芯片力量：全球半导体征程与AI智造实录

《芯片力量》的主要内容分成三个部分：第一篇（第 1 ～ 3 章）是机遇篇，阐述历史机遇与产业历程，包括半导体产业在过去一个世纪中带给全球经济发展的机遇，以及大国在半导体机遇中的竞合与博弈历程。第二篇（第 4 ～ 7 章）是技术篇，阐述交叉跨界技术创新，以及新一代信息技术在半导体产业正发挥的、愈发重要的作用，这涉及芯片设计、制造、封测，也包括芯片制造设备厂商的应用实践与重要成果。第三篇（第 8、9 章）是管理篇，展望未来产业发展，包括如何看待和理解半导体产业在 21 世纪的爆发式增长，以及从产业发展管理及企业管理的视角出发，阐述如何更好地实现智能制造的升级管理。

前 言

亲爱的读者朋友，您好！

非常幸运与您一起踏入半导体征程与AI智造新世界！无论您是半导体集成电路产业或工业软件行业的从业人员，还是政策的谏言者或制定者，或是企业高管与专业人士，以及正加入到这一产业的投资者，希望本书可以帮助您一窥产业全貌，做出正确的决策。事实上，从2022年开始，无论是政府、投资机构，还是从业者，对新一代信息技术如何赋能半导体集成电路的发展都给予了高度关注和厚望。

从全球大国的科技与贸易博弈来看，美国近年一系列新的半导体产业发展政策相继出台，正试图把美国拉回先进制造大国的行列。数字化时代先进制造的代表就是芯片，而美国发展这一产业的大量投资源头便是中国，因为在过去的五年（2017—2022年），我国已成为世界最大的芯片消费国，为美国贡献了巨额的利润。制造与消费本是一对完美的上下游组合，但美国为了霸占科技鳌头并把持最为丰厚的产业利润，与其盟国在芯片领域持续打压中国，以遏制中国在科技领域不可逆转的崛起。芯片产业自主可控的战略意义不言而喻，国务院2020年发布的相关数据显示，我国芯片自给率在2025年要达到70%，而2019年的自给率仅约30%。近年我国芯片自给率不断提升的成果，从海关总署的数据中可见一斑:2022年的前7个月，我国进口芯片数量为3246.7亿个，同比下降11.8%，相当于减少了434亿个。

从发展历史来看，半导体集成电路产业最初是在20世纪60年代由军工需求带动而迅猛发展。根据IC Insights的统计，在近30年的时间里，半导体产业一直推动着全球GDP的增长。全球资本与产业智慧凝结于人类科技皇冠—“芯片”，它也成为全球科技战与贸易战的中心。芯片难，难于在全产业链进行资源的配置与协同，其中芯片制造更在产业链中处于投资规模最大、建设周期最长、尖端技术最密、良率要求最高的高价值环节。缺失了先进制造能力的芯片产业链就如“铁嘴豆腐脚”—在全球产业链竞争中容易被对手“卡脖子”，跑不出应有的速度。

从软硬件的组合来看，半导体集成电路是数字时代的硬件底座，而支持硬件智造与智能终端产品使用的软件大致可分为两类：操作系统的基础平台和各类应用程序。工业软件作为集成电路的灵魂，最初也是随着军工需求与硬件同步发展起来的。如今，毋庸置疑的是，以“数据科技（大数据）”和“计算科技（AI）”为代表的新一代信息技术，正成为按下第四次工业革命快进键的新生力量。在芯片智造中，除了我们熟知的工艺、设备、材料等必备条件外，难以突破的还有工业软件。工艺研发离不开专属软件，先进机台设备都内嵌智控软件，材料研发也需要通过软件来分析、配比和升级。而打通这三类生产要素形成智能一体化的稳定生产环境，也同样离不开工业智造软件。随着芯片的工艺制程不断微缩并朝着原子级尺寸迈进，无论是芯片的制造厂商，还是芯片的设备厂商，都在大量运用新一代信息技术助力前沿科研、产品的设计研发和制造。新型信息技术正发挥着至关重要且愈发重要的作用，这是一条已在国际先进制造业获得广泛验证、认同、推崇的必经之道，因此也被业界媒体称为神秘的“黑武器”。但“黑武器”并不容易掌握或获得，由于半导体及集成电路与国防军工密切相关，西方在产业背后建立了强固的政治贸易壁垒c；加上行业竞争异常激烈、技术垄断与霸权盛行，各厂商对核心技术研发和应用都是退藏于密；再就是专业领域知识的缄默性与抽象性提高了效仿和学习的门槛，使对手难以触及和掌握。试图打开这扇“芯片的新一代信息技术智造之门”正是本书的价值所在。

从新时代双循环的格局来看，要参与全球竞争，势必要先发挥自身优势建立内部的闭环。在我国半导体集成电路产业的闭环格局建设中，与国际对手正面的竞争是无法回避的，这是我国持续、坚定、大力支持发展工艺、设备、材料方面自主可控能力的原因。但我们也要看到正面战场的局限性与被动性。由于受《瓦森纳条约》、总体产业投资规模及目前国内尖端人才短缺等诸多方面的客观限制，在传统赛道的后发跟跑策略并不能令我们获得领先优势。所谓以正和、以奇胜，通过数据科技和计算科技进行赋能，则可以充分借力并发挥我国在新信息技术方面自主可控的优势，特别是近年已积累的海量数据及不断精进的深度学习AI算法，可以对芯片的制造特别是先进制造起到“四两拨千斤”的效果。

从宏观来看，我国在芯片领域积极推广和应用“数据科技”和“计算科技”的智造能力，已有了良好的政策、产业、技术与资本的基础。

首先，2015年3月国务院首次提出“推行国家大数据战略”之后，大数据及AI政策频出，工业制造领域在利好形势的引导和支持下，开始重视并挖掘数据科技和计算科技的巨大应用价值，跨界创新应用层出不穷，开启了“数智化”时代。中国已成为全球AI强国之一，一贯严谨保守的半导体产业也参与了进来。之后在

2018年12月，中央经济工作会议提出“新基建”，无疑将数字化转型提到了战略发展新高度。这并非只是一场技术革命，它与战略融合，已成为行业或企业顶层设计的重要部分。

其次，数字化转型在中国工业领域已走过数个年头，工业制造领域已打下了计算机/现代集成制造系统（Computer/Contemporary Integrated Manufacturing Systems，CIMS）的坚实基础。在2021年数字化转型发展高峰论坛上，信通院发布的数字化转型成果显示：“十三五”期间，我国数字经济增速超过了16.6%。产业数字化需要依靠企业实现，产业数字化聚焦的是传统产业的核心生产场景，提高的是整个产业的竞争力和经济水平。

最后，国家大基金一、二期共撬动万亿元的地方及社会资金大举进入半导体产业，科创板等二级市场兴起也为优秀科创企业铺平了一条新的科研兴国之路。半导体软件领域的投资规模超前，速度惊人。虽然一、二级市场随着行业周期与中美博弈而出现震荡，但没有人否认半导体集成电路将成为未来最为重要的数字化基建，它将推动全球从“数字化”到“数智化”再到“元宇宙”的发展进程，新技术、新概念、新机遇将层出不穷。

数据科技、计算科技与产业应用场景的深度融合，即半导体集成电路的产业数字化，是集成电路半导体企业本身技术发展的诉求与投资方向。此外，对于前瞻性的、行业共性的攻坚难题，也可以纳入政府的专项中来，一方面可以立足于AI大数据等新信息技术专项，另一方面也可以立足于芯片先进智造专项。多年以来，半导体集成电路与AI是科技兴国战略发展的两条线，新信息技术在半导体集成电路产业的加速应用推动了两条战略路线的融合。在半导体产业，工业智造软件的投入与重型基础设施的投入比起来只是九牛一毛，而可能产出的价值却是巨大的。基础科研不仅可以在工艺、设备、材料方面发力，也可以将生产要素与数据科技、计算科技融合起来，从而有望开创出一条更能发挥软实力竞争优势的、因地制宜的、扬长避短的、事半功倍的，对企业、产业甚至国家产生更大协同价值和溢出价值的新技术路径。

工业智造软件归属于工业软件，但为了聚焦芯片智造技术的阐述并与国际产业界的统称接轨，本书将使用工业软件作为切入点，但更多会使用“智造软件”来展开本书的主要内容。工业软件作为一个应用广泛、包罗万象的集合，其中能转化成商用的部分只是冰山一角，更具核心价值的“专用技术”往往秘而不宣（例如，美国波音公司研制了8000多种工业软件，进行商业化向市场开放了不到1000种）。本书聚焦于庞大的芯片智造工业软件体系中更具核心价值的部分，期望掀开其神秘面纱。第四次工业革命已经来到，工业软件也迎来智能时代，传统工业软件是必备的生存之本，而智造软件才是发展的制胜之道。对于产业界来说，由于采取不同技术路线的试错成本太高，非头部厂商极少开辟自己独有的技术路线，一般做法就是沿袭头部厂商的成功做法并进行二次创新。因此，了解这一维度的发展历史、价值起源、应用现状与未来趋势对于我国的芯片制造厂商来说意义重大。

但是，智造软件仅作为一类新的技术，并不是产业发展的护身符。技术与产业的结合应用，归根到底还是靠企业来实施和推动。那些举足轻重的领军企业特别需要先试先行，它们一是有足够的资本和资源，二是可以起到产业链的带动作用。诚然，芯片产业整体上除了需要弥补工艺、设备、材料等硬件与智造软件方面的短板外，还需要提升企业经营管理的水平。核心技术的背后是核心人才，核心人才的背后是文化与价值观、圈子与利益，而跨文化管理又是国际化半导体公司的普遍挑战。在半导体产业中，管理学流派似乎是被排斥在决策层之外的，因为需要依靠先进技术发挥竞争优势。那么，半导体产业的管理模式与经验真的与其他行业有天壤之别吗？事实上，全球半导体领军企业的实践证明并非如此。在全球大型半导体企业的CEO中，华人占据的比例越来越大。无论他们出生于何处，都有着共同的特质：既有工于科技、乐在坚毅的专业能力，又有敏锐的市场前瞻与商业洞察。懂科学可以搞科研，唯有懂经营才能做企业。这可能也是全球十大芯片设计厂商中有八家的掌门人为华人的重要原因吧。半导体是一个充分参与全球激烈市场竞争的产业，除了要有技术，更要有创新的商业思维。抛开地缘政治和自然资源因素，半导体产业发展有两大能力支柱：

（1）戈登·摩尔a在其“摩尔定律”中体现的持续性技术创新。按ITRS在2005年白皮书中的定义，又分化为“后摩尔”（More Moore）和“超摩尔”（More than Moore）两大趋势。

（2）张忠谋先生阐释台积电成功秘诀时谈及商业模式创新的重要性。他与台积电的成功给予了行业两大管理思想的贡献：其一是大家熟知的、始于1987年台积电创立之时的晶圆代工模式Foundry，当然这一模式同时催生了Fabless；其二则更为重要，就是台积电于2008年创立的开放式创新平台OIP，以此发展出的台积电大联盟已成为全球半导体顶尖公司的“盟友圈”。如果说第一个商业模式创新是把晶圆代工从IDM中“分”出来，第二个则是通过联盟的形式又“合”进去，从而形成一个全球化协同的“虚拟IDM”。30多年来，台积电的成功就在这样一分一合的商业模式创新中塑造起来，通过弯道超车一步一步令对手望尘莫及。

在这里我们强调企业管理的价值，除了如上的原因外，更重要的是数字化智造意味着一场制造的变革，它涉及公司内部的倡议、先试先行的预算投入、人才与团队的配置、对跨界融合创新的鼓励以及对初始失败的包容等，它不仅是一个技术问题，更是一个战略问题。数字化赋能智造是一项战略任务，对芯片制造业来说是实现战略目标的重要支撑力量，决策层只有在这个方面达成共识，并由主要负责人挂帅，数字化智造变革才能成功，这并非只是IT部门的工作。

从宏观的产业链发展与全球市场竞争来看，中国的情况与美国不完全相同，硬拼传统赛道的工艺、设备、材料等技术也可能落入“追赶者陷阱”，因此需要发挥我国特有的制度优势、管理优势和新一代信息技术优势，在某种程度上甚至可以通过工业智造软件来弥补硬件的不足。换句话说，就是“把握机遇+技术领先+管理卓越”。除了产业机遇之外，我国的制度优势为产业发展创造了前所未有的机遇；而技术创新的源头是解放的思想，这包括本书倡议的数智化跨界创新；商业成败的源头是决策，正确的决策则需要卓越的管理。芯片制造业由于精细的分工，行政权力被约束在“尊重科学与客观数据、掌握关键信息并集体决策”的框架内，更多的共识与协力是国际化团队特别需要加强的。因此，我们还是强调，在芯片智造的转型中，需要“把握机遇、技术领先、管理卓越”三者的结合。

所以，本书的主要内容也分成这三个部分：

第一篇是机遇篇，阐述历史机遇与产业历程，包括半导体产业在过去一个世纪中带给全球经济发展的机遇，以及大国在半导体机遇中的竞合与博弈历程。

第二篇是技术篇，阐述交叉跨界技术创新，以及新一代信息技术在半导体产业正发挥的、愈发重要的作用，这涉及芯片设计、制造、封测，也包括芯片制造设备厂商的应用实践与重要成果。

第三篇是管理篇，展望未来产业发展，包括如何看待和理解半导体芯片产业在21世纪的爆发式增长，以及从产业发展管理及企业管理的视角出发，阐述如何更好地实现智能制造的升级管理。

在资本与科技密集的半导体产业，技术通常是第一位的。但是在一个以领军企业为龙头的产业链竞争中，必须认清和把握时代机遇，顺势而为，同时充分提升现代科技企业的管理水平并优化治理结构，才能更好地培养和发挥技术优势，实现领先和超越。

我们正沿着20世纪工业化的步伐迈向21世纪的数字化、数智化和智人化，如果说我们在半导体过去30年的发展中忽略或错过了一些重要机遇，那么现在是我们树立远见卓识，更好地把握半导体未来10年辉煌发展的关键时刻。从第四次工业革命中提出的信息物理系统（Cyber-Physical System）到工业4.0的数字孪生（Digital Twins），再到元宇宙（Metaverse）……这一切都因芯片的发展成果而生，又推动芯片自身迈向更辉煌的未来。实现芯片强国梦是志在必得、砥砺奋进的征程。在半导体领域，从来少有投机取巧的一本万利，也鲜见文人骚客的闲情逸致，更无一厢情愿的天马行空，只能靠扎实的内功。

最后，由于半导体芯片产业涉及的知识纷繁复杂，本书作为在该领域的一次探索性尝试，必然存在错漏与不足之处，特别是由于地区差异造成的专有名词的不同叫法，容易翻译错误或引起误读。因此，您的任何批评、纠正和建议将是我们的宝贵财富！

本书最新配套视频内容，可在B站搜索“海俊频道”并访问“芯片力量”专栏查看。

谨以此书献给在半导体时代继续同行的我们！

李海俊

目 录

第1篇 机遇篇：半导体芯片全球进程与智造机遇

第1章 集成电路推动全球GDP增长与工业革命 ···············································2

1.1 集成电路推动全球GDP增长三十年与中国成就 ·······························2

1.1.1 半导体集成电路一直所处的战略领地 ···················································2

1.1.2 俄罗斯集成电路产业现状与未来电子战 ··············································18

1.1.3 我国软件与集成电路行业发展的三个十年 ···········································20

1.2 工业革命与不死摩尔定律 ··························································26

1.2.1 五次工业革命与半导体发展 ·····························································26

1.2.2 不死摩尔定律正从纳米深入埃米 ·······················································34

1.2.3 投资成本增势与产能预期 ································································38

第2章 美国科技制裁与中国自主替代 ······························································43

2.1 美国科技长臂管辖45年 ····························································43

2.2 美国白宫科技智囊与半导体军工组织 ···········································46

2.2.1 政府：近90年白宫科技智囊及其盟国科技智囊 ···································46

2.2.2 军工：DAPPA引领科研66年 ···························································51

2.3 凭使命改宿命、靠替代对制裁 ····················································56

2.3.1 国家科技咨询委员会智囊团呼之欲出 ·················································56

2.3.2 政、经、金、产四位一体推动集成电路行业发展 ··································58

2.3.3 《中国制造2025》中的集成电路 ························································63

2.3.4 中国行业巨头的跨界重塑 ································································67

第3章 集成电路与新信息技术交叉融合的智造机遇 ·········································73

3.1 智造工业软件生逢其时 ·····························································73

3.1.1 制造强国必强于工业软件 ································································73

3.1.2 工业软件支撑起全球最强工业企业 ····················································77

3.1.3 智造工业软件是半导体发展的黑武器 ·················································83

3.2 智能2—芯片与AI的交叉赋能 ·················································89

3.2.1 AI芯片引燃半导体产业爆发 ····························································89

3.2.2 AI是半导体智造的软核心 ·······························································93

3.2.3 集成电路与AI的互促成就 ······························································97

3.3 AI应用于集成电路的投资回报分析 ···········································100

第2篇技术篇：集成电路与New IT的跨界融合与智造技术

第4章智造软件持续加码全球半导体制造 ·····················································106

4.1 开启先进半导体智造之窗 ························································106

4.1.1 台湾的AI智造与竞争基础：工业3.5 ················································106

4.1.2 从生产自动化迈向工程自动化 ·························································111

4.2 半导体智造软件的极致力量 ·····················································114

4.2.1 半导体制造三大极致挑战 ·······························································114

4.2.2 工业互联数据汇聚的平台化 ····························································122

4.2.3 数据科技在半导体制造中崭露头角 ···················································125

4.3 智能学习仓库与数字孪生 ························································131

4.4 智造软件提升芯片制造的KPI ··················································134

4.4.1 良率是晶圆生产的生命线与终极挑战 ················································134

4.4.2 数据科技应用于制程良率管理 ·························································135

4.4.3 AI模糊神经网络赋能良率预测与生产排程 ·········································139

第5章智造软件为半导体产业提供全程价值 ·················································144

5.1 头部半导体厂商对AI应用的洞察·············································144

5.1.1 英伟达 ·······················································································145

5.1.2 科磊 ··························································································146

5.1.3 泛林 ··························································································147

5.1.4 欧洲微电子研究中心 ·····································································148

5.1.5 迈康尼 ·······················································································150

5.2 半导体服务厂商的智造方案 ·····················································151

5.2.1 DataProphet的AI即服务方案 ··························································151

5.2.2 Onto Innovation的创新数据驱动解决方案··········································152

5.2.3 D2S的GPU加速方案 ···································································154

第6章 来自世界头部半导体制造厂商的智造验证 ··········································156

6.1 英特尔20年的AI智造之路 ·····················································158

6.1.1 AI在英特尔整厂应用的方法论 ························································158

6.1.2 优化AI应用排序以提升商业价值 ····················································159

6.1.3 英特尔实现AI智造的典型案例 ·······················································161

6.2 台积电11年的AI智造与大联盟OIP··········································168

6.2.1 台积电的智造战略 ········································································169

6.2.2 台积电的智造案例 ········································································178

6.2.3 台积电向客户提供的虚拟晶圆厂 ······················································184

6.2.4 台积电大联盟的开放式创新平台 ······················································185

6.3 中芯国际的10年智造之路 ·······················································193

6.3.1 2011年打造云端工厂的成果 ···························································194

6.3.2 2018年关于实施智能制造战略的成果 ···············································196

6.3.3 2020年打造中芯国际工业互联网平台 ···············································198

6.4 其他知名半导体厂商的智造实践 ···············································199

6.4.1 格芯 ··························································································199

6.4.2 美光 ·························································································200

第7章 来自世界头部半导体设备厂商的智造验证 ··········································202

7.1 阿斯麦是卓越的工业软件公司 ··················································202

7.1.1 智控软件是光刻三十年来的灵魂 ······················································203

7.1.2 阿斯麦拥有世界最大开放软件社区 ···················································204

7.1.3 智能软件应用场景及案例 ·······························································205

7.1.4 EUV光刻机与 F-35隐身战机 ··························································207

7.2 应材的软硬一体 ····································································209

7.2.1 AI赋能晶圆缺陷检测 ····································································210

7.2.2 AI赋能晶圆制造产能爬坡及良率提升 ···············································211

7.2.3 AI赋能晶圆制造走向无人化“自动驾驶” ···········································213

7.2.4 应材的“全自动化半导体工厂”方案 ················································215

7.3 泛林的设备智能 ····································································217

7.3.1 泛林设备智能 ··············································································217

7.3.2 数字孪生/数字主线 ······································································217

7.3.3 虚拟工艺开发、智能工具与数字服务 ················································218

第3篇管理篇：未来科技与产业发展借鉴

第8章未来科技与半导体智造 ·······································································222

8.1 超级人类与未来科技······························································222

8.1.1 从“增长的极限”到“超级人类” ····················································222

8.1.2 中国“十四五”规划的七大前沿科技 ················································228

8.1.3 欧美未来科技预测及策略 ·······························································233

8.2 半导体智造的远景方略 ···························································235

8.2.1 半导体未来十年发展与挑战 ····························································235

8.2.2 半导体智造方略 ···········································································237

8.2.3 面向未来的工业4.0晶圆工厂 ··························································241

8.2.4 5G在半导体领域的前瞻性应用 ························································244

第9章半导体产业展望及工业4.0创新 ·························································246

9.1 美国半导体行业组织管理借鉴 ··················································246

9.1.1 SIA推动美国半导体产业发展 ·························································246

9.1.2 SEMATECH推动美国半导体制造 ····················································247

9.2 半导体工业4.0转型中的关键管理 ·············································250

9.2.1 数字化冠军企业转型的战略定位 ······················································250

9.2.2 数字化冠军企业转型的变革管理 ······················································252

9.2.3 英特尔、台积电与三星的创新转型案例 ·············································253

9.3 半导体产业工业4.0转型的框架应用 ··········································262

9.3.1 TüV工业4.0成熟度模型九宫格 ······················································262

9.3.2 EDB工业智能成熟度指数 ······························································264

9.3.3 IMPLUS工业4.0成熟度自评 ··························································267

结语 ··················································································································270

致谢 ··················································································································272