英飞凌扩大与英伟达合作：氮化镓高压转换器面向 800 伏直流 AI 数据中心
慕尼黑——英飞凌科技 10 月 21 日宣布，扩大与英伟达在 800 伏直流（VDC）AI 数据中心供电架构上的合作，重点是供电链中最关键、要求最高的环节之一：高压中间母线转换器（IBC）。英飞凌表示，其氮化镓（GaN）技术正成为这一环节的主选技术。此次宣布恰逢英伟达 GTC 柏林开发者大会，并承接两家公司 2025 年 5 月首次公布、2025 年 10 月进一步细化的合作——当时的重点是可维护、可热插拔的 800 伏直流服务器机架。
英飞凌首席执行官 Jochen Hanebeck 表示：「人工智能正在从根本上重塑我们对电力基础设施的要求，而半导体正是解决这一挑战的核心。我们与英伟达在 800 伏供电架构上的合作，展示了深厚的半导体专业能力与人工智能所需创新速度结合时能够实现什么。基于氮化镓的系统解决方案是下一代数据中心的关键使能技术，英飞凌很自豪能帮助从电网到核心为人工智能时代供电。」
据英飞凌介绍，AI 负载的极端功耗推动数据中心采用中间母线架构：由 IBC 把主母线电压降至较低的轨电压，供下游环节为 AI 处理器生成严格的核心电压与负载点电压。随着 AI 基础设施扩张，800 伏直流正作为一种新增的配电选项，与既有交流及低压直流架构并行，面向更高密度的计算。机架功率持续攀升至兆瓦量级。
英飞凌称，与硅相比，氮化镓开关速度显著更快、损耗更低，使 IBC 环节的开关频率可达 1 MHz；频率提高令磁性元件与无源器件体积大幅缩小，直接转化为更高的功率密度与更低的板高。英飞凌的 GaN 产品组合覆盖高压 IBC 环节所需的全部转换比，包括 800 伏转 50 伏、800 伏转 12 伏与 800 伏转 6 伏，并可与 PSoC 控制 C3 及 EiceDRIVER 产品家族组成完整的 IBC 系统方案；其产品组合同时覆盖硅、氮化镓与碳化硅三条路线的 IBC 方案。
两家公司的合作贯穿数据中心供电链：除 IBC 外，还包括 2025 年 10 月宣布的电源路径保护方案，涉及热插拔控制器等安全与可维护功能，使未来服务器主板能在 800 伏直流架构下运行。英飞凌将自身定位为「从电网到核心」的 AI 基础设施功率半导体伙伴，方案覆盖从电网侧电源单元到处理器负载点的完整供电链。
英飞凌在 GSM《全球芯片公司排名榜》中居第 12 名（2025 财年营收约 168 亿美元），并以 23.2% 的份额居全球微控制器市场第一（Omdia，2025 年）。
Infineon Expands NVIDIA Collaboration: GaN-Based High-Voltage Converters for 800 VDC AI Data Centers
MUNICH — Infineon Technologies said on October 21 that it is advancing its collaboration with NVIDIA on 800-volt direct-current power architectures for AI data centers, centered on the high-voltage intermediate bus converter (IBC), one of the most critical and demanding conversion stages in the data center power delivery chain. The company's gallium nitride (GaN) technology is emerging as the technology of choice for this stage, Infineon said. The announcement, timed with NVIDIA's GTC Berlin developer conference, builds on joint work first announced in May 2025 and detailed further in October 2025, which centered on serviceable, hot-swappable 800 VDC server racks.
"AI is fundamentally reshaping the demands we place on power infrastructure, and semiconductors are at the heart of solving that challenge," said Jochen Hanebeck, CEO of Infineon Technologies. "Our collaboration with NVIDIA on 800 Volt power architectures shows what is possible when deep semiconductor expertise meets the pace of innovation required by AI. Gallium nitride-based system solutions are a key enabler of this next generation of data centers, and Infineon is proud to help power the AI era from grid to core."
Driven by the extreme power draw of AI workloads, data centers have adopted intermediate-bus architectures in which an IBC steps down the main bus voltage feeding accelerator cards into a lower-voltage rail, Infineon said. As AI infrastructure scales, 800 VDC is being introduced alongside existing AC and lower-voltage DC architectures as an additional distribution option for higher-density compute, as power levels per rack climb toward the megawatt range.
Compared with silicon, GaN switches significantly faster with lower losses, enabling switching frequencies of up to 1 MHz at the IBC stage, which shrinks magnetic components and passives and translates directly into higher power density and lower board heights, the company said. Infineon's GaN portfolio supports conversion ratios of 800 V to 50 V, 800 V to 12 V and 800 V to 6 V, and combines with its PSoC control C3 and EiceDRIVER families for full IBC system solutions; its IBC portfolio also spans silicon and silicon carbide. The two companies collaborate along the entire data center power chain, including power path protection solutions announced in October 2025, such as hot-swap controller functionality for future 800 VDC server boards. Infineon describes itself as a grid-to-core power semiconductor partner for AI infrastructure, from grid-facing power supply units to the point of load at the processor.
Infineon ranks No. 12 in GSM's Global Chip Company Rankings, with fiscal 2025 revenue of about $16.8 billion, and leads the global microcontroller market with a 23.2% share (Omdia, 2025).
来源：GSM·美国（原文链接）
图片来源：图：数据中心机房服务器机柜资料图。照片：Gideonwills44，CC BY-SA 4.0，经 Wikimedia Commons。Photo: Server racks in a data center (file photo). Photo: Gideonwills44, CC BY-SA 4.0, via Wikimedia Commons.
