u7981\u4ee4\uff0c\u4ee5\u9632\u635f\u5bb3\u672c\u56fdAI\u521d\u521b\u516c\u53f8\u53d1\u5c55\u53ca\u52a0\u5267\u4e2d\u7f8e\u7d27\u5f20\u5173\u7cfb\u3002\u5317\u4eac\u6b64\u4e3e\u65e8\u5728\u4e3a\u56fd\u5185AI\u82af\u7247\u5236\u9020\u5546\u5f00\u62d3\u5e02\u573a\u7a7a\u95f4\uff0c\u5e76\u51c6\u5907\u5728\u6f5c\u5728\u66f4\u591a\u7f8e\u56fd\u9650\u5236\u6761\u4ef6\u4e0b\u4fdd\u969c\u672c\u571f\u79d1\u6280\u4f01\u4e1a\u7684\u9700\u6c42\u3002<\/p>\n

\u5177\u4f53\u800c\u8a00\uff0c\u76d1\u7ba1\u673a\u6784\u5411\u4f01\u4e1a\u53d1\u51fa\u7a97\u53e3\u6307\u5bfc\uff0c\u9f13\u52b1\u4f7f\u7528\u56fd\u4ea7\u4f9b\u5e94\u5546\u534e\u4e3a\u4e0eCambricon\u7684\u4ea7\u54c1\u4ee5\u51cf\u5c11\u5bf9Nvidia\u7684\u4f9d\u8d56\uff1b\u540c\u65f6\u901a\u8fc7\u5730\u65b9\u5546\u534f\u4f1a\u8fdb\u4e00\u6b65\u4f20\u8fbe\u6307\u793a\u3002\u7136\u800c\uff0c\u5317\u4eac\u4ea6\u5bf9\u672c\u5730\u516c\u53f8\u63d0\u51fa\u671f\u671b\uff1a\u5c3d\u53ef\u80fd\u6784\u5efa\u6700\u4f18\u79c0\u7684AI\u7cfb\u7edf\uff0c\u5373\u4fbf\u6b64\u8fc7\u7a0b\u4e2d\u53ef\u80fd\u9700\u91c7\u8d2d\u5916\u56fd\u534a\u5bfc\u4f53\u82af\u7247\u66ff\u4ee3\u56fd\u5185\u4ea7\u54c1\uff0c\u653f\u5e9c\u4e5f\u4f1a\u5bbd\u5bb9\u8fd9\u4e00\u60c5\u51b5\u3002<\/p>\n

\u5c31\u5168\u7403\u6570\u636e\u4e2d\u5fc3\u5e02\u573a\u800c\u8a00\uff0c\u82f1\u4f1f\u8fbe(Nvidia)\u4f5c\u4e3a\u6700\u5177\u4ef7\u503c\u7684\u82af\u7247\u5236\u9020\u5546\uff0c\u6b63\u53d7\u5230\u5404\u56fd\u7ade\u76f8\u8d2d\u4e70\u5176\u5904\u7406\u5668\u7684\u70ed\u6f6e\u9a71\u52a8\u3002\u5c3d\u7ba1\u4e2d\u7f8e\u8d38\u6613\u9650\u5236\u9020\u6210\u4e86\u5f71\u54cd\uff0c\u4e2d\u56fd\u5e02\u573a\u4f9d\u7136\u4fdd\u6301\u589e\u957f\u6001\u52bf\uff1b\u4eca\u5e74\u4e0a\u534a\u5e74\u4ece\u4e2d\u56fd\u53ca\u9999\u6e2f\u5730\u533a\u83b7\u5f97\u7684\u6536\u5165\u6bd4\u4f8b\u7ea6\u4e3a12%\uff0c\u603b\u8ba137\u4ebf\u7f8e\u5143\uff0c\u76f8\u6bd4\u53bb\u5e74\u589e\u957f\u4e86\u903e\u4e09\u6210\u3002<\/p>\n

\u6b64\u5916\uff0c\u7f8e\u56fd\u653f\u5e9c\u66fe\u7981\u6b62\u82f1\u4f1f\u8fbe\u5411\u4e2d\u56fd\u5ba2\u6237\u9500\u552e\u6700\u524d\u6cbfAI\u82af\u7247\u4ee5\u904f\u5236\u5317\u4eac\u7684\u6280\u672f\u8fdb\u6b65\u3002\u4e0d\u8fc7\uff0c\u516c\u53f8\u5df2\u8c03\u6574\u540e\u7eed\u7248\u672c\u4ea7\u54c1\u8bbe\u8ba1\uff0c\u4f7f\u5176\u7b26\u5408\u7f8e\u56fd\u5546\u52a1\u90e8\u89c4\u5219\u89c4\u5b9a\u3002<\/p>\n

\u9488\u5bf9Nvidia H20\u7cfb\u5217\u82af\u7247\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u4e2d\u56fd\u653f\u5e9c\u76f8\u5173\u90e8\u95e8\u8fd1\u671f\u901a\u8fc7\u7a97\u53e3\u6307\u5bfc\u65b9\u5f0f\u63d0\u793a\u4f01\u4e1a\u51cf\u5c11\u4f9d\u8d56\u3002\u76ee\u6807\u662f\u9f13\u52b1\u8f6c\u5411\u534e\u4e3a\u4e0eCambricon\u7b49\u56fd\u5185\u4f9b\u5e94\u5546\uff0c\u540c\u65f6\u5730\u65b9\u5546\u534f\u4f1a\u4e5f\u53c2\u4e0e\u5176\u4e2d\u52a0\u5927\u4e86\u5ba3\u4f20\u529b\u5ea6\u3002\u4e0e\u6b64\u540c\u65f6\uff0c\u5317\u4eac\u652f\u6301\u534a\u5bfc\u4f53\u4ea7\u4e1a\uff0c\u5e76\u7ed9\u4e88\u76f8\u5e94\u8865\u8d34\u653f\u7b56\uff0c\u4f46\u56fd\u4ea7AI\u82af\u7247\u76ee\u524d\u4ecd\u843d\u540e\u4e8eNvidia\u3002<\/p>\n

\u5c3d\u7ba1\u9762\u4e34\u7f8e\u56fd\u7684\u5236\u88c1\u6311\u6218\uff0c\u4e2d\u56fd\u7684AI\u9886\u57df\u4f9d\u7136\u53d1\u5c55\u8fc5\u901f\uff0c\u5b57\u8282\u8df3\u52a8\u3001\u963f\u91cc\u5df4\u5df4\u7b49\u516c\u53f8\u6301\u7eed\u6295\u5165\u5e76\u79ef\u6781\u63a8\u52a8\uff1b\u4f17\u591a\u521d\u521b\u4f01\u4e1a\u4e5f\u7ade\u76f8\u5bfb\u6c42\u9886\u5148\u5730\u4f4d\u3002\u5f53\u524d\u5728\u5927\u8bed\u8a00\u6a21\u578b\u8fd9\u4e00\u5173\u952e\u4eba\u5de5\u667a\u80fd\u6280\u672f\u65b9\u9762\u6709\u516d\u5bb6\u201c\u72ec\u89d2\u517d\u201d\u4f01\u4e1a\u5904\u4e8e\u7ade\u4e89\u524d\u6cbf\uff1a01.AI\u3001\u767e\u5ddd\u3001\u6708\u5149\u5c04\u7ebf\u3001MiniMax\u3001\u6b65\u98de\u70df\u4ee5\u53ca\u81f4\u6734\u3002<\/p>\n

\u6709\u6d88\u606f\u900f\u9732\uff0c\u5728\u9488\u5bf9H20\u7cfb\u5217\u82af\u7247\u7684\u653f\u5e9c\u6307\u4ee4\u4e0b\uff0c\u67d0\u4e9b\u4e2d\u4f01\u4ecd\u9009\u62e9\u5ffd\u89c6\u5b98\u65b9\u5efa\u8bae\u5e76\u6025\u4e8e\u5728\u5e74\u5e95\u524d\u53ef\u80fd\u5230\u6765\u7684\u7f8e\u56fd\u5236\u88c1\u4e4b\u524d\u5927\u91cf\u91c7\u8d2d\u6b64\u6b3e\u82af\u7247\uff1b\u540c\u65f6\uff0c\u4e3a\u6ee1\u8db3\u4e2d\u56fd\u653f\u5e9c\u7684\u8981\u6c42\uff0c\u8fd9\u4e9b\u4f01\u4e1a\u4ea6\u8d2d\u4e70\u534e\u4e3a\u81ea\u6709\u54c1\u724c\u7684\u82af\u7247\u3002 <\/p>\n

\n

\u65b0\u95fb\u6765\u6e90\uff1a<\/b>www.bloomberg.com
\u539f\u6587\u5730\u5740\uff1a<\/b>China Urges Local Companies to Stay Away From Nvidia\u2019s AI Chips<\/a>
\u65b0\u95fb\u65e5\u671f\uff1a<\/b>2024-09-27
\u539f\u6587\u6458\u8981\uff1a<\/b><\/font><\/p>\n

Beijing is stepping up pressure on Chinese companies to buy locally produced artificial intelligence chips instead of  products, part of the nation\u2019s effort to expand its semiconductor industry and counter US sanctions.Chinese regulators have been discouraging companies from purchasing Nvidia\u2019s H20 chips, which are used to develop and run AI models, according to people familiar with the matter. The policy has taken the form of guidance rather than an outright ban, as Beijing wants to avoid handicapping its own AI startups and escalating tensions with the US, said the people, who asked not to be identified because the matter is private.The move is designed to help domestic Chinese AI chipmakers gain more market share while preparing local tech companies for any potential additional US restrictions, the people said. The country\u2019s leading makers of AI processors include  and  Earlier this year, Beijing  local electric-vehicle makers to procure more of their supplies from local chipmakers, part of its campaign to reach self-sufficiency in critical technologies. Nvidia shares fell as much as 3.9% to $119.26 on Friday, extending an earlier decline, after Bloomberg reported the news. The stock has more than doubled this year. Read More: The US government banned Nvidia from selling its most advanced AI processors to Chinese customers , part of an attempt to limit Beijing\u2019s technological advances. Nvidia, based in Santa Clara, California, modified subsequent versions of the chips so they could be sold under US Commerce Department regulations. The H20 line fits that criteria.In recent months, several Chinese regulators, including the powerful Ministry of Industry and Information Technology, issued so-called window guidance \u2014 instructions without the force of law \u2014 to reduce the use of Nvidia, the people said. The notice was aimed at encouraging companies to rely on domestic vendors like Huawei and Cambricon, they added. Beijing also amplified the message via a local trade group, according to another. At the same time, Chinese officials want local companies to build the best AI systems possible. If that means they need to buy some foreign semiconductors over domestic alternatives, Beijing will still tolerate that, according to people familiar with China\u2019s AI policy. Nvidia declined to comment. China\u2019s Ministry of Commerce, Ministry of Information and Technology, and Cyberspace Administration didn\u2019t respond to faxed requests for comment. Separately, Nvidia Chief Executive Officer   that he\u2019s doing his best to serve customers in China and stay within the requirements of US government restrictions.\u201cThe first thing we have to do is comply with whatever policies and regulations that are being imposed,\u201d he said in an interview with Bloomberg Television. \u201cAnd, meanwhile, do the best we can to compete in the markets that we serve. We have a lot of customers there that depend on us, and we\u2019ll do our best to support them.\u201dNvidia, the world\u2019s most valuable chipmaker, has seen sales soar as data center operators across the globe scramble to buy more of its processors. China continues to be part of that growth, though trade restrictions have taken a toll. In the July quarter, it got 12% of its revenue, or about $3.7 billion, from the country, including Hong Kong. That was up more than 30% from a year earlier.\u201cOur data center revenue in China grew sequentially in Q2 and is a significant contributor to our data center revenue,\u201d Chief Financial Officer   during an earnings call in August. \u201cAs a percentage of total data center revenue, it remains below levels seen prior to the imposition of export controls. We continue to expect the China market to be very competitive going forward.\u201d Nvidia chips are the gold standard for companies looking to develop artificial intelligence services. The likes of , OpenAI and  have been  to scoop up its most cutting-edge products so they can build leading AI models. Several Chinese tech companies, including ByteDance Ltd. and Tencent Holdings Ltd.,  Nvidia\u2019s chips before the export controls took effect.Chinese chip designers and manufacturers, meanwhile, are working to introduce alternatives to Nvidia. Beijing has offered  to the semiconductor sector, but local AI chips remain well behind Nvidia\u2019s fare.China does have a burgeoning AI sector, though, despite the US restrictions. ByteDance and  have been investing aggressively, while a flock of startups are vying for leadership. There are six so-called tigers in developing large language models, the key technology behind generative AI: 01.AI, Baichuan, Moonshot, MiniMax, Stepfun and Zhipu.Some of the companies are turning a blind eye to the Chinese decree to avoid H20 chips and rushing to buy more of them before an anticipated sanction from the US by the end of this year, one of the people said, though they are also buying homemade Huawei chips to please Beijing.<\/pre>\n
<\/p>\n
<\/div>
\r\n