[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"company-c-148":3},{"queryId":4,"companyId":5,"companySlug":6,"companyName":7,"summary":8,"tags":9,"keyData":10,"markdownContent":11,"lifecycleStage":12,"sections":13,"sources":61,"status":17,"updatedAt":68},"2037100642908217346","2037101192915689474","c-148","Ambient Scientific, Inc.","Ambient Scientific, Inc. 是一家成立于2018年、总部位于美国加利福尼亚州圣何塞\u002F圣克拉拉的私营半导体公司。公司专注于开发基于其专有DigAn®架构的超低功耗、模拟-数字混合人工智能处理器，旨在为边缘设备提供端侧训练和推理能力。公司定位为一家技术驱动的“软件定义”AI芯片设计公司。","[\"Ambient Scientific, Inc.\",\"成长期\"]","{\"email\": \"info@ambient.science; support@ambient.science\", \"phone\": \"信息不足\", \"address\": \"2350 Mission College Blvd, Suite 105, Santa Clara, CA 95054 (Principal Executive Office)\", \"revenue\": \"信息不足\", \"website\": \"https:\u002F\u002Fwww.ambient.science\u002F\", \"legal_name\": \"Ambient Scientific, Inc.\", \"duns_number\": \"信息不足\", \"founded_year\": \"2018\", \"headquarters\": \"美国加利福尼亚州圣何塞\u002F圣克拉拉\", \"listing_info\": \"未上市\", \"core_business\": \"设计并销售基于专有DigAn®架构的超低功耗、模拟-数字混合人工智能处理器，专注于边缘设备端侧训练与推理。\", \"target_market\": \"全球市场，重点为医疗可穿戴设备、工业传感器、智能家居基础设施等边缘AI应用领域。\", \"employee_count\": \"51-200人\", \"identity_number\": \"C4142105 (California File Number)\"}","# Ambient Scientific, Inc.\n\n## 1. 核心身份标签\n\nAmbient Scientific, Inc. 是一家成立于2018年、总部位于美国加利福尼亚州圣何塞\u002F圣克拉拉的私营半导体公司。公司专注于开发基于其专有DigAn®架构的超低功耗、模拟-数字混合人工智能处理器，旨在为边缘设备提供端侧训练和推理能力。公司定位为一家技术驱动的“软件定义”AI芯片设计公司。\n\n## 2. 战略愿景与使命\n\n根据公开信息，公司的战略愿景是成为边缘人工智能计算领域的领导者，通过其创新的模拟-数字混合计算架构，突破传统数字AI处理器的功耗和性能瓶颈。其使命是赋能下一代“始终在线”的智能设备，使AI能够在极低功耗下直接在终端设备上运行和学习，减少对云端计算的依赖。\n\n## 3. 业务架构与产品矩阵\n\n公司业务围绕其专有技术栈展开，涵盖硬件、软件和开发者生态。\n1.  **核心硬件产品**：\n    *   **GPX-10 AI微处理器**：旗舰产品，基于TSMC 7nm工艺制造，宣称是全球功耗最低的可编程AI处理器，支持低于100μW的“始终在线”功耗，峰值性能达512 GOPs。该芯片支持端侧训练和推理，目前处于向选定合作伙伴提供样片阶段。\n    *   **DigAn®架构**：公司所有处理器的技术基础，是一种可扩展的软件定义AI核心架构，集成了混合信号电路与可重构数字逻辑。支持从5到8，000个核心的动态扩展，并兼容40nm至5nm工艺节点。\n2.  **软件与开发生态**：\n    *   **Ambient SDK**：核心软件开发工具包，最新版本为v2.4（2026年1月发布），包含首个公开文档化的端侧微调工具包，支持在GPX-10上直接进行迁移学习。提供PyTorch兼容前端、自动模拟感知量化和运行时内存分析器。\n    *   **开发者门户**：提供技术文档、SDK访问和评估板申请入口。\n    *   **评估套件**：售价499美元，包含JTAG调试器、传感器子板和1年SDK许可。\n3.  **技术合作与标准参与**：公司于2024年第二季度以贡献成员身份加入RISC-V International，参与RISC-V AI\u002FML和模拟扩展工作组，其DigAn®架构使用RISC-V RV32IMC作为控制平面。\n\n## 4. 核心能力与护城河\n\n1.  **模拟-数字混合计算技术**：其DigAn®架构及GPX-10芯片的核心创新，利用电压模式模拟乘累加单元、动态精度缩放和数字控制逻辑进行纠错与校准，据称能效比领先主流数字NPU 8倍以上。\n2.  **超低功耗设计**：实现低于100μW的“始终在线”操作，并采用浮栅非易失性模拟存储权重，实现零启动延迟和即时唤醒。\n3.  **端侧训练能力**：区别于多数仅支持推理的边缘AI芯片，GPX-10支持直接在设备上进行微调和在线学习，无需依赖主机PC或云端。\n4.  **软件可编程性**：尽管采用模拟计算，但仍通过SDK提供C\u002FC++和Python的完全软件可编程性，降低了开发门槛。\n5.  **知识产权**：已申请核心专利（如US20240342092A1），覆盖其DigAn®架构和训练-推理协同执行方法。\n\n## 5. 目标市场与客户\n\n公司瞄准对功耗和实时性要求极高的边缘AI应用市场。\n*   **目标市场**：主要包括医疗可穿戴设备、工业传感器和智能家居基础设施。\n*   **客户情况**：未公开具体客户名称，但多个来源提及已获得“全球一级医疗设备OEM”和“智能家居基础设施提供商”的设计订单。在Embedded World 2025的演示中，也展示了面向可穿戴传感器的跌倒检测应用。\n\n## 6. 品牌形象与价值观\n\n公司通过技术发布会、行业展会（如Embedded World）和行业媒体（如EE Times）报道，塑造了“前沿技术突破者”和“模拟AI复兴领导者”的品牌形象。其价值观强调通过底层硬件创新解决实际能效问题，推动AI普惠至资源受限的边缘设备。公司注重开发者生态建设，但未公开其GitHub或开源仓库，技术细节和完整生产工具链的获取需要签署保密协议。\n\n## 7. 发展生命周期\n\n公司处于成长期。\n*   **成立与研发**：公司于2018年4月12日正式注册成立，但自称研发活动始于2016年。\n*   **融资历程**：已完成三轮融资，总额6800万美元（种子轮600万\u002F2019年，A轮2000万\u002F2021年，B轮4200万\u002F2023年），投资方包括DCVC、Eclipse Ventures、SineWave Ventures等。\n*   **产品状态**：旗舰产品GPX-10已发布并处于客户送样阶段。公司正在招聘人员以推进下一代GPX-20的研发（提及Q3 2026有流片计划）。\n*   **估值**：据PitchBook估计，B轮后估值约为4.2亿美元（未确认）。\n*   **上市状态**：私营公司，CEO确认暂无IPO计划。\n\n## 8. 潜在挑战与风险\n\n1.  **市场竞争**：边缘AI芯片市场竞争激烈，需面对来自传统半导体巨头和众多初创公司的挑战。\n2.  **技术商业化**：模拟计算在精度、可编程性和量产一致性方面存在固有挑战，需要强大的软件工具链和生态支持来吸引开发者并实现大规模商用。\n3.  **客户依赖**：目前披露的客户集中在少数高端领域（如医疗），客户集中度可能较高。\n4.  **财务不透明**：作为私营公司，营收、利润等关键财务数据未披露，难以评估其商业化和盈利能力。\n5.  **供应链风险**：依赖TSMC等先进晶圆代工厂，可能受全球半导体供应链波动影响。\n\n## 9. 综述\n\nAmbient Scientific是一家在超低功耗边缘AI芯片领域具有显著技术特色的初创公司。其核心优势在于将模拟计算的高能效与数字可编程性相结合，并率先支持端侧训练，在特定细分市场（如医疗可穿戴设备）形成了差异化竞争力。公司已获得知名风投支持，产品进入客户验证阶段。然而，其长期成功将取决于能否将技术优势转化为广泛的商业应用，构建强大的开发者生态，并在激烈的市场竞争中持续保持领先。目前公司仍处于“技术验证”向“市场扩张”过渡的关键阶段。\n\n## 结构化关键数据\n\n\u003C!--KEY_DATA-->\n{\"email\": \"info@ambient.science; support@ambient.science\", \"phone\": \"信息不足\", \"address\": \"2350 Mission College Blvd, Suite 105, Santa Clara, CA 95054 (Principal Executive Office)\", \"revenue\": \"信息不足\", \"website\": \"https:\u002F\u002Fwww.ambient.science\u002F\", \"legal_name\": \"Ambient Scientific, Inc.\", \"duns_number\": \"信息不足\", \"founded_year\": \"2018\", \"headquarters\": \"美国加利福尼亚州圣何塞\u002F圣克拉拉\", \"listing_info\": \"未上市\", \"core_business\": \"设计并销售基于专有DigAn®架构的超低功耗、模拟-数字混合人工智能处理器，专注于边缘设备端侧训练与推理。\", \"target_market\": \"全球市场，重点为医疗可穿戴设备、工业传感器、智能家居基础设施等边缘AI应用领域。\", \"employee_count\": \"51-200人\", \"identity_number\": \"C4142105 (California File Number)\"}","成长期",[14,21,26,31,36,41,46,51,56],{"id":15,"profileId":16,"sectionNo":17,"sectionName":18,"content":8,"sortOrder":17,"identityId":19,"createdAt":20},2037101188264206300,2037101188054491100,1,"核心身份标签",null,"2026-03-26T09:35:45.000+00:00",{"id":22,"profileId":16,"sectionNo":23,"sectionName":24,"content":25,"sortOrder":23,"identityId":19,"createdAt":20},2037101188473921500,2,"战略愿景与使命","根据公开信息，公司的战略愿景是成为边缘人工智能计算领域的领导者，通过其创新的模拟-数字混合计算架构，突破传统数字AI处理器的功耗和性能瓶颈。其使命是赋能下一代“始终在线”的智能设备，使AI能够在极低功耗下直接在终端设备上运行和学习，减少对云端计算的依赖。",{"id":27,"profileId":16,"sectionNo":28,"sectionName":29,"content":30,"sortOrder":28,"identityId":19,"createdAt":20},2037101188679442400,3,"业务架构与产品矩阵","公司业务围绕其专有技术栈展开，涵盖硬件、软件和开发者生态。\n1.  **核心硬件产品**：\n    *   **GPX-10 AI微处理器**：旗舰产品，基于TSMC 7nm工艺制造，宣称是全球功耗最低的可编程AI处理器，支持低于100μW的“始终在线”功耗，峰值性能达512 GOPs。该芯片支持端侧训练和推理，目前处于向选定合作伙伴提供样片阶段。\n    *   **DigAn®架构**：公司所有处理器的技术基础，是一种可扩展的软件定义AI核心架构，集成了混合信号电路与可重构数字逻辑。支持从5到8，000个核心的动态扩展，并兼容40nm至5nm工艺节点。\n2.  **软件与开发生态**：\n    *   **Ambient SDK**：核心软件开发工具包，最新版本为v2.4（2026年1月发布），包含首个公开文档化的端侧微调工具包，支持在GPX-10上直接进行迁移学习。提供PyTorch兼容前端、自动模拟感知量化和运行时内存分析器。\n    *   **开发者门户**：提供技术文档、SDK访问和评估板申请入口。\n    *   **评估套件**：售价499美元，包含JTAG调试器、传感器子板和1年SDK许可。\n3.  **技术合作与标准参与**：公司于2024年第二季度以贡献成员身份加入RISC-V International，参与RISC-V AI\u002FML和模拟扩展工作组，其DigAn®架构使用RISC-V RV32IMC作为控制平面。",{"id":32,"profileId":16,"sectionNo":33,"sectionName":34,"content":35,"sortOrder":33,"identityId":19,"createdAt":20},2037101188889157600,4,"核心能力与护城河","1.  **模拟-数字混合计算技术**：其DigAn®架构及GPX-10芯片的核心创新，利用电压模式模拟乘累加单元、动态精度缩放和数字控制逻辑进行纠错与校准，据称能效比领先主流数字NPU 8倍以上。\n2.  **超低功耗设计**：实现低于100μW的“始终在线”操作，并采用浮栅非易失性模拟存储权重，实现零启动延迟和即时唤醒。\n3.  **端侧训练能力**：区别于多数仅支持推理的边缘AI芯片，GPX-10支持直接在设备上进行微调和在线学习，无需依赖主机PC或云端。\n4.  **软件可编程性**：尽管采用模拟计算，但仍通过SDK提供C\u002FC++和Python的完全软件可编程性，降低了开发门槛。\n5.  **知识产权**：已申请核心专利（如US20240342092A1），覆盖其DigAn®架构和训练-推理协同执行方法。",{"id":37,"profileId":16,"sectionNo":38,"sectionName":39,"content":40,"sortOrder":38,"identityId":19,"createdAt":20},2037101189094678500,5,"目标市场与客户","公司瞄准对功耗和实时性要求极高的边缘AI应用市场。\n*   **目标市场**：主要包括医疗可穿戴设备、工业传感器和智能家居基础设施。\n*   **客户情况**：未公开具体客户名称，但多个来源提及已获得“全球一级医疗设备OEM”和“智能家居基础设施提供商”的设计订单。在Embedded World 2025的演示中，也展示了面向可穿戴传感器的跌倒检测应用。",{"id":42,"profileId":16,"sectionNo":43,"sectionName":44,"content":45,"sortOrder":43,"identityId":19,"createdAt":20},2037101189304393700,6,"品牌形象与价值观","公司通过技术发布会、行业展会（如Embedded World）和行业媒体（如EE Times）报道，塑造了“前沿技术突破者”和“模拟AI复兴领导者”的品牌形象。其价值观强调通过底层硬件创新解决实际能效问题，推动AI普惠至资源受限的边缘设备。公司注重开发者生态建设，但未公开其GitHub或开源仓库，技术细节和完整生产工具链的获取需要签署保密协议。",{"id":47,"profileId":16,"sectionNo":48,"sectionName":49,"content":50,"sortOrder":48,"identityId":19,"createdAt":20},2037101189509914600,7,"发展生命周期","公司处于成长期。\n*   **成立与研发**：公司于2018年4月12日正式注册成立，但自称研发活动始于2016年。\n*   **融资历程**：已完成三轮融资，总额6800万美元（种子轮600万\u002F2019年，A轮2000万\u002F2021年，B轮4200万\u002F2023年），投资方包括DCVC、Eclipse Ventures、SineWave Ventures等。\n*   **产品状态**：旗舰产品GPX-10已发布并处于客户送样阶段。公司正在招聘人员以推进下一代GPX-20的研发（提及Q3 2026有流片计划）。\n*   **估值**：据PitchBook估计，B轮后估值约为4.2亿美元（未确认）。\n*   **上市状态**：私营公司，CEO确认暂无IPO计划。",{"id":52,"profileId":16,"sectionNo":53,"sectionName":54,"content":55,"sortOrder":53,"identityId":19,"createdAt":20},2037101189715435500,8,"潜在挑战与风险","1.  **市场竞争**：边缘AI芯片市场竞争激烈，需面对来自传统半导体巨头和众多初创公司的挑战。\n2.  **技术商业化**：模拟计算在精度、可编程性和量产一致性方面存在固有挑战，需要强大的软件工具链和生态支持来吸引开发者并实现大规模商用。\n3.  **客户依赖**：目前披露的客户集中在少数高端领域（如医疗），客户集中度可能较高。\n4.  **财务不透明**：作为私营公司，营收、利润等关键财务数据未披露，难以评估其商业化和盈利能力。\n5.  **供应链风险**：依赖TSMC等先进晶圆代工厂，可能受全球半导体供应链波动影响。",{"id":57,"profileId":16,"sectionNo":58,"sectionName":59,"content":60,"sortOrder":58,"identityId":19,"createdAt":20},2037101189925150700,9,"综述","Ambient Scientific是一家在超低功耗边缘AI芯片领域具有显著技术特色的初创公司。其核心优势在于将模拟计算的高能效与数字可编程性相结合，并率先支持端侧训练，在特定细分市场（如医疗可穿戴设备）形成了差异化竞争力。公司已获得知名风投支持，产品进入客户验证阶段。然而，其长期成功将取决于能否将技术优势转化为广泛的商业应用，构建强大的开发者生态，并在激烈的市场竞争中持续保持领先。目前公司仍处于“技术验证”向“市场扩张”过渡的关键阶段。",[62,69,74,79,84,89,94,99,104,109,114,119,124],{"id":63,"profileId":16,"sectionNo":17,"title":64,"url":65,"sourceType":66,"publishTime":19,"credibility":38,"snippet":67,"createdAt":68},2037101190130671600,"Ambient Scientific, Inc. – California Secretary of State Business Entity Detail","https:\u002F\u002Fbizfileonline.sos.ca.gov\u002Fsearch\u002Fbusiness",0,"Ambient Scientific, Inc. is a California corporation filed on April 12, 2018 (File Number C4142105). Status: Active. Registered Agent: Corporation Service Company (CSC), 2710 Gateway Oaks Dr., Suite 150N, Sacramento, CA 95833. Principal Executive Office: 2350 Mission College Blvd, Suite 105, Santa Clara, CA 95054. Officers: Gajendra Prasad Singh (President\u002FCEO), Anil K. Jain (Secretary\u002FTreasurer). No delinquencies or suspensions reported. Annual Statement due March 15, 2026 (confirmed filed). Entity type: Stock Corporation; Authorized shares: 10,000,000 common.","2026-03-26T09:35:46.000+00:00",{"id":70,"profileId":16,"sectionNo":23,"title":71,"url":72,"sourceType":17,"publishTime":19,"credibility":38,"snippet":73,"createdAt":68},2037101190340386800,"Ambient Scientific Developer Portal Launches SDK v2.4 with On-Device Fine-Tuning Toolkit","https:\u002F\u002Fdev.ambient.science\u002Fblog\u002Fsdk-v2-4-release","Ambient Scientific released SDK v2.4 (January 2026) featuring its first publicly documented on-device fine-tuning toolkit — enabling transfer learning directly on GPX-10 without host PC or cloud dependency. Includes PyTorch-compatible frontend, automatic analog-aware quantization, and runtime memory profiler. Requires NDA for full documentation and production toolchain access. Evaluation kits ($499) ship with JTAG debugger, sensor daughterboard, and 1-year SDK license. Support email: support@ambient.science; SLA: 3-business-day response for licensed partners.",{"id":75,"profileId":16,"sectionNo":28,"title":76,"url":77,"sourceType":66,"publishTime":19,"credibility":38,"snippet":78,"createdAt":68},2037101190554296300,"U.S. Patent US20240342092A1: Analog-Digital Hybrid Processor for Low-Power Neural Network Execution","https:\u002F\u002Fpatents.google.com\u002Fpatent\u002FUS20240342092A1","Filed by Ambient Scientific, Inc. (2023-04-18), granted priority to provisional application 63\u002F398,201 (2022-07-15). Describes core innovations of DigAn®: voltage-mode analog MAC units, dynamic precision scaling, and digital control logic for error correction and calibration. Assignee: Ambient Scientific, Inc., San Jose, CA. Inventors include G. P. Singh and M. Chen. Claims cover GPX-10’s architecture and training-inference co-execution methodology. No licensing terms disclosed; patent status: Published (not yet granted as of 2026-03-26).",{"id":80,"profileId":16,"sectionNo":33,"title":81,"url":82,"sourceType":17,"publishTime":19,"credibility":38,"snippet":83,"createdAt":68},2037101190759817200,"Ambient Scientific Unveils DigAn® Architecture: Scalable, Software-Defined AI Cores from 5 to 8,000 Units","https:\u002F\u002Fwww.ambient.science\u002Ftechnology\u002Fdigan","Ambient Scientific’s proprietary DigAn® (Digital-Analog Neural) architecture forms the foundation of all its processors. It integrates mixed-signal circuitry with reconfigurable digital logic, enabling ultra-low-power analog AI computation while retaining full software programmability in C\u002FC++ and Python. DigAn supports dynamic scaling across process nodes (40nm to 5nm), and each core is independently configurable for inference, fine-tuning, or online learning. Documentation, SDK access, and evaluation board requests are available via developer portal (dev.ambient.science). No public GitHub or open-source repositories disclosed.",{"id":85,"profileId":16,"sectionNo":38,"title":86,"url":87,"sourceType":17,"publishTime":19,"credibility":38,"snippet":88,"createdAt":68},2037101190969532400,"Ambient Scientific Announces GPX-10: World’s Lowest-Power Programmable AI Processor for Edge Training & Inference","https:\u002F\u002Fwww.ambient.science\u002Fpress\u002Fgpx-10-launch","Ambient Scientific, Inc. officially launched its GPX-10 AI microprocessor — a fully programmable, analog-digital hybrid chip enabling on-device training and inference at sub-100μW 'always-on' power. The GPX-10 delivers up to 512 GOPs peak performance in a compact package, fabricated on TSMC’s 7nm process. It supports custom neural networks via Ambient’s proprietary DigAn® architecture and software stack. The company states GPX-10 is sampling to select partners; no public pricing or distributor list is published. Contact: info@ambient.science. Headquarters: San Jose, CA. Founded April 2018 (operational since 2016 R&D).",{"id":90,"profileId":16,"sectionNo":43,"title":91,"url":92,"sourceType":33,"publishTime":19,"credibility":33,"snippet":93,"createdAt":68},2037101191212802000,"Ambient Scientific — LinkedIn Company Profile","https:\u002F\u002Fwww.linkedin.com\u002Fcompany\u002Fambient-scientific","Ambient Scientific’s official LinkedIn (verified) lists 51–200 employees, founded 2018, industry: Semiconductors. Location: San Jose, CA. Highlights include ‘GPX-10 sampling now’, ‘DigAn® architecture deployed in medical wearables & industrial sensors’, and ‘hiring for compiler engineers, analog IC designers, and AI firmware developers’. No investor logos or financial metrics shown. Leadership section confirms Gajendra Prasad Singh as CEO and co-founder; Anil Jain as VP Engineering. Careers page links to Greenhouse.io (jobs.ambient.science), listing 12 active roles (as of March 2026), all remote\u002Fhybrid eligible.",{"id":95,"profileId":16,"sectionNo":48,"title":96,"url":97,"sourceType":66,"publishTime":19,"credibility":33,"snippet":98,"createdAt":68},2037101191422517200,"Ambient Scientific, Inc. – Crunchbase Profile","https:\u002F\u002Fwww.crunchbase.com\u002Forganization\u002Fambient-scientific","Crunchbase lists Ambient Scientific, Inc. as a private semiconductor company with total funding of $68M across three rounds (Seed: $6M in 2019; Series A: $20M in 2021; Series B: $42M in 2023). Investors include DCVC, Eclipse Ventures, SineWave Ventures, and undisclosed strategic partners. Key differentiators cited: analog-AI compute, on-device training capability, and sub-100μW always-on operation. No revenue figures, employee count beyond ‘51–200’, or customer names disclosed. Last funding update: November 2023. Data last verified by Crunchbase on 2026-02-28.",{"id":100,"profileId":16,"sectionNo":53,"title":101,"url":102,"sourceType":66,"publishTime":19,"credibility":33,"snippet":103,"createdAt":68},2037101191632232400,"Ambient Scientific Job Posting: Senior Analog IC Design Engineer (Santa Clara, CA)","https:\u002F\u002Fboards.greenhouse.io\u002Fambientscientific\u002Fjobs\u002F4278921004","Greenhouse-hosted job post (active as of 2026-03-20) seeks analog IC design engineer with expertise in 7nm\u002F5nm CMOS, current-mode DACs, and neural ADCs. Requires experience with Cadence Virtuoso, Spectre, and EM\u002FIR analysis. Mentions ‘tape-out scheduled for Q3 2026’ and ‘next-gen GPX-20 roadmap’. Reports to CTO Mei Lin Chen. Benefits include equity grants, health coverage, and flexible PTO. Application requires resume + cover letter explaining interest in ‘analog-first AI’. No public investor or customer references in posting.",{"id":105,"profileId":16,"sectionNo":58,"title":106,"url":107,"sourceType":66,"publishTime":19,"credibility":33,"snippet":108,"createdAt":68},2037101191841947600,"Ambient Scientific — PitchBook Company Profile","https:\u002F\u002Fpitchbook.com\u002Fprofiles\u002Fcompany\u002F11843385","PitchBook reports Ambient Scientific, Inc. as headquartered in Santa Clara, CA, with incorporation date April 2018. Valuation estimate: $420M (post-Series B, unconfirmed). Key tech focus areas: edge AI chips, analog computing, neuromorphic hardware. Customers cited (per unnamed sources): ‘global Tier-1 medical device OEM’ and ‘smart home infrastructure provider’. No revenue or EBITDA disclosed. Board composition includes G. P. Singh (CEO), H. R. Kim (DCVC Partner), and A. L. Tan (Eclipse Ventures). Data last updated: 2026-01-30.",{"id":110,"profileId":16,"sectionNo":17,"title":111,"url":112,"sourceType":23,"publishTime":19,"credibility":33,"snippet":113,"createdAt":68},2037101192072634400,"Ambient Scientific at Embedded World 2025: Live Demo of GPX-10 Running Real-Time Fall Detection on Coin Cell","https:\u002F\u002Fwww.embedded-world.de\u002Fen\u002Fnews\u002Fambient-scientific-gpx10-demo","At Embedded World Nuremberg (March 11–13, 2025), Ambient Scientific demonstrated GPX-10 running a quantized LSTM model for fall detection on wearable sensors — powered solely by a CR2032 coin cell for >6 months. The system achieved \u003C15ms latency, 98.2% accuracy (on MobiAct dataset), and consumed 83μW average during continuous sensing. Demo included integration with Ambience SDK v2.3 and sensor fusion API. No product SKUs or OEM partnerships named, but stated ‘design wins secured in Tier-1 medical device and smart home OEMs’.",{"id":115,"profileId":16,"sectionNo":23,"title":116,"url":117,"sourceType":23,"publishTime":19,"credibility":33,"snippet":118,"createdAt":68},2037101192286544000,"Ambient Scientific Featured in EE Times: ‘The Analog AI Comeback’","https:\u002F\u002Fwww.eetimes.com\u002Fambient-scientific-analog-ai-comeback","EE Times (June 2024) profiled Ambient Scientific as a leader in the analog AI resurgence, citing its ability to bypass von Neumann bottlenecks via in-memory compute. Interview with CTO Dr. Mei Lin Chen confirmed GPX-10’s analog cores operate at 0.4V supply and achieve 12.8 TOPS\u002FW — 8× higher than leading digital NPUs at sub-1mW. The article notes Ambient avoids SRAM-based weight storage, instead using floating-gate nonvolatile analog storage for weights — enabling instant wake-up and zero boot latency. No sales figures or customer names revealed.",{"id":120,"profileId":16,"sectionNo":28,"title":121,"url":122,"sourceType":66,"publishTime":19,"credibility":33,"snippet":123,"createdAt":68},2037101192496259000,"Ambient Scientific Joins RISC-V International as Contributing Member","https:\u002F\u002Friscv.org\u002Fmembers\u002Fambient-scientific-inc\u002F","Ambient Scientific, Inc. joined RISC-V International in Q2 2024 as a Contributing Member. The company contributes to the RISC-V AI\u002FML and Analog Extension working groups. Its DigAn® architecture uses RISC-V RV32IMC as control plane, while offloading neural compute to custom analog accelerators. Public technical whitepapers and ISA extension proposals are accessible to members only; no public spec sheets available. Membership confirms ongoing engagement with open standards but no open-core or open-hardware commitment disclosed.",{"id":125,"profileId":16,"sectionNo":33,"title":126,"url":127,"sourceType":23,"publishTime":19,"credibility":33,"snippet":128,"createdAt":129},2037101192705974300,"Ambient Scientific Raises $42M Series B to Scale GPX-10 Production and Expand AI-on-Edge Ecosystem","https:\u002F\u002Ftechcrunch.com\u002F2023\u002F11\u002F07\u002Fambient-scientific-series-b-ai-chip","Ambient Scientific, Inc. secured $42 million in Series B funding led by DCVC and joined by existing investors including Eclipse Ventures and SineWave Ventures. The round brings total disclosed funding to $68 million. Funds will accelerate GPX-10 silicon validation, expand tape-outs with TSMC, and grow its developer ecosystem—including tools, libraries, and hardware reference designs. CEO Gajendra Prasad Singh confirmed the company remains privately held with no IPO plans. No board member names or financial disclosures were released. Corporate address listed as 2350 Mission College Blvd, Suite 105, Santa Clara, CA 95054.","2026-03-26T09:35:47.000+00:00"]