观察台积电的碳化硅战略：碳化硅进入先进封装的主舞台

Original Articles By SemiVision Research (Nvidia ,TSMC, Globalwafers, Wolfspeed)SemiVision Research 原创文章（英伟达、台积电、环球晶圆、Wolfspeed） SEP 21, 20252025 年 9 月 21 日·PAID·付费SEMIVISION As the complexity and power demands of AI chip designs continue to escalate,Power Delivery Networks (PDNs)and thermal management have emerged asunavoidable core challenges for the industry. Traditional power deliverythrough PCBs or ABF substrates is becoming increasingly inadequate due tolong power paths and high parasitic inductance, leading to IR drops and transient voltage droops. With single GPUs now requiring over 1000A ofcurrent, these legacy approaches are approaching their limits. 随着 AI 芯片设计的复杂性和功耗需求不断增加，电源传输网络（PDNs）和热管理已成为行业不可避免的核心挑战。传统通过 PCB 或 ABF 基板进行供电的方法，由于电源路径长和寄生电感高，逐渐变得不够用，导致压降和瞬态电压波动。如今，单个 GPU 的电流需求已超过 1000A，这些传统方法正逐渐接近其极限。 To address these challenges, foundries and OSAT (Outsourced SemiconductorAssembly and Test) providers are actively proposing solutions across differentintegration levels. For instance, Marvell has introduced the PIVR (Package-Integrated Voltage Regulator), embedding the VRM directly into the package to shorten the power path and reduce PDN impedance, enabling faster voltageresponse. ASE, through its VIPack and FOCoS-Bridge (with TSV) platforms, isoptimizing PDN, signal interconnects, and thermal performance simultaneously—especially for Chiplet + HBM packages that demand high bandwidth and I/O,enabling reduced power loss without compromising performance. 为应对这些挑战，晶圆厂和 OSAT（封装测试）供应商积极提出不同集成层级的解决方案。例如，Marvell 推出了 PIVR（封装集成电压调节器），将 VRM 直接集成到封装中，以缩短电源路径并降低 PDN 阻抗，从而实现更快的电压响应。ASE 通过其VIPack 和 FOCoS-Bridge（带有 TSV）平台，同时优化 PDN、信号互连和热性能——特别是针对需要高带宽和 I/O 的 Chiplet + HBM 封装，实现降低能耗而不影响性能。 TSMC, on the other hand, is pursuing a more diversified integration strategy. Inits CoWoS-L platform, TSMC embedsIVRsandeDTCs(embedded Deep TrenchCapacitors) to enhance on-die decoupling density and power stability, whilealso developing BSPDN (Backside Power Delivery Network) to physicallyseparate power and signal layers. This reduces voltage drop and frees upvaluable front-side routing resources. TSMC is also pioneering thermal co-design, incorporating Direct Liquid Cooling (DLC), microchannel technologies,and exploring new materials for TIM (Thermal Interface Material) and underfill, with the goal of enabling reliable operation of kilowatt-class AI chips. 台积电则采取了更为多元化的集成策略。在其 CoWoS-L 平台中，台积电嵌入了 IVR（电压调节器模块）和 eDTC（嵌入式深沟电容器），以增强芯片内的去耦密度和电源稳定性，同时还开发了 BSPDN（背面电源传输网络），以在物理上分离电源层和信号层。这不仅减少了电压降，还释放了宝贵的前端布线资源。台积电还在引领热设计的协同创新，结合直接液冷（DLC）、微通道技术，并探索新型 TIM（热界面材料）和灌封材料，以实现千瓦级 AI 芯片的可靠运行。 Under these trends,NVIDIA and other hyperscalersare beginning to viewmaterialsas a critical pillar in AI chip and system design, actively pushing thesupply chain to explore diversified material options. Among these,SiliconCarbide (SiC)has suddenly become a focal point of attention. 在这些趋势下，英伟达等超大规模云服务提供商开始将材料视为 AI 芯片和系统设计中的关键支柱，积极推动供应链探索多样化的材料选择。其中，碳化硅（SiC）突然成为关注的焦点。 Proven over the past three decades in the power electronics sector,SiChasemerged as a star material due to itswide bandgap (WBG),high thermalconductivity,high breakdown field strength,excellent thermal stability, andhigh Young’s modulus. It has already been successfully commercialized inelectric vehicle inverters, industrial power modules, and high-voltage gridsystems. Now, with the explosive growth in AI computing demand, SiC’s role israpidly expanding intoadvanced packaging and heterogeneous integration. 在过去三十年的电力电子领域中，碳化硅（SiC）凭借其宽禁带（WBG）、高热导率、高击穿场强、优异的热稳定性和高杨氏模量，已成为一颗明星材料。它已在电动汽车逆变器、工业电源模块和高压电网系统中成功实现商业化。如今，随着人工智能计算需求的爆炸性增长，碳化硅的应用正迅速扩展到先进封装和异构集成领域。 SiC holds the potential to become thesubstrate material for high-voltage ICs(HVICs)andon-chip power delivery, supporting developments inBSPDNandIVRarchitectures. It also shows great promise as anoptical interposer, offeringa thermally conductive, low-loss platform forsilicon photonics (SiPh)andlithium niobate (LN) modulators. Furthermore, its high rigidity and thermalconductivity are beneficial for large-area3D stacking of Chiplets and HBM,helping reduce package warpage while enhancing mechanical reliability. 碳化硅有望成为高压集成电路（HVIC）和片上电源的基板材料，支持 BSPDN 和 IVR架构的发展。它在光学中介层方面也展现出巨大潜力，提供一种热导性强、低损耗的平台，用于硅光子（SiPh）和铌酸锂（LN）调制器。此外，其高刚性和热导率对于大面积芯片堆叠（Chiplets）和高带宽存储（HBM）具有积极作用，有助于减少封装翘曲，同时提升机械可靠性。 These unique properties positionSiC as a key material linking PDN, thermal,and optical interconnect domains, driving strong investment anddevelopment efforts from the Taiwanese supply chain. The goal: to establishnew industrial advantages in the age of AI and HPC. 这些独特特性使 SiC 成为连接 PDN、热学和光互连领域的关键材料，并推动台湾供应链进行大量投资和研发。目标是在 AI 和 HPC 时代建立新的产业优势。 Driven bygenerative AIandlarge-scale model training, companies likeNVIDIA,AMD,Google, andAWSare successively launching ultra-large-scaleGPUs and ASICspriced at tens of thousands of dollars each. These chipscontain hundreds of billions or even trillions of transistors, with powerconsumption often exceeding1 kW—plac