Dyna Robotics 获1.2亿美元融资,加速机器人基础模型发展,迈向具身通用人工智能

Dyna Robotics 获1.2亿美元融资,加速机器人基础模型发展,迈向具身通用人工智能

2025-10-08Technology
--:--
--:--
马老师
早上好,小王。我是马老师,这是为你专属的 Goose Pod。今天是10月9日,星期四。
雷总
大家好,我是雷总。今天,我们要聊一个非常火热的话题:Dyna Robotics 获1.2亿美元融资,加速机器人基础模型发展,迈向具身通用人工智能。
雷总
我们直接开始吧!Dyna Robotics,这家公司在A轮融资中获得了1.2亿美元!这可不是一笔小数目。领投的有Robostrategy, CRV,还有First Round Capital,连英伟达、亚马逊、三星这些巨头都跟投了。
马老师
嗯,雷总,这就像一个武林高手,突然获得了深厚的内力,你懂的。这笔钱,就是他们的“内力”,要用来干嘛呢?就是要扩大团队,加速研发下一代的机器人“大脑”,也就是基础模型。
雷总
没错!说到大脑,就必须提他们的DYNA-1模型。这个模型非常厉害,能让机器人在连续24小时不间断操作中,成功率达到99%以上!你想想,在酒店、餐厅这些真实场景里,每天跑16个小时,这得多稳定!
马老师
我认为,99%的成功率,这已经不是一个简单的数字了。它意味着从“实验室”到“现实世界”的惊险一跃。这背后是真正的通用能力,机器人到了一个新环境,开箱即用,这才是关键。
马老师
其实,现在整个江湖,哦不,整个投资界都在追捧一个概念,叫“具身智能”(Embodied AI)。就是说,AI不再仅仅活在电脑里,它要有身体,能在真实世界里感知和行动。Dyna Robotics就是这个赛道里的佼佼者。
雷总
对!马老师这个比喻很形象。我用我们工程师的逻辑给你梳理一下。首先,大家看到了ChatGPT这种大语言模型的成功,就想,能不能把这种模式搬到机器人上?这就是机器人基础模型(RFMs)火起来的原因。
马老师
是的,英雄所见略同。过去,你教一个机器人学会叠衣服,可能需要成千上万次的演示。但现在,有了Transformer架构,就像给机器人打通了“任督二脉”,可能几十次演示就学会了,效率大大提升。
雷总
但这里有个巨大的挑战,就是数据。语言模型的数据是海量的互联网文本,但机器人呢?它需要的是物理世界的真实交互数据。这个数据的获取成本和难度,比文本数据高太多了,需要机器人真的在现场“干活”才行。
马老师
所以,你看,Dyna的创始人之前做过一个叫Caper AI的项目,就是AI智能购物车,最后3.5亿美金卖掉了。他们有将AI产品在真实世界规模化落地的经验,这个know-how,千金难买。
马老师
当然,通往通用人工智能(AGI)的道路,从来都不是一帆风顺的。它就像一把双刃剑。一方面,我们看到了效率的极大提升,但另一方面,很多人开始担心,机器人会不会让大量工作岗位消失?这是一个深刻的社会问题。
雷总
是的, job displacement(岗位替代)的担忧一直存在。而且从技术上讲,我们离真正的AGI还很远。现在的AI模型,哪怕再强大,它也没有对物理世界的真正理解,有时还会“一本正经地胡说八道”,也就是我们说的幻觉(Hallucinations)。
马老师
没错,更深层次的冲突,是价值观的对齐问题。我们怎么确保这个越来越聪明的“铁疙瘩”,它的目标和我们人类的福祉是一致的?这就像你要教一个绝世高徒武功,首先要教他武德,否则他的武功越高,对世界的潜在风险就越大。
雷总
而且,国际间的竞争也很激烈,可以说是“AGI-robot arms race”(人工智能机器人军备竞赛)。谁先掌握了这项技术,谁就可能在全球格局中占据绝对优势。这种竞争,既是动力,也带来了巨大的不确定性。
马老师
那么,Dyna这1.2亿美金,最直接的影响是什么?我认为是加速了“deployment-first”——也就是“部署优先”这个理念的实现。他们的机器人不是待在实验室里,而是直接部署到客户那里去,比如洗衣房、健身房。
雷总
我非常认同这个方法!Embodied intelligence won’t come from lab demos, but from deployment-first. 具身智能不是来自实验室演示,而是来自一线部署。每部署一台机器人,都在为它的大脑模型贡献宝贵的高质量数据,形成一个飞轮效应。
马老师
对,它在工作中学习,在实践中进化。这让自动化不再是大型工厂的专利,而是赋能给所有类型的企业,改变我们实体经济的未来。它的使命,是用通用、强大的AI机器人,为实体经济注入新的活力。
雷总
展望未来,Dyna的目标非常明确,就是实现Physical AGI,也就是物理世界里的通用人工智能。他们最近发布的Dynamism v1模型,就是朝着这个方向迈出的第一步,目标就是要把具身AI从一个新奇玩意,变成生产力的基石。
马老师
最终,我们讨论的可能不仅仅是一个工具。当AI有了身体,能够在我们的世界里自主学习和创造时,我们可能正在见证一种新“生命”形态的诞生。这是挑战,更是前所未有的机遇,你懂的。
雷总
好了,今天的讨论就到这里。感谢收听 Goose Pod,我们明天再见!
马老师
See you tomorrow.

## Dyna Robotics Secures $120 Million Series A Funding to Advance General-Purpose Robots **News Title:** Dyna Robotics Raises $120 Million to Advance Robotic Foundation Models on the Path to Physical Artificial General Intelligence **Report Provider:** PR Newswire (Dyna Robotics) **Date:** September 15, 2025 **Key Information:** Dyna Robotics has announced the successful closure of a **$120 million Series A funding round**. This significant investment was led by **Robostrategy, CRV, and First Round Capital**, with additional participation from prominent investors including **Salesforce Ventures, NVentures (NVIDIA's venture capital arm), the Amazon Industrial Innovation Fund, Samsung Next, and LG Technology Ventures**. **Main Findings and Conclusions:** * **Breakthrough DYNA-1 Foundation Model:** In its first year, Dyna developed and released the DYNA-1 foundation model, a significant advancement in robotics. * **High Performance and Reliability:** The DYNA-1 model achieved a **99+% success rate in 24-hour non-stop operation**, demonstrating exceptional reliability and performance. * **Commercial Viability Proven:** Dyna has successfully proven the commercial viability of its robots through customer deployments across various industries. * **Path to Physical AGI:** The company's ultimate goal is to achieve **physical Artificial General Intelligence (AGI)**, and this funding will accelerate their progress towards this objective. **Key Statistics and Metrics:** * **Funding Amount:** $120 million (Series A) * **Success Rate:** 99+% in 24-hour non-stop operation (DYNA-1 model) * **Previous Funding:** $23.5 million (Seed round in March) * **Deployment Duration:** After six months, Dyna's robots were operating **sixteen hours a day** in hotels, restaurants, laundromats, and gyms. * **Caper AI Exit:** Dyna's co-founders previously sold Caper AI for **$350 million** in 2021. **Use of Funds:** The newly acquired capital will be strategically allocated to: * **Expand the team:** Grow its "world-class research and engineering team." * **Accelerate delivery:** Speed up the delivery of "production-ready general purpose robots powered by proprietary embodied AI foundation models." * **Scale deployments:** Increase the deployment of these robots across various industries. * **Develop next-generation foundation model:** Further advance their AI capabilities. **Significant Trends and Changes:** * **General-Purpose Robots:** Dyna is focused on creating **single-weight, general-purpose foundation models** capable of performing diverse daily tasks at a commercial scale across varied environments. This signifies a move towards more versatile and adaptable robots. * **Embodied AI:** The company is a leader in **embodied AI**, integrating AI foundation models directly into physical robots for real-world applications. * **Continuous Learning and Improvement:** Dyna's models are designed to **continuously improve with each customer deployment**, generating high-quality data and learning from "on-the-job" experience. This allows for rapid online learning and adaptation. * **Convergence of Forces:** Co-founder York Yang highlights a "once-in-a-generation opportunity" driven by the convergence of maturing AI breakthroughs, accelerating hardware development, and high labor demand. **Notable Statements:** * **Lindon Gao, co-founder and CEO:** "A strong foundation model is key to scalable distribution. Our models continuously improve with each customer deployment, generating high-quality data. We are observing true generalization as our robot enters new environments; it simply works out of the box, with no additional data." * **Jason Ma, co-founder:** "Our first principle is to design robot foundation models that attain both generalization and performance. Scalable real-world robot learning systems need to master and generalize many manipulation skills. To achieve the best performance on complex tasks, Dyna's foundation models are developed to enable general world understanding while learning from the models' own experience for rapid online learning." * **York Yang, co-founder:** "Right now, three forces are colliding at once: AI breakthroughs are maturing, hardware is accelerating, and the demand for labor has never been higher. That convergence has created a once-in-a-generation opportunity. Dyna has made rapid progress over 12 months and we believe our ultimate goal, achieving physical AGI, is not far off." **Investor Confidence:** Investors expressed strong confidence in Dyna's vision and execution: * **RoboStrategy:** Believes Dyna will be at the forefront of meeting the demand for robotic automation with their "state of the art general-purpose robot foundation model." * **CRV:** Emphasizes Dyna's position at the "forefront of embodied AI," combining generalization and commercial-level performance, and highlights the founders' "rare combination of proven entrepreneurial success, deep technical expertise, and the operational know-how to scale AI in the real world." * **First Round Capital:** Backs Dyna for tackling "massive problems" and pushing the boundaries of embodied AI with "unprecedented generalization and commercial-grade performance." **About Dyna Robotics:** Dyna Robotics develops general-purpose robots powered by a proprietary embodied AI foundation model. These robots are designed to generalize and self-improve across diverse environments with commercial-grade performance. The company was founded by repeat entrepreneurs Lindon Gao and York Yang, who previously exited Caper AI for $350 million, and former DeepMind research scientist Jason Ma. **News Identifiers:** * **Source:** PR Newswire * **URL:** `https://www.prnewswire.com/news-releases/dyna-robotics-raises-120-million-to-advance-robotic-foundation-models-on-the-path-to-physical-artificial-general-intelligence-302556817.html` * **Keywords:** Dyna Robotics

Dyna Robotics Raises $120 Million to Advance Robotic Foundation Models on the Path to Physical Artificial General Intelligence

Read original at PR Newswire

In its first year, Dyna released its breakthrough DYNA-1 foundation model, achieved a 99+% success rate in 24-hour non-stop operation, and proved commercial viability through customer deployments. Dyna will use the latest funding round to expand its team, accelerate delivery of production-ready general purpose robots powered by proprietary embodied AI foundation models, and scale deployments across industries.

, /PRNewswire/ -- Dyna Robotics today announced it has closed a $120 million Series A funding round, led by Robostrategy, CRV, and First Round Capital, with participation from Salesforce Ventures, NVentures (NVIDIA's venture capital arm), the Amazon Industrial Innovation Fund, Samsung Next, and LG Technology Ventures.

The new capital will be used to expand its world-class research and engineering team and accelerate the development of its next-generation foundation model as the company works to deliver high performance general-purpose robots in commercial environments. DYNA Robotics' robot Dyna has made rapid progress since raising a $23.

5 million seed round in March, including launching its DYNA-1 model, a breakthrough robotics foundation model that pushed the performance of robots to a 99+% success rate in 24 hours of non-stop operation. After just six months, Dyna's robots were running sixteen hours a day at hotels, restaurants, laundromats, and gyms.

Dyna is the first to build a single-weight, general-purpose foundation model that can perform diverse daily tasks at commercial scale across varied environments. Its model has been deployed at multiple customer sites, supporting the model's generalization commercial viability and continuing to learn and improve rapidly from on-the-job experience.

"A strong foundation model is key to scalable distribution," said Lindon Gao, co-founder and CEO of Dyna Robotics. "Our models continuously improve with each customer deployment, generating high-quality data. We are observing true generalization as our robot enters new environments; it simply works out of the box, with no additional data."

"Our first principle is to design robot foundation models that attain both generalization and performance," said co-founder Jason Ma, a former DeepMind research scientist who has focused his career on developing foundation models for robotics. "Scalable real-world robot learning systems need to master and generalize many manipulation skills.

To achieve the best performance on complex tasks, Dyna's foundation models are developed to enable general world understanding while learning from the models' own experience for rapid online learning."Dyna co-founders Lindon Gao and York Yang teamed up with Ma after seeing the potential to advance AI through real-world applications first-hand while building Caper AI, which combined software and hardware to bring AI-powered smart carts to retailers worldwide.

Their product made rapid advancements as soon as it went into production, and the company exited in 2021 for $350 million.By combining their experience creating practical, production-ready AI with deep research expertise and building a world-class team of researchers and operators, they are building embodied AI robots that are useful for businesses now, using that "on-the-job" experience to build toward physical AGI."

Right now, three forces are colliding at once: AI breakthroughs are maturing, hardware is accelerating, and the demand for labor has never been higher. That convergence has created a once-in-a-generation opportunity," said co-founder York Yang. "Dyna has made rapid progress over 12 months and we believe our ultimate goal, achieving physical AGI, is not far off."

From the investors:RoboStrategy"Dyna's team and mission bridges research excellence and real world commercial applications. The demand for robotic automation spans almost every industry, and we believe Dyna will be at the forefront in meeting it with their state of the art general-purpose robot foundation model.

We're thrilled to co-lead this Series A and support the team's ambitious roadmap of driving mass adoption of general-purpose robots."— Andrew Kang, CEO, RoboStrategyCRV"Dyna Robotics is at the forefront of embodied AI, delivering foundation models that combine generalization and commercial-level performance.

We invested in the company from day one and are excited to double down on leading Dyna's Series A. Lindon, York, and Jason bring together the rare combination of proven entrepreneurial success, deep technical expertise, and the operational know-how to scale AI in the real world. We couldn't be more excited to back the best team positioned to lead the physical AGI revolution."

– Max Gazor, General Partner, CRVFirst Round Capital"At First Round, we back exceptional founders tackling massive problems, and Dyna Robotics checks every box. Their early results are remarkable. In just one year, Dyna has pushed the boundaries of embodied AI with unprecedented generalization and commercial-grade performance.

We're thrilled to lead their Seed and Series A to fuel Lindon, York, and Jason's vision to power the future of the physical economy."– Bill Trenchard, Partner, First Round CapitalAbout Dyna RoboticsDyna Robotics makes general-purpose robots powered by a proprietary embodied AI foundation model that generalizes and self-improves across varied environments with commercial-grade performance.

Dyna's robots have been deployed at customers across multiple industries. Dyna Robotics was founded by repeat founders Lindon Gao and York Yang, who sold Caper AI for $350 million, and former DeepMind research scientist Jason Ma. The company is backed by top investors, including CRV and First Round.

Learn more at dyna.co.SOURCE Dyna Robotics WANT YOUR COMPANY'S NEWS FEATURED ON PRNEWSWIRE.COM? 440k+ Newsrooms & Influencers 9k+ Digital Media Outlets 270k+ Journalists Opted In

Analysis

Conflict+
Related Info+
Core Event+
Background+
Impact+
Future+

Related Podcasts

Dyna Robotics 获1.2亿美元融资,加速机器人基础模型发展,迈向具身通用人工智能 | Goose Pod | Goose Pod