高通将骁龙平台从智能手机扩展至个人AI设备
高通已将其骁龙平台产品线从智能手机扩展至面向下一代个性化（个人）人工智能（AI）设备。该公司正寻求借此拓展业务。
原文深度解读
本文报道高通在2026年骁龙峰会（Snapdragon Summit 2026，毛伊岛，当地时间9月24日）上宣布将其骁龙（Snapdragon）平台从智能手机扩展至面向下一代个性化（个人）人工智能（AI）设备。高通XR、可穿戴与个人AI业务高级副总裁兼总经理Ziad Asghar介绍了六大骁龙平台家族（骁龙8系、骁龙X、骁龙Wear、骁龙AR、骁龙Reality、骁龙Sound），并发布第二代骁龙Sound Elite音频平台，以及面向AI眼镜的AR1、AR1+、AR2系列。关键数字包括：微NPU架构优化使AI算力翻倍、芯片尺寸缩小30%、能效提升40%；支持1-bit AI模型（数据中心多用16–32位、传统端侧AI多用4–8位）；用户日均调用AI智能体在智能手机上约2次、在智能眼镜上最高达13次。与电子元器件供应链的关系在于：这些平台定义了可穿戴/音频/AI眼镜等新形态终端的SoC、微NPU、Wi-Fi 6E连接、存储（DRAM）与低功耗芯片需求方向，可能影响相关芯片、存储与低功耗器件的需求结构。
- 高通已将其骁龙平台从智能手机扩展为六大平台家族：面向智能手机的骁龙8系、面向PC的骁龙X、面向智能手表与可穿戴的骁龙Wear、面向智能眼镜的骁龙AR、面向扩展现实（XR）设备的骁龙Reality，以及面向无线耳机与音频眼镜的骁龙Sound。
- Ziad Asghar称高通拥有从毫瓦级骁龙Sound到千瓦级数据中心产品的独特技术谱系，这一广度使其能更高效地开发超低功耗高性能技术，并在多产品间自由扩展以形成协同。
- 第二代骁龙Sound Elite是高通首个面向个人AI时代的音频平台，面向AI pin、无线耳机和音频眼镜等智能音频可穿戴设备；其采用更小的微NPU而非大型Hexagon NPU，并显著改进微NPU的IP模块与内部架构。
- 微NPU架构优化使AI算力较上一代翻倍，芯片尺寸缩小30%，能效提升40%；超低功耗Wi-Fi 6E支持设备不经智能手机直接与外部网络服务器或云端通信，即“直连云端”，从而支持独立的AI pin、AI吊坠和音频眼镜。
- 第二代骁龙Sound Elite与骁龙AR1、AR1+支持1-bit AI模型架构；比特数代表AI模型精度，数据中心主要使用16–32位模型，传统端侧AI通常使用4–8位模型；降低比特数可显著减少模型文件大小与内存（DRAM）需求，计算从复杂浮点乘法转向加减等更简单运算，提升处理速度并降低功耗，代价是误差可能性更大，高通与PrismML合作以尽量减少精度损失。
- Asghar称10秒的AI响应时间会赶走消费者；对于内存受限的可穿戴设备，量化1-bit AI模型可直接在设备上运行，即时处理实时命令与控制、视障用户环境感知以及实时语音转写；智能眼镜是AI智能体无需用户输入即可理解并交互周围环境与上下文的最佳形态，用户在智能手机上日均调用AI智能体约2次，在智能眼镜上最高达13次。
供应链影响
- 若个人AI设备（AI pin、音频眼镜、智能眼镜等）按原文所述形成多设备并存的趋势，可能带动对超低功耗SoC、微NPU、Wi-Fi 6E连接芯片及小型化封装的需求结构变化，具体取决于终端出货节奏与消费者采纳程度。
- 原文称1-bit AI模型可显著降低模型文件大小与DRAM需求，这可能影响可穿戴设备中存储器件（DRAM）的容量配置方向，但实际影响取决于模型量化在终端侧的普及程度与厂商设计选择。
- 芯片尺寸缩小30%与能效提升40%若在量产中实现，可能改变音频可穿戴与AI眼镜对芯片面积、功耗与散热相关元器件的规格要求，具体取决于代工与封装良率。
- 高通将感知算法（头部追踪、眼动追踪、手势识别）直接集成进芯片以超低功耗运行，可能减少部分外置传感器或协处理器的需求，也可能改变传感器与芯片的配套方式，取决于终端厂商的架构设计。
- 原文以2011年智能手机平台迁移类比当前个人AI形态早期阶段，若该判断成立，可能意味着相关芯片与元器件供应链将围绕新形态终端逐步重构，但时间与规模取决于AI智能体体验成熟度与生态投入。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
高通高级副总裁Ziad Asghar讲解第二代骁龙Sound Elite。（照片：THE ELEC，毛伊岛）
高通已将其骁龙平台产品线从智能手机扩展至面向下一代个性化（个人）人工智能（AI）设备。该公司正寻求在由AI智能体驱动的新兴设备市场中占据领先地位。
高通XR、可穿戴与个人AI业务高级副总裁兼总经理Ziad Asghar当地时间9月24日在毛伊岛举行的Snapdragon Summit 2026上表示：“高通拥有独特的技术谱系，涵盖从毫瓦级骁龙Sound到千瓦级数据中心产品的全部范围。”他补充说：“这一广泛的技术谱系使我们能够更高效、更强大地开发超低功耗、高性能技术，并在多个产品之间自由扩展以创造协同效应。”
从面向智能手机的骁龙8系列开始，高通目前拥有六个骁龙平台家族：面向PC的骁龙X、面向智能手表与可穿戴设备的骁龙Wear、面向智能眼镜的骁龙AR、面向扩展现实（XR）设备的骁龙Reality，以及面向无线耳机与音频眼镜的骁龙Sound。
◇ 智能体AI最大化新形态设备的潜力
高通之所以将骁龙平台多元化，是因为“数字体验”正因智能体AI而发生变化。过去，智能手机凭借便携性与可用性的结合，是能够提供完整数字体验的主要设备。包括智能手表在内的可穿戴产品在很大程度上扮演辅助角色，或仅扩展智能手机的部分功能。
在智能体AI时代，个人AI理解用户意图与上下文，并自动提供定制化服务。这使针对单个产品形态量身定制的差异化AI智能体体验成为可能。用户即使没有像智能手机那样支持电容触控的显示屏，也能获得数字体验。
Asghar说：“我们已经走过键盘鼠标和触摸屏的时代，AI智能体将成为下一代计算体验与交互界面的核心。”“一个新的形态设备时代正在出现，用户在移动中无需操作屏幕，而是通过语音和视觉上下文进行交互。”高通据此开发了针对每种形态优化的骁龙平台。
高通认为当今智能眼镜市场仍处于早期阶段。该公司认为这与2011年的智能手机市场相似，当时行业正逐步从Windows CE、Palm OS和Symbian等较旧的移动操作系统转向Android。
Asghar说：“当时早早投资新移动平台的公司，今天正引领市场。”“现在投资个人AI形态设备的公司，将成为下一代大型科技领导者。”
◇ 第二代骁龙Sound Elite升级微NPU架构
高通高级副总裁Ziad Asghar展示第二代骁龙Sound Elite。（照片：THE ELEC，毛伊岛）
高通在Snapdragon Summit 2026期间发布了第二代骁龙Sound Elite。这是该公司首个为个人AI时代设计的高端音频平台。该片上系统（SoC）面向智能音频可穿戴设备（即听觉设备，hearables），包括AI pin、无线耳机和音频眼镜。
Asghar说：“骁龙Sound Elite平台没有采用大型Hexagon神经处理单元（NPU），而是使用更小的微NPU。”“特别是，第二代产品显著改进了微NPU的知识产权（IP）模块和内部架构。”
微NPU的架构优化使AI计算性能较上一代翻倍。芯片尺寸缩小了30%，而能效提升了40%。更小的芯片和更低的功耗要求，使在耳机和眼镜等紧凑产品中实现个人AI能力变得更加容易。
超低功耗Wi-Fi 6E允许设备不经智能手机直接与外部网络服务器或云端通信。“直连云端”技术使独立的AI pin、AI吊坠和音频眼镜成为可能。
第二代骁龙Sound Elite以及骁龙AR1和AR1+采用支持1-bit AI模型的架构。比特数代表AI模型的精度。数据中心主要使用16至32位模型，而传统端侧AI通常使用4至8位模型。
降低比特数可显著削减AI模型文件大小和内存（即DRAM）需求。计算也从复杂的浮点乘法转向加减法等更简单的运算，从而提高处理速度并降低功耗。代价是误差可能性更大。高通与PrismML合作以尽量减少精度损失。
Asghar说：“10秒的AI响应时间会让消费者转身离开。”“对于内存受限的可穿戴设备，量化的1-bit AI模型可以直接在设备上运行，即时处理实时命令与控制、面向视障用户的环境感知以及实时语音转写。”
◇ AR1和AR2面向能理解用户注视与上下文的AI眼镜
高通高级副总裁Ziad Asghar回答韩国记者提问。（照片：THE ELEC，毛伊岛）
智能眼镜目前是高通整个骁龙产品组合中优先级最高的领域之一。这就是该公司为该形态准备了第二代骁龙Sound Elite、AR1、AR1+和AR2系列的原因。
Asghar说：“智能眼镜是最适合AI智能体在无需用户输入的情况下理解并与周围环境和上下文交互的形态。”“用户在智能手机上平均每天调用AI智能体两次，但在智能眼镜上这一频率最高可升至每天13次。”
第二代骁龙Sound Elite专注于音频，而骁龙AR系列则以通过摄像头和其他组件实现的视觉感知为中心。AR1系列支持从无显示屏的智能眼镜（如三星电子的智能眼镜）到带单目或双目显示屏的完全沉浸式增强现实（AR）眼镜等产品。
Asghar说：“仅仅把智能手机技术搬进智能眼镜是不够的。功耗、性能和用户体验必须针对该形态进行优化和彻底重新设计。”“高通已将经过10多年研发的头部追踪、眼动追踪和手势识别等感知算法直接实现在其芯片中，使其能够以超低功耗运行。”
◇ 个人AI设备将与智能手机共存，扩大市场
高通并不预期包括智能眼镜在内的个人AI设备的兴起会削弱智能手机的作用。Asghar以PC为例。
当智能手机刚出现时，一些人预测PC时代将终结，但这两类产品划分了各自角色并继续共存。高通预计智能手机与智能眼镜之间也会是同样的关系。
Asghar说：“我们经常用智能手机处理简单邮件，但在写长邮件时仍会回到PC。”“在个人AI设备时代，每种设备将承担最适合自己的角色，它们将共存。”
他补充说：“人们通常每人拥有一部智能手机，但在个人AI时代，一个人将同时使用智能手表、智能眼镜、无线耳机和AI吊坠等多台设备。”“这正是个人AI设备市场取得重大成功的潜力所在。”
Ziad Asghar, Qualcomm senior vice president, explains the second-generation Snapdragon Sound Elite. (Photo: THE ELEC, Maui)
Qualcomm has built out its Snapdragon platform lineup beyond smartphones to address next-generation personalized, or personal, artificial intelligence (AI) devices. The company is seeking to take the lead in emerging device markets driven by AI agents.
Ziad Asghar, senior vice president and general manager of XR, Wearables and Personal AI at Qualcomm, said at Snapdragon Summit 2026 in Maui on Sept. 24 local time, "Qualcomm has a unique technology spectrum spanning everything from milliwatt-level Snapdragon Sound to kilowatt-level data center products." He added, "This broad technology spectrum allows us to develop ultra-low-power, high-performance technologies much more efficiently and powerfully, and freely scale them across multiple products to create synergies."
Starting with the Snapdragon 8 series for smartphones, Qualcomm now has six Snapdragon platform families: Snapdragon X for PCs, Snapdragon Wear for smartwatches and wearables, Snapdragon AR for smart glasses, Snapdragon Reality for extended reality (XR) devices, and Snapdragon Sound for wireless earbuds and audio glasses.
◇ Agentic AI Maximizes the Potential of New Form Factors
Qualcomm has diversified its Snapdragon platforms because of changes in the "digital experience" being driven by agentic AI. In the past, smartphones, with their combination of portability and usability, were the primary devices capable of delivering a full digital experience. Wearable products, including smartwatches, largely played a supporting role or extended only some smartphone functions.
In the agentic AI era, personal AI understands a user's intent and context and automatically provides customized services. This enables differentiated AI agent experiences tailored to individual product form factors. Users can access digital experiences even without a display supporting capacitive touch like a smartphone.
"We have moved through the eras of keyboards and mice and touchscreens, and AI agents will be at the heart of the next-generation computing experience and interface," Asghar said. "A new era of form factors is emerging in which users interact through voice and visual context without operating a screen while on the move." Qualcomm has accordingly developed Snapdragon platforms optimized for each form factor.
Qualcomm views today's smart glasses market as still being in its early stages. The company sees similarities with the smartphone market in 2011, when the industry was gradually shifting from older mobile operating systems such as Windows CE, Palm OS and Symbian to Android.
"Companies that invested early in new mobile platforms at the time are leading the market today," Asghar said. "Companies investing in personal AI form factors now will become the next generation of Big Tech leaders."
◇ Second-Generation Snapdragon Sound Elite Upgrades Micro NPU Architecture
Ziad Asghar, Qualcomm senior vice president, presents the second-generation Snapdragon Sound Elite. (Photo: THE ELEC, Maui)
Qualcomm unveiled the second-generation Snapdragon Sound Elite during Snapdragon Summit 2026. It is the company's first premium audio platform designed for the personal AI era. The system-on-chip (SoC) is designed for intelligent audio wearables, or hearables, including AI pins, wireless earbuds and audio glasses.
"Instead of the large Hexagon neural processing unit (NPU), the Snapdragon Sound Elite platform uses a smaller micro NPU," Asghar said. "In particular, the second-generation product significantly improves the micro NPU's intellectual property (IP) blocks and internal architecture."
Architectural optimization of the micro NPU has doubled AI computing performance compared with the previous generation. Chip size has been reduced by 30%, while power efficiency has improved by 40%. The smaller chip and lower power requirements make it easier to implement personal AI capabilities in compact products such as earbuds and glasses.
Ultra-low-power Wi-Fi 6E allows devices to communicate directly with external network servers, or the cloud, without going through a smartphone. The "direct-to-cloud" technology enables standalone AI pins, AI pendants and audio glasses.
The second-generation Snapdragon Sound Elite and Snapdragon AR1 and AR1+ feature architectures supporting 1-bit AI models. The bit count represents the precision of an AI model. Data centers primarily use 16- to 32-bit models, while conventional on-device AI typically uses 4- to 8-bit models.
Reducing the number of bits significantly cuts AI model file sizes and memory, or DRAM, requirements. Computation also shifts from complex floating-point multiplication to simpler operations such as addition and subtraction, increasing processing speed and reducing power consumption. The trade-off is a greater potential for errors. Qualcomm worked with PrismML to minimize accuracy loss.
"AI response times of 10 seconds will turn consumers away," Asghar said. "For memory-constrained wearables, quantized 1-bit AI models can run directly on the device to instantly handle real-time command and control, environmental awareness for visually impaired users, and real-time voice transcription."
◇ AR1 and AR2 Target AI Glasses That Understand Users' Gaze and Context
Ziad Asghar, Qualcomm senior vice president, answers questions from Korean reporters. (Photo: THE ELEC, Maui)
Smart glasses are currently one of Qualcomm's highest-priority areas across its Snapdragon portfolio. That is why the company has prepared the second-generation Snapdragon Sound Elite, AR1, AR1+ and AR2 series for the form factor.
"Smart glasses are the form factor best suited for AI agents to understand and interact with the surrounding environment and context without user input," Asghar said. "Users invoke AI agents an average of twice a day on smartphones, but that frequency rises to as many as 13 times a day on smart glasses."
While the second-generation Snapdragon Sound Elite focuses on audio, the Snapdragon AR series centers on visual perception through cameras and other components. The AR1 series supports products ranging from smart glasses without displays, such as Samsung Electronics' smart glasses, to fully immersive augmented reality (AR) glasses with monocular or binocular displays.
"Simply bringing smartphone technology into smart glasses is not enough. Power, performance and user experience have to be optimized and completely redesigned for the form factor," Asghar said. "Qualcomm has implemented perception algorithms such as head tracking, eye tracking and gesture recognition, developed through more than 10 years of research and development (R&D), directly into its chips so they can operate at ultra-low power."
◇ Personal AI Devices to Coexist With Smartphones, Expanding the Market
Qualcomm does not expect the rise of personal AI devices, including smart glasses, to diminish the role of smartphones. Asghar cited PCs as an example.
When smartphones first emerged, some predicted the end of the PC era, but the two products have divided their roles and continue to coexist. Qualcomm expects the same relationship between smartphones and smart glasses.
"We often use smartphones for simple emails, but we still return to PCs when writing long emails," Asghar said. "In the personal AI device era, each device will take on the roles it is best suited for, and they will coexist."
"People typically own one smartphone each, but in the personal AI era, a single person will simultaneously use multiple devices such as a smartwatch, smart glasses, wireless earbuds and an AI pendant," he added. "That is where the potential for major success in the personal AI device market lies."