NVIDIA 与微软推出搭载 RTX Spark 和 AI Agent 的 Windows PC

周三在旧金山举行的一场微软活动上，黄仁勋与萨提亚·纳德拉阐述了 NVIDIA 和微软如何联合开发硬件与软件，使 AI Agent 能够在 Windows PC 上运行。

黄仁勋表示，NVIDIA 因 Windows 而创立。如今，AI Agent 正在进入 Windows。

“如果你纵观我们公司的整个历程，Windows 始终处于核心位置，”黄仁勋说道。他回顾了 NVIDIA 数十年来对 Windows 和微软的支持，“没有 Windows，就不会有 GeForce。”

周三，NVIDIA 创始人兼 CEO 黄仁勋在旧金山 Dogpatch Studios 举行的 Windows AI 与 Surface 活动上，与微软 CEO 萨提亚·纳德拉进行了一场炉边对话。对话由前白宫 AI 政策高级顾问 Sriram Krishnan 主持。黄仁勋将这段历史及其长期愿景，与 Agent 时代联系在一起。

“坦率地说，我真正要把所有功劳都归给 Jensen 的，是他对于这一方向的愿景始终如一——不只是看今天或明天，而是将其视为一种长期趋势，”纳德拉说。“这正是我们走到今天这一时刻的原因。”

“个人电脑是终极工具，是我的终极工具；对于整整一代人来说，它也是我们的终极工具，”黄仁勋说。“在 Agent 时代，当 Agent 运行在你的电脑上时，会发生什么？它现在就是你的个人助理。”

活动由纳德拉开场，随后微软 Windows 与设备执行副总裁 Pavan Davuluri 介绍了为 Agent 时代设计的 Windows 更新。

为将 Windows 打造为 Agent 的安全平台，微软宣布 Microsoft Execution Containers（MXC）正式全面可用。这一操作系统级基础设施可让 Agent 在操作系统控制下安全且持续地在后台运行。

“我们正将这些基础原语直接构建进 Windows，使 Agent 能够被保护、观测和治理，”Davuluri 解释道。“通过 MXC、Microsoft Security 和 Agent 365，我们正在释放 Windows 上 Agent 的全部潜力。”

纳德拉在与黄仁勋的对话中重申了这一主题。“我们需要让桌面成为 Agent 执行任务最安全的场所，”纳德拉说。

“Satya 刚才所说的内容，将成为我们所知的下一代 IT 基础，”黄仁勋说。“正如 Windows 和 DirectX 曾彻底改变应用构建方式，MXC 也将彻底改变 Agent 的构建与部署方式。”

Windows PC 的新起点：RTX Spark 笔记本今日开启预订

在黄仁勋和纳德拉周三介绍的多项发布中，RTX Spark 将完整的 NVIDIA AI 软件栈带入 Windows 笔记本和紧凑型台式机。笔记本产品今日开放预订，并将于 10 月 16 日上市；紧凑型台式机则将于 11 月开售。

Windows 正让本地 AI 更贴近日常工作，而新硬件则为开发者提供了在 PC 上直接运行能力日益强大的模型的空间。Davuluri 在发布 Surface Laptop Ultra 时，将这一愿景与 NVIDIA RTX Spark 联系起来。

“我们围绕 NVIDIA RTX Spark 打造了 Surface Laptop Ultra。它最高配备 128GB 统一内存和 1 PFLOPS AI 算力，因此你可以在这台笔记本上运行根本无法装入传统机器的模型，”Davuluri 说。

宏碁、华硕、戴尔、惠普、联想、微软、微星和技嘉均将推出相关系统，产品形态涵盖轻薄笔记本到专为始终在线 Agent 设计的紧凑型台式机。

RTX Spark 结合了拥有最高 6,144 个核心的 NVIDIA Blackwell RTX GPU，以及最高 20 核、通过 600 GB/s 互连的 NVIDIA Grace CPU。

1 PFLOPS 的 FP4 AI 性能及最高 128GB 统一内存，使 RTX Spark 适合本地 AI 应用。它可运行 Qwen 3.8 Flash Next 等模型；该模型拥有 1,250 亿参数和 510 亿 n-gram，其智能水平可媲美许多云端模型，同时无需按量计费，也无需将数据发送到云端。

RTX Spark 运行完整的 NVIDIA CUDA 平台——与 NVIDIA 各类硬件所使用的软件栈相同——并针对人们使用 PC 的各种方式打造：

• 开发者可以无需重写代码即可迁移模型、工具和工作流，在从 RTX Spark 到 DGX Station 的环境中运行同一套 NVIDIA AI 软件栈。
• 创作者可获得支持 NVFP4 的第五代 Tensor Core、硬件加速 AV1、4:2:2 视频编码与解码能力，以及覆盖完整制作流程的 DLSS 和 RTX 光线追踪。
• 游戏玩家可以借助 DLSS 5、Reflex 和 G-SYNC，以 1440p 分辨率和超过每秒 100 帧的速度运行 AAA 游戏。

紧凑型台式机配置将同一款 RTX Spark 超级芯片装入专为 24/7 运行设计的小型机箱中，构成一个让 Agent 持续运行的专用本地 AI 系统。今日预订 RTX Spark。

面向 Windows 的 NVIDIA DGX Station：企业桌面上的前沿 AI 算力

该活动当天预览了面向 Windows 的 NVIDIA DGX Station——这是一款首度将 GB300 Grace Blackwell 级 AI 基础设施直接带入 Windows 生态系统的桌边 AI 超级计算机。

“这释放了在本地运行前沿级模型的能力，”Davuluri 表示。“过去需要租用集群才能获得的能力，如今可通过一台桌边超级计算机实现。”

在此之前，DGX Station 运行在 Linux 上——这意味着企业开发者必须维护两套独立环境：一套 Linux 环境用于重型 AI 工作负载，另一套 Windows 环境用于生产力工具、应用程序和工作流。

绝大多数《财富》500 强公司都已标准化采用 Windows，而这种鸿沟一直在消耗开发者的时间和资源——开发者要么迁移到 Linux 以获得 AI 算力，要么留在 Windows 中并面对用于重型模型开发和多 Agent 工作负载的硬件选择有限的问题。

面向 Windows 的 NVIDIA DGX Station 基于 GB300 Grace Blackwell Ultra Desktop Superchip，提供 748GB 连贯内存和最高 20 PFLOPS 的 FP4 AI 算力——足以在本地运行最高万亿参数规模的模型。

AI 原生公司、领先研究实验室及企业中的开发者和研究人员团队，可构建并运行始终在线的 AI Agent，使其直接连接到现有的 Windows 应用程序和基础设施；他们还可在不离开主力机器的情况下，对大模型进行微调和推理。需要时，Linux AI 工具链仍可通过 WSL 访问。

了解更多有关面向 Windows 的 NVIDIA DGX Station的信息，并注册以便在其上市时第一时间获得通知。

At a Microsoft event in San Francisco on Wednesday, Jensen Huang and Satya Nadella outlined how NVIDIA and Microsoft are co-engineering hardware and software for AI agents to run on Windows PCs.

NVIDIA was founded because of Windows, Huang said. Now AI agents are coming to Windows.

“If you look at the entire journey of our company, Windows was at the core of it,” Huang said, tracing back NVIDIA’s support for Windows and Microsoft over several decades. “If not for Windows there would be no GeForce.”

NVIDIA founder and CEO Jensen Huang connected that history — and his long-term vision — to the agentic era Wednesday during a fireside chat with Microsoft CEO Satya Nadella hosted by Sriram Krishnan, a former senior White House policy advisor on AI at the Windows AI and Surface event, held at Dogpatch Studios in San Francisco.

“The thing that I really give Jensen all the credit for, quite frankly, is that consistency of vision of what this can be—not today, not tomorrow, but as a secular trend,” Nadella said. “And that’s what brings us to this moment.”

“The personal computer is the ultimate tool, it’s my ultimate tool, and for a whole generation of people, it’s our ultimate tool,” Huang said. “What happens in the era of agents, when the agent is on your computer? It is now your personal assistant.”

The event opened with Nadella before Pavan Davuluri, Microsoft’s EVP of Windows and Devices, walked through the Windows updates designed for the agentic era.

Turning Windows into a secure platform for agents, Microsoft announced general availability of Microsoft Execution Containers (MXC), the OS-level infrastructure that lets agents run safely and persistently in the background, under operating system control.

“We are building these primitives directly into Windows, so agents can be secured, observed and governed,” Davuluri explained. “With MXC, Microsoft Security, and Agent 365, we’re unlocking the full potential of agents on Windows.”

Nadella returned to that theme in his talk with Huang. “We needed to make the desktop the most secure place for agents to execute,” Nadella said.

“What Satya just said is going to be the foundation of the next generation of IT as we know it,” Huang said. “Just as Windows and DirectX revolutionized how applications were built, MXC is going to revolutionize how agents are built and deployed.”

A New Beginning for Windows PCs: Preorder RTX Spark Laptops Today

Among the announcements Huang and Nadella outlined Wednesday, RTX Spark puts the full NVIDIA AI stack into Windows laptops and compact desktops, with laptop preorders open today and available October 16. Compact desktops will be available for sale in November.

Windows is bringing local AI closer to everyday work, while new hardware gives developers room to run increasingly capable models right on their PC. Introducing Surface Laptop Ultra, Pavan Davuluri tied that vision to NVIDIA RTX Spark.

“We built Surface Laptop Ultra around NVIDIA RTX Spark. With up to 128 gigs of unified memory and up to a petaflop of AI compute, you can run models on this laptop that simply don’t fit on a traditional machine,” Davuluri said.

Systems are coming from Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte with designs ranging from slim laptops to compact desktops built for always-on agents.

RTX Spark combines an NVIDIA Blackwell RTX GPU with up to 6,144 cores, and an up to 20-core NVIDIA Grace CPU connected at 600 GB/s.

One petaflop of FP4 AI performance and up to 128GB unified memory makes RTX Spark effective for local AI. It can run models such as Qwen 3.8 Flash Next, a 125B model with 51B n-gram that matches the intelligence of many cloud models, unmetered and without sending data to the cloud.

RTX Spark runs the full NVIDIA CUDA platform — the same software stack that runs across NVIDIA hardware — and is built for every way people use a PC:

• Developers can move models, tools and workflows without rewriting, running the same NVIDIA AI stack from RTX Spark to DGX Station.
• Creators get 5th-generation Tensor Cores with NVFP4 support, hardware-accelerated AV1, and 4:2:2 video encode and decode, DLSS and RTX ray tracing across the full production pipeline.
• Gamers can run AAA games at 1440p over 100 frames per second with DLSS 5, Reflex and G-SYNC.

The compact desktop configuration puts the same RTX Spark superchip in a small chassis designed for 24/7 operation — a dedicated local AI system that keeps agents running continuously. Preorder RTX Spark today.

NVIDIA DGX Station for Windows: Frontier AI Compute on the Enterprise Desktop

The event previewed NVIDIA DGX Station for Windows today — the first deskside AI supercomputer to bring GB300 Grace Blackwell-class AI infrastructure directly into the Windows ecosystem.

“This unlocks the power to run frontier-class models locally,” Davuluri said. “Capabilities that once required renting a cluster, now in a deskside supercomputer.”

Until now, DGX Station ran on Linux — which meant enterprise developers maintained two separate environments: Linux for heavy AI workloads and Windows for the productivity tools, applications and workflows.

The vast majority of Fortune 500 companies are standardized on Windows, and that gap has cost developers time and resources — developers either moved to Linux to access AI compute, or stayed in Windows with limited hardware options for heavy-duty model development and multi-agent workloads.

NVIDIA DGX Station for Windows runs on the GB300 Grace Blackwell Ultra Desktop Superchip, delivering 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute — enough to run models up to a trillion-parameter scale locally.

Teams of developers and researchers at AI-native companies, leading research labs and enterprises can build and run always-on AI agents that connect directly to the Windows applications and infrastructure they already use, and fine-tune and inference large models without leaving their primary machine. Linux AI toolchains remain accessible through WSL when needed.

Learn more about NVIDIA DGX Station for Windows and sign up to be the first to know when it’s available.

Follow NVIDIA RTX Spark on X ,Instagram ,TikTok and Facebook — and stay informed by subscribing to the NVIDIA Local AI newsletter . Follow NVIDIA Workstation on LinkedIn and X .

See notice regarding software product information.