最高可能罚款超50亿美元，中国为何会对英伟达动手？
最高可能罚款超50亿美元，中国为何会对英伟达动手？
GPU巨头在中国惹上麻烦。
12月9日晚间，中国国家市场监管总局发布消息称，近日，因英伟达公司涉嫌违反《中华人民共和国反垄断法》及《市场监管总局关于附加限制性条件批准英伟达公司收购迈络思科技有限公司股权案反垄断审查决定的公告》（市场监管总局公告〔2020〕第16号），市场监管总局依法对英伟达公司开展立案调查。
消息一出，英伟达股价应声下跌。截至当地时间12月9日收盘，英伟达收跌近3%，市值一夜蒸发6400亿元；英特尔、AMD、阿斯麦、台积电、Arm等美股大型半导体公司也普遍收跌，像AMD最高跌近6%。
今天（12月10日），A股半导体概念集体大涨，尤其是AI芯片相关概念，例如寒武纪盘中涨超3%，海光信息涨超4%，景嘉微涨超5%，中芯国际、华虹半导体、南大光电、中微公司等产业链企业也纷纷飘红。
随后，英伟达方面回应国内媒体称，“英伟达凭借实力取胜，这反映在我们的基准测试结果和对客户的价值上，客户可以选择任何最适合他们的解决方案。我们努力在每个地区提供最好的产品，并在我们开展业务的任何地方履行我们的承诺。我们很乐意回答监管机构对我们业务的任何问题。”
英伟达这次被立案调查的具体原因是什么？近两年，美光、英特尔、英伟达等美国芯片巨头频频在中国被点名，在美国制裁不断扰乱供应链的背景下，中国四大行业协会也已明确指出，美国芯片已不再安全可靠，这又释放了哪些信号？对中国各行业发展又会带来哪些影响？
英伟达CEO黄仁勋
最高或被罚50亿美元，英伟达为何被调查？
这次对英伟达立案调查，监管机构是有充分依据的。
国家市场监管总局表示，英伟达涉嫌违反《中华人民共和国反垄断法》及《市场监管总局关于附加限制性条件批准英伟达公司收购迈络思科技有限公司股权案反垄断审查决定的公告》（市场监管总局公告〔2020〕第16号），因此依法决定对英伟达立案调查。
说起来，这件事可以追溯到5年多前。
2019年3月，英伟达宣布以69亿美元（约合人民币500亿元）的价格，收购美股上市公司迈络思，但考虑到交易金额、在华收入等因素，该收购需要中国反垄断机构的批准。
在经过一年多的审查后，中国国家市场监管总局认为，这项收购（经营者集中）对全球和中国GPU加速器、专用网络互联设备和高速以太网适配器市场，具有或者可能具有排除、限制竞争效果，因此决定附加限制性条件批准此项经营者集中。
随后，根据申报方提交的附加限制性条件承诺方案，国家市场监督管理总局要求英伟达、迈络思和集中后实体共履行五条义务。
这其中前两条是：
一、向中国市场销售英伟达GPU加速器与迈络思高速网络互联设备时，不得以任何方式强制进行搭售，或者附加任何其他不合理的交易条件；不得阻碍或限制客户单独购买或使用上述产品；不得在服务水平、价格、软件功能等方面歧视单独购买上述产品的客户。
二、依据公平、合理、无歧视原则向中国市场继续供应英伟达GPU加速器、迈络思高速网络互联设备和相关软件、配件。
这些限制性条件自生效日起6年后，交易双方和集中后实体可以向国家市场监管总局提出解除条件的申请，国家市场监管总局将依申请并根据市场竞争状况作出是否解除的决定。未经解除，交易双方和集中后实体应继续履行限制性条件。
但众所周知，自2020年以来，英伟达高端GPU已宣布对中国市场断供，转而对华销售“阉割版”的GPU，显然这违背了“不得阻碍或限制客户单独购买或使用上述产品”、“依据公平、合理、无歧视原则向中国市场继续供应英伟达GPU加速器”等承诺。
因此，从法律层面来看，中国这次对英伟达立案调查是有理有据的。
根据《中华人民共和国反垄断法》（下称：《反垄断法》）第五十八条等条款规定，经营者违反本法规定实施集中，且具有或者可能具有排除、限制竞争效果的，处上一年度销售额百分之十以下的罚款；不具有排除、限制竞争效果的，处五百万元以下的罚款。
财报显示，2024财年（截至今年1月底），英伟达在华收入为103.1亿美元，同比增长78%。
按照这个数据计算，英伟达可能面临1.03亿美元-10.3亿美元的罚款。
而按照《反垄断法》六十三条规定，违反本法规定，情节特别严重、影响特别恶劣、造成特别严重后果的，国务院反垄断执法机构可以在本法第五十六条、第五十七条、第五十八条、第六十二条规定的罚款数额的二倍以上五倍以下确定具体罚款数额。
因此，一旦英伟达的违法行为被认定情节特别严重、影响特别恶劣、造成特别严重后果，罚款金额可能达到原罚款额的2-5倍，最高可能达到20.6亿美元-50.15亿美元。
国产软硬件不是花架子，中国处置违规更有底气
曾几何时，由于中国科技产业存在一些短板，尤其在核心芯片领域，为了不对供应造成影响，市场普遍预期中国执法者不会轻易对芯片巨头“动手”。
但近两年，从美光未通过中央网信办网络安全审查，到中国网络空间安全协会点名英特尔“安全漏洞问题频发”，再到国家市场监管总局对英伟达立案调查，中国处置芯片巨头违法违规的底气明显更足了。
近期在美国再度升级对华制裁后，中国四大行业协会也集体发声，表示坚决反对美国对华采取出口限制，认为美国相关芯片产品不再安全、不再可靠，建议国内企业谨慎采购美国芯片。
这种底气背后，是中国科技产业的集体突围。
从CPU、GPU、NPU、DPU等各种核心处理器，到操作系统、数据库、AI框架等多种基础软件，再到已成为“国产之光”的存储芯片产业，国产软硬件早已不再是“花架子”。虽然部分领域的国产芯片和软件，在先进性上与国际巨头还有一定差距，但现有成果已可以支撑产业发展，并且在制裁围堵下加速迭代。
以AI领域为例，华为基于昇腾软硬件打造了昇腾云服务，百度昆仑与飞桨及其他国产软硬件组合打造了国产化AI技术生态。在AI芯片领域，海光信息推出了GPGPU架构的DCU处理器，寒武纪推出了思元370等系列加速卡，摩尔线程推出了MTT S4000系列等智算加速卡，壁仞科技也推出了壁砺106M等产品。
今年4月，中国移动智算中心（呼和浩特）投产使用，部署约2万张AI加速卡，AI芯片国产化率超85%，智能算力规模高达6.7EFLOPS（每秒670亿亿次半精度浮点运算）。
8月底，中国移动智算中心（哈尔滨）建成并投产使用，AI芯片国产化率100%，首次通过国产网络设备探索1.8万张智算卡单集群部署规模上限，可提供6.9EFLOPS智能算力。
不仅如此，中国联通北京公司建设了自主创新人工智能算力中心，可提供全国产化AI基础软硬件的300P公共算力能力；在上海临港，中国联通首个全液冷国产超万卡智算集群已建成发布；中国电信今年也已经完成了首个全国产化万卡万参大模型训练。
中国移动智算中心（呼和浩特）
这些案例印证了国产芯片的可用性。
有人可能会说，在先进制程被卡脖子的情况下，国产AI芯片的单卡算力与英伟达高端芯片存在明显差距，并且AI软硬件生态也没有英伟达完善，这会制约大模型训练和推理的效率。
这种说法有一定道理。毕竟就连华为轮值董事长徐直军也坦承，美国在AI芯片领域对中国的制裁长期不会取消，而中国半导体制造工艺由于也受美国制裁，将在相当长时间处于落后状态，这就意味着我们所能制造的芯片的先进性将受到制约。
但需要指出的是，随着AI模型参数越来越大、应用场景越来越复杂，单卡算力的优势正被逐步弱化，AI计算走向集群化是大势所趋，这就给国产算力中心的系统化方案带来契机。
徐直军今年曾直言，人工智能正在成为主导性算力需求，促使计算系统正在发生结构性变化，需要的是系统算力，而不仅仅是单处理器的算力。这些结构性变化，为我们通过架构性创新，开创出一条自主可持续的计算产业发展道路，提供了机遇。
“我们的战略核心就是，充分抓住人工智能变革机遇，基于实际可获得的芯片制造工艺，计算、存储和网络技术协同创新，开创计算架构，打造‘超节点+集群’系统算力解决方案，长期持续满足算力需求。”徐直军说道。
华为云近期也透露，基于CloudMatrix全新架构的昇腾AI云服务将于明年正式上线，CloudMatrix将CPU、NPU、DPU、存储和内存等资源全部互联和池化，从单体算力向矩阵算力演进，以架构性优势提升算力集群的性能。
可见，一方面国产AI芯片正逐步在产业中应用，另一方面中国AI产业也在探索系统化的解决方案，突破单卡算力瓶颈，提供可持续的算力解决方案。
事实上，在全球科技产业的发展中，中国一直秉持的是开放态度，但美国挑动逆全球化大行其道，不断围堵中国AI产业发展，让美国芯片变得不再安全可靠，中国AI产业势必将探索更多“中国方案”支撑产业智能化，到时候该着急的或将是英伟达。
多国反垄断调查英伟达，黄仁勋踩政治红线
在中国对英伟达立案调查之前，这家美国芯片巨头已经在多个国家惹上事。
今年7月，法国反垄断机构宣布正对英伟达进行反垄断调查。该国监管机构在一份生成式AI竞争调查报告中指出，对AI行业依赖英伟达CUDA平台来编写AI软件感到担忧。该报告还提到，英伟达投资美国AI云端基建初创CoreWeave，同样令监管机构感到不安。
根据财报，2024财年英伟达全球营收609.2亿美元，同比增长126%。若英伟达违反法国反垄断法的行为被坐实，该公司将面临全球年收入10%的罚款，约60.9亿美元。
同样在欧洲，近期欧盟委员会向英伟达的竞争对手和客户发出调查问卷，以调查英伟达是否在合同中将其图形处理器（GPU）与网络设备等其他产品捆绑销售。
这个调查可能会招致欧盟对英伟达的正式调查。
英伟达处理器
在英伟达主场，美国也已启动对该公司的反垄断调查。该公司被竞争对手投诉，在出售人工智能芯片时可能滥用市场支配地位。美国司法部正在调查英伟达是否强迫云计算供应商购买多款英伟达产品。调查同样涉及，一旦客户想购买AMD等英伟达同业对手的AI芯片，英伟达是否会对这些客户提高网络设备售价。
如果说英伟达涉嫌违反反垄断法更多是市场行为，那该公司CEO黄仁勋错误地将台湾称为“国家”，则是明显踩了政治红线。今年5月黄仁勋在参访台企时，被问及台湾在全球科技领域的地位，他以英语回答“台湾是世界上最重要的‘国家’之一”。在该公司财报中，“中国”、“台湾”和“美国”也处于并列位置。
国台办发言人陈斌华回应指出，对于这种极其错误的言论，大陆民众、网民已经纷纷表达强烈不满。“我要强调的是，两岸同属一个中国，台湾是中国的一部分，历史经纬清晰、法理事实清楚。台湾从来不是一个国家，过去不是，今后更不可能是。这是国际社会普遍共识和国际关系基本常识。希望他好好补补课。”
从招致各国反垄断调查，到踩中国政治红线，英伟达在全球似乎正处于引发众怒的境况。如果黄仁勋等高管不能妥善处理，这家美国芯片巨头或将遭遇更多麻烦。
Why would China take action against Nvidia when the maximum possible fine could exceed $5 billion?
GPU giants are causing trouble in China.
On the evening of December 9th, the State Administration for Market Regulation of China announced that it had recently launched an investigation into Nvidia due to suspected violations of the Anti Monopoly law of the People's Republic of China and the Announcement of the State Administration for Market Regulation on the Anti Monopoly review decision on the Approval of Nvidia's Acquisition of the Equity of Maxus Technology Co., Ltd. with Additional Restrictive Conditions (Announcement [2020] No. 16).
As soon as the news came out, Nvidia's stock price immediately fell.
As of the close on December 9th local time, Nvidia closed down nearly 3%, with its market value evaporating 640 billion yuan overnight;
Intel
AMD、
ASML, TSMC, Arm and other large semiconductor companies in the US stock market also generally closed down, with AMD falling as much as nearly 6%.
Today (December 10th), A-share semiconductor concepts collectively surged, especially AI chip related concepts. For example, Cambrian rose more than 3%, Haiguang Information rose more than 4%, Jingjia Microelectronics rose more than 5%, and industrial chain enterprises such as SMIC, Huahong Semiconductor, Nanda Optoelectronics, and Zhongwei Company also rose to fame.
Subsequently, Nvidia responded to domestic media by saying, "Nvidia wins with its strength, which is reflected in our benchmark test results and value to customers. Customers can choose any solution that best suits them. We strive to provide the best products in every region and fulfill our commitments wherever we conduct business. We are happy to answer any questions regulatory agencies may have about our business
What are the specific reasons for Nvidia being investigated this time?
In the past two years, American chip giants such as Micron, Intel, and Nvidia have been frequently named in China. Against the backdrop of US sanctions disrupting supply chains, the four major industry associations in China have also clearly pointed out that American chips are no longer safe and reliable. What signals does this send?
What impact will it have on the development of various industries in China?
picture
NVIDIA CEO Huang Renxun
Why is Nvidia under investigation with a maximum penalty of $5 billion?
The regulatory authorities have sufficient evidence to initiate an investigation into Nvidia this time.
The State Administration for Market Regulation stated that Nvidia is suspected of violating the Anti Monopoly Law of the People's Republic of China and the Announcement of the State Administration for Market Regulation on the Anti Monopoly Review Decision on the Approval of Nvidia's Acquisition of the Equity of Melos Technology Co., Ltd. with Additional Restrictive Conditions (Announcement [2020] No. 16 of the State Administration for Market Regulation), and therefore has decided to initiate an investigation into Nvidia in accordance with the law.
Speaking of which, this matter can be traced back to over 5 years ago.
In March 2019, Nvidia announced its acquisition of US listed company Maxus for $6.9 billion (approximately RMB 50 billion), but considering factors such as transaction amount and revenue in China, the acquisition requires approval from Chinese antitrust authorities.
After more than a year of review, the State Administration for Market Regulation of China believes that this acquisition (concentration of operators) has or may have the effect of excluding or restricting competition in the global and Chinese GPU accelerator, dedicated network interconnection equipment, and high-speed Ethernet adapter markets. Therefore, it has decided to approve this concentration of operators with additional restrictive conditions.
Subsequently, according to the additional restrictive condition commitment plan submitted by the applicant, the State Administration for Market Regulation required Nvidia, Maxus, and the centralized entity to fulfill a total of five obligations.
The first two of these are:
1、 When selling Nvidia GPU accelerators and Melos high-speed network interconnection devices to the Chinese market, bundling or attaching any other unreasonable transaction conditions shall not be enforced in any way;
Do not obstruct or restrict customers from purchasing or using the above-mentioned products separately;
Customers who purchase the above-mentioned products separately shall not be discriminated against in terms of service level, price, software functionality, etc.
2、 Based on the principles of fairness, reasonableness, and non discrimination, we will continue to supply Nvidia GPU accelerators, Maxus high-speed network interconnect devices, and related software and accessories to the Chinese market.
After 6 years from the effective date of these restrictive conditions, both parties to the transaction and the centralized entity may apply to the State Administration for Market Regulation to lift the conditions. The State Administration for Market Regulation will make a decision on whether to lift the conditions based on the application and the market competition situation.
Without termination, both parties to the transaction and the centralized entity shall continue to fulfill the restrictive conditions.
However, as is well known, since 2020, Nvidia's high-end GPUs have announced the discontinuation of supply to the Chinese market and instead sold "castrated versions" of GPUs to China, which clearly violates commitments such as "not obstructing or restricting customers from purchasing or using the above-mentioned products separately" and "continuing to supply Nvidia GPU accelerators to the Chinese market based on principles of fairness, reasonableness, and non discrimination".
Therefore, from a legal perspective, China's investigation into Nvidia is reasonable and well founded.
According to Article 58 and other provisions of the Anti Monopoly Law of the People's Republic of China (hereinafter referred to as the "Anti Monopoly Law"), if an operator violates the provisions of this Law by implementing concentration and has or may have the effect of excluding or restricting competition, a fine of no more than 10% of the previous year's sales revenue shall be imposed;
Those who do not have the effect of excluding or restricting competition shall be fined up to five million yuan.
According to the financial report, in the fiscal year 2024 (ending at the end of January this year), Nvidia's revenue in China was $10.31 billion, a year-on-year increase of 78%.
Based on this data, Nvidia may face fines ranging from $103 million to $103 million.
According to Article 63 of the Anti Monopoly Law, if a violation of the provisions of this Law is particularly serious, has a particularly adverse impact, or causes particularly serious consequences, the anti-monopoly law enforcement agency of the State Council may determine the specific amount of the fine between two and five times the amount specified in Articles 56, 57, 58, and 62 of this Law.
Therefore, once Nvidia's illegal behavior is determined to be particularly serious, have a particularly adverse impact, and cause particularly serious consequences, the fine amount may reach 2-5 times the original fine amount, with a maximum of 2.06 billion US dollars to 5.015 billion US dollars.
Domestic software and hardware are not just empty shelves, China has more confidence in dealing with violations
Once upon a time, due to some shortcomings in China's technology industry, especially in the core chip field, in order to avoid affecting supply, the market generally expected that Chinese law enforcement would not easily "take action" against chip giants.
But in the past two years, from Micron's failure to pass the Cyberspace Administration of China's cybersecurity review, to the China Cyberspace security Association's naming of Intel as having "frequent security vulnerabilities," to the State Administration for Market Regulation's investigation into Nvidia, China's confidence in dealing with the illegal and irregular behavior of chip giants has clearly increased.
After the recent escalation of US sanctions against China, the four major industry associations in China have collectively expressed their strong opposition to US export restrictions on China. They believe that US related chip products are no longer safe and reliable, and suggest that domestic companies purchase US chips with caution.
Behind this confidence lies the collective breakthrough of China's technology industry.
From various core processors such as CPUs, GPUs, NPUs, DPUs, to various basic software such as operating systems, databases, AI frameworks, and to the storage chip industry that has become the "light of domestic production", domestic software and hardware are no longer just "flower racks".
Although domestic chips and software in some fields still lag behind international giants in progressiveness, existing achievements can support industrial development and accelerate iteration under sanctions.
Taking the AI field as an example, Huawei has created Shengteng cloud service based on Shengteng software and hardware, and Baidu Kunlun has created a domestic AI technology ecosystem with PaddlePaddle and other domestic software and hardware combinations.
In the field of AI chips, Haiguang Information has launched DCU processors with GPGPU architecture, Cambrian has launched the Siyuan 370 series of acceleration cards, Moore Thread has launched the MTT S4000 series of intelligent computing acceleration cards, and Biren Technology has also launched products such as the Biren 106M.
In April of this year, the China Mobile Intelligent Computing center (Hohhot) was put into operation, deploying about 20000 AI acceleration cards. The localization rate of AI chips exceeded 85%, and the intelligent computing power scale reached 6.7EFLOPS (67 billion billion times per second of semi precision floating-point operations).
At the end of August, the China Mobile Intelligent Computing Center (Harbin) was completed and put into operation, with a 100% localization rate of AI chips. For the first time, the maximum deployment scale of 18000 intelligent computing cards per cluster was explored through domestic network equipment, which can provide 6.9EFLOPS intelligent computing power.
Not only that, China Unicom Beijing has established an independent innovation artificial intelligence computing power center, which can provide 300P public computing power for the national production of AI basic software and hardware;
In Lingang, Shanghai, China Unicom's first fully liquid cooled domestically produced intelligent computing cluster with over 10000 cards has been completed and released;
China Telecom has also completed the first nationwide production of a 10000 card, 10000 parameter large-scale model training this year.
picture
China Mobile Intelligent Computing Center (Hohhot)
These cases confirm the usability of domestic chips.
Some people may argue that in the case of advanced processes being constrained, there is a significant gap between the single card computing power of domestic AI chips and Nvidia's high-end chips, and the AI software and hardware ecosystem is not as perfect as Nvidia, which will constrain the efficiency of large model training and inference.
This statement has some truth.
After all, even Xu Zhijun, Huawei's rotating chairman, admitted that the US sanctions against China in the field of AI chips will not be lifted for a long time, and China's semiconductor manufacturing process will be in a backward state for a long time due to US sanctions, which means that the progressiveness of the chips we can manufacture will be restricted.
However, it should be pointed out that as the parameters of AI models become larger and the application scenarios become more complex, the advantage of single card computing power is gradually weakening. The trend of AI computing towards clustering is inevitable, which brings opportunities for the systematic solution of domestic computing power centers.
Xu Zhijun stated this year that artificial intelligence is becoming the dominant demand for computing power, leading to structural changes in computing systems. What is needed is system computing power, not just single processor computing power.
These structural changes provide us with opportunities to create an independent and sustainable development path for the computing industry through architectural innovation.
Our strategic core is to fully seize the opportunities of artificial intelligence transformation, innovate in computing, storage, and network technologies based on actual available chip manufacturing processes, create computing architectures, and create 'super node+cluster' system computing power solutions to continuously meet computing power needs in the long run, "said Xu Zhijun.
Huawei Cloud recently revealed that the Ascend AI cloud service based on the new CloudMatrix architecture will be officially launched next year. CloudMatrix will integrate CPU
NPU、DPU、
All storage and memory resources are interconnected and pooled, evolving from individual computing power to matrix computing power, and improving the performance of computing power clusters with architectural advantages.
It can be seen that on the one hand, domestic AI chips are gradually being applied in the industry, and on the other hand, China's AI industry is also exploring systematic solutions to break through the bottleneck of single card computing power and provide sustainable computing solutions.
In fact, in the development of the global technology industry, China has always adhered to an open attitude. However, the United States has provoked anti globalization and constantly blocked the development of China's AI industry, making American chips no longer safe and reliable. China's AI industry is bound to explore more "Chinese solutions" to support industrial intelligence, and Nvidia may be the one in a hurry.
Multinational anti-monopoly investigation into Nvidia, Huang Renxun steps on political red line
Before China launched an investigation into Nvidia, this American chip giant had already caused trouble in multiple countries.
In July of this year, the French antitrust agency announced that it was conducting an antitrust investigation into Nvidia.
The regulatory agency of the country expressed concerns in a report on the competition investigation of generative AI, stating that the AI industry relies on the NVIDIA CUDA platform to develop AI software.
The report also mentioned that Nvidia's investment in CoreWeave, a US AI cloud infrastructure startup, has also made regulators uneasy.
According to the financial report, Nvidia's global revenue for the fiscal year 2024 was $60.92 billion, a year-on-year increase of 126%.
If Nvidia's violation of French antitrust laws is confirmed, the company will face a fine of 10% of its global annual revenue, approximately $6.09 billion.
In Europe, the European Commission recently sent a survey questionnaire to Nvidia's competitors and customers to investigate whether Nvidia bundled its graphics processing units (GPUs) with other products such as network equipment in contracts for sale.
This investigation may lead to a formal investigation by the European Union into Nvidia.
picture
NVIDIA processors
At Nvidia's home ground, the United States has also launched an antitrust investigation into the company.
The company has been complained by competitors that it may abuse its market dominance when selling artificial intelligence chips.
The US department of Justice is investigating whether Nvidia forced cloud computing providers to purchase multiple Nvidia products.
The investigation also involves whether Nvidia will increase the price of network equipment for customers who want to purchase AI chips from competitors such as AMD.
If Nvidia's alleged violation of antitrust laws is more of a market behavior, then CEO Huang Renxun's mistake of referring to Taiwan as a "country" is clearly crossing a political red line.
In the company's financial report, "China", "Taiwan" and "the United States" are also in a parallel position.
Chen Binhua, spokesperson for the Taiwan Affairs Office of the State Council, responded by stating that mainland Chinese people and netizens have expressed strong dissatisfaction with such extremely erroneous remarks.
"What I want to emphasize is that both sides belong to one China, and Taiwan is a part of China, with clear historical longitude and latitude, and clear legal facts. Taiwan has never been a country, and it has not been, and will never be. This is the general consensus of the international community and the basic common sense of international relations. I hope he can make up for it."
From attracting anti-monopoly investigations from various countries to stepping on China's political red line, Nvidia seems to be in a situation of causing public outrage globally.
If executives such as Huang Renxun cannot handle it properly, this American chip giant may encounter more troubles.
1. For example, Cambrian rose more than 3%, Haiguang Information rose more than 4%, Jingjia Microelectronics rose more than 5%, and industrial chain enterprises such as SMIC, Huahong Semiconductor, Nanda Optoelectronics, and Zhongwei Company also rose to fame.
句子成分分析：
[For example], Cambrian rose [more] [than 3%], || Haiguang Information rose [more] [than 4%], || Jingjia Microelectronics rose [more] [than 5%], || and industrial chain enterprises (such as SMIC, Huahong Semiconductor, Nanda Optoelectronics, and Zhongwei Company) also rose [to fame].
句子语法结构详解：
(rose 为 rise 的过去式。)
* 第1个 rose 为谓语，采用一般过去时。
* Haiguang 开头为陈述句。
* 第2个 rose 为谓语，采用一般过去时。
* Jingjia 开头为陈述句。
* 第3个 rose 为谓语，采用一般过去时。
* 第1个 and 为并列连词，连接并列句。
* 第4个 rose 为谓语，采用一般过去时。
句子相关词汇解释：
Phrase:
| for example | 例如 |
| such as | 1) 例如，比如 2) 像……那样的，诸如……之类的 |
| rise to fame | 成名 |
Vocabulary:
| Cambrian ['kæmbriәn] | n. | 寒武纪, 威尔士人, 寒武系 |
| rise [raiz] | vi. | 1) 上升，攀升，提高，达到较高水平（或位置） 2) 起床，起立，站起来 |
| vi. | 1) 上升，攀升，提高，达到较高水平（或位置） 2) 起床，起立，站起来 | |
| more [mɔ:] | ad. | 1) (与形容词或副词连用，构成比较级)更 2) 更加; 更甚 |
| than [ðæn] | prep. | 比 |
| information [,infә'meiʃәn] | n. | 1) 信息，消息，情报，资料，资讯 2) 书的信息内容 |
| microelectronics | n. | 微电子学 |
| industrial [in'dʌstriәl] | a. | 1) 工业的，产业的 2) 用于工业的 |
| chain [tʃein] | n. | 1) 链条,链 2) 一连串 |
| enterprise ['entәpraiz] | n. | 1) 企业, 公司 2) 事业，规划 |
| semiconductor [,semikәn'dʌktә] | n. | 1) [术]半导体 2) 半导体装置 |
| and [ænd] | conj. | 1) 和, 与, 同, 并 2) 然后，接着 |
| company ['kʌmpәni] | n. | 1) 公司 2) 陪伴 |
| also ['ɔ:lsәu] | ad. | 而且；此外；也；同样 |
句子相关学习点：
2.Subsequently, according to the additional restrictive condition commitment plan submitted by the applicant, the State Administration for Market Regulation required Nvidia, Maxus, and the centralized entity to fulfill a total of five obligations.
句子成分分析：
[Subsequently], [according to the additional restrictive condition commitment plan submitted [by the applicant]], the State Administration [for Market Regulation] required Nvidia, Maxus, and the centralized entity to fulfill a total (of five obligations).
句子语法结构详解：
* submitted 为过去分词，作后置定语。
* state 开头为祈使句。
* state 为谓语。
* required 为谓语，采用一般过去时。
* to fulfill 为不定式，作宾语补足语(或称复合宾语)。
* the 为定冠词。a 为不定冠词。
句子相关词汇解释：
Phrase:
| according to | 根据，据（某人）所述 |
Vocabulary:
| subsequently ['sʌbsikwәntli] | ad. | 随后，后来，之后，接着 |
| additional [ә'diʃәnәl] | a. | 外加的，附加的，额外的 |
| restrictive [ri'striktiv] | a. | 1) 限制性的；约束的 2) 限制性的 |
| condition [kәn'diʃәn] | n. | 1) 环境，条件 2) 状态 |
| commitment [kә'mitmәnt] | n. | 1) 承诺，许诺，保证 2) 献身，奉献，投入 |
| plan [plæn] | n. | 1) 计划，打算 2) （详细）规划，方案，精心安排 |
| submit [sәb'mit] | vt. | 1) 提交，呈递（文件、建议等） 2) 顺从，屈服，投降，不得已接受 |
| applicant ['æplikәnt] | n. | 申请人（尤指求职、进高等学校等） |
| state [steit] | vt. | 1) 陈述，说明，声明 2) 规定，公布 |
| administration [әd,mini'streiʃәn] | n. | 1) （企业、学校等的）管理，行政 2) 施行，执行 |
| market ['mɑ:kit] | n. | 1) 市场，集市，商场 2) 交易，买卖，交易量 |
| regulation [,regju'leiʃәn] | n. | 1) 章程，规章制度，规则，法规 2) （运用规则条例的）管理，控制 |
| require [ri'kwaiә] | vt. | 1) 需要，有赖于 2) 要求，规定 |
| and [ænd] | conj. | 1) 和, 与, 同, 并 2) 然后，接着 |
| entity ['entiti] | n. | 独立存在物，实体 |
| fulfill [ful'fil] | vt. | 1) 履行，执行，符合，责备 2) 实现 |
| total ['tәutl] | n. | 总数，总额，合计，总计 |
| five [faiv] | num | 五 |
| obligation [,ɔbli'geiʃәn] | n. | 1) 义务，职责，责任 2) （已承诺的或法律等规定的）义务，责任 |
3.According to Article 58 and other provisions of the Anti Monopoly Law of the People's Republic of China (hereinafter referred to as the "Anti Monopoly Law"), if an operator violates the provisions of this Law by implementing concentration and has or may have the effect of excluding or restricting competition, a fine of no more than 10% of the previous year's sales revenue shall be imposed;
句子成分分析：
According to Article 58 and other provisions (of the [Anti Monopoly Law (of the People's Republic of China)])
(hereinafter referred to [as the "[Anti Monopoly Law]]"),
if an operator violates the provisions (of this Law) [by implementing concentration] and has or may have the effect (of excluding or restricting competition), || a fine (of no more than 10% (of the previous year's sales revenue)) shall be imposed;
句子语法结构详解：
(has 为 have 的第三人称单数形式。)
* hereinafter referred 为过去分词。
* if 为连词，引导条件状语从句。
* violates 为谓语，采用一般现在时。动词采用第三人称单数形式。
* implementing 为动名词。
* has 为谓语，采用一般现在时。动词采用第三人称单数形式。
* have 为谓语。
* excluding 为动名词，作介词宾语。
* restricting 为动名词。
* a 开头为陈述句。
* imposed 为谓语，采用被动语态。
* may, shall 为情态动词。be 为助动词。year's 为名词所有格。the 为定冠词。a/an 为不定冠词。
句子相关词汇解释：
Phrase:
| according to | 根据，据（某人）所述 |
| the People's Republic of China | 中华人民共和国 |
| no more than | 不超过, 仅仅 |
Vocabulary:
| article ['ɑ:tikl] | n. | 1) 文章，论文，报道 2) 物品,物件 |
| and [ænd] | conj. | 1) 和, 与, 同, 并 2) 然后，接着 |
| other ['ʌðә] | a. | 1) 其他的, 另外的, 其余的 2) 另一 |
| provision [prә'viʒәn] | n. | 1) 提供，供给，给养，供应品 2) （为将来做的）准备 |
| anti ['ænti] | prep. | 反对 |
| monopoly [mә'nɔpәli] | n. | 1) 垄断，专营服务，被垄断的商品（或服务） 2) 独占，专利，专利品 |
| law [lɔ:] | n. | 1) 法律，法规 2) 原理，定律 |
| hereinafter [,hiәrin'ɑ:ftә] | ad. | 以下, 在下文中 |
| refer [ri'fә:] | vi. | 提到, 涉及, 查阅, 查询, 咨询 |
| as [æz] | prep. | 1) 作为, 当作 2) 像，如同 |
| if [if] | conj. | 1) 如果, 倘若 2) 是否 |
| operator ['ɔpәreitә] | n. | 1) 操作人员，技工 2) 电话员，接线员 |
| violate ['vaiәleit] | vt. | 1) 侵犯（隐私等），使不得安宁，搅扰 2) 违反，违犯，违背（法律、协议等） |
| this [θis] | a. | 1) 这(个) 2) 今,本 |
| implement ['implimәnt] | vt. | 使生效，贯彻，执行，实施 |
| concentration [,kɔnsәn'treiʃәn] | n. | 1) 专注, 专心 2) 关注，重视 3) 浓度,含量 |
| or [ɔ:] | conj. | 1) 或，或者，还是 2) 否则，要不然 |
| have [hæv] | vt. | 1) 有，持有 2) 吃，喝, 抽(烟) 3) 患病，出现(某症状) |
| effect [i'fekt] | n. | 1) 效果，影响，作用 2) （艺术家，作家创作想得到的）效果，印象 |
| restrict [ri'strikt] | vt. | 1) （以法规）限制 2) 限制、限定(数量、范围等） |
| competition [,kɔmpi'tiʃәn] | n. | 1) 竞争 2) 比赛 |
| fine [fain] | n. | 罚款，罚金 |
| previous ['pri:viәs] | a. | 1) 先前的，以往的 2) （时间上）稍前的 |
| year [jiә] | n. | 1) 年 2) 一年时间 |
| sale [seil] | n. | 1) 销售，出售 2) 销售量 |
| revenue ['revinju:] | n. | （also revenues）财政收入，税收收入，收益 |
| impose [im'pәuz] | vt. | 1) 迫使，把……强加于 2) 推行，采用（规章制度），强制实行 |
句子相关学习点：
4. In the field of AI chips, Haiguang Information has launched DCU processors with GPGPU architecture, Cambrian has launched the Siyuan 370 series of acceleration cards, Moore Thread has launched the MTT S4000 series of intelligent computing acceleration cards, and Biren Technology has also launched products such as the Biren 106M.
句子成分分析：
[In the field (of AI chips)], Haiguang Information has launched DCU processors [with GPGPU architecture], || Cambrian has launched the Siyuan (370 series (of acceleration) cards), || Moore Thread has launched the MTT S4000 series (of intelligent computing acceleration cards), || and Biren Technology has [also] launched products (such as the Biren 106M).
句子语法结构详解：
* 第1个 launched 为谓语，采用现在完成时。
* Cambrian 开头为陈述句。
* 第2个 launched 为谓语，采用现在完成时。
* 370 开头为定语从句，引导词 that 被省略。
* 第1个 cards 为谓语，采用一般现在时。动词采用第三人称单数形式。
* Moore 开头为陈述句。
* 第3个 launched 为谓语，采用现在完成时。
* and 为并列连词，连接并列句。
* 第4个 launched 为谓语，采用现在完成时。
* has 为助动词。the 为定冠词。
句子相关词汇解释：
Phrase:
| such as | 1) 例如，比如 2) 像……那样的，诸如……之类的 |
Vocabulary:
| field [fi:ld] | n. | 1) 领域,界, 专业 2) 田,田地，牧场 3) (某种用途的)场地, 场所 |
| chip [tʃip] | n. | 1) 薄片，碎片，碎屑 2) 芯片 3) (chips)炸土豆条 |
| information [,infә'meiʃәn] | n. | 1) 信息，消息，情报，资料，资讯 2) 书的信息内容 |
| launch [lɔ:ntʃ] | vt. | 1) 发射，把（航天器、武器等）发射上天，水中发射 2) 开始从事，发起，发动（尤指有组织的活动） |
| processor [prә'sesә] | n. | 1) 加工机，处理机 2) [计算机]处理器 |
| architecture ['ɑ:kitektʃә] | n. | 1) 建筑学 2) 建筑设计，建筑风格 |
| Cambrian ['kæmbriәn] | n. | 寒武纪, 威尔士人, 寒武系 |
| launch [lɔ:ntʃ] | vt. | 1) 发射，把（航天器、武器等）发射上天，水中发射 2) 开始从事，发起，发动（尤指有组织的活动） |
| series ['siәri:z] | n. | 1) 一系列 2) 系列节目 |
| acceleration [әk,selә'reiʃәn] | n. | 1) 加速，加快 2) [物理]加速度 |
| card [kɑ:d] | vt. | 1) （用钢丝刷）梳理 2) 要求出示身份证（以确认年龄，如购酒） |
| n. | 1) 卡，卡片 2) 名片 | |
| thread [θred] | n. | 1) 线，棉线，毛线，丝线 2) 线索，脉络，思绪，思路，贯穿的主线 |
| launch [lɔ:ntʃ] | vt. | 1) 发射，把（航天器、武器等）发射上天，水中发射 2) 开始从事，发起，发动（尤指有组织的活动） |
| intelligent [in'telidʒәnt] | a. | 1) 有才智的，悟性强的，聪明的 2) （动物、生物等）有智力的，有理解和学习能力的 |
| computing | n. | 计算；计算机技术；信息处理技术 |
| technology [tek'nɔlәdʒi] | n. | 1) 技术，科技，工艺，工程技术，技术学，工艺学 2) 技术性机器（或设备） |
| also ['ɔ:lsәu] | ad. | 而且；此外；也；同样 |