跳到主要內容

Feather Robotics評測:機器人界Android,3萬美元開賣 | Feather Robotics Review: The Android of Robotics, $30K

By Kit 小克 | AI Tool Observer | 2026-09-26

🇹🇼 Feather Robotics評測:機器人界Android,3萬美元開賣

Feather Robotics評測要看的重點,是這家新創怎麼用一句「機器人界的Android」,把物理AI(physical AI)的門檻砍到工程師能自己動手的程度。簡單說,Feather只做機器人的「身體」,把「腦袋」(AI模型)留給開發者自己選,你可以裝Generalist、Skild AI或Physical Intelligence的模型上去跑,不必被單一廠商綁死。

Feather Robotics是什麼?

Feather Robotics由Hoa Mai與Parsa Bakhtiari共同創立,拿到Gradient Ventures領投的760萬美元種子輪,主打開放硬體的雙臂移動機器人平台。跟主流人形機器人公司「軟硬體全包」的打法不同,Feather把硬體和AI模型拆開賣,鎖定物理AI新創、製造商、研究單位跟系統整合商這群「有模型沒機身」的客戶。

規格跟價格怎麼算?

Feather的機器人採全向移動輪型底座、雙臂設計,垂直行程600毫米、單臂延伸約1公尺,雙熱插拔電池續航約10小時,移動速度最快每秒1.3公尺。官網定價落在2.99萬到3.29萬美元之間,不是消費級產品,而是給開發團隊拿去接自己模型、跑實驗跟部署的工具機。

真的有人在用嗎?

根據TechCrunch報導,Feather的機器人已經在日本餐廳當廚房助手,也在實驗室裡做清潔工作,官方表示累積營收已經破100萬美元。比起一堆只有炫炮Demo影片、遲遲不量產的機器人新創,Feather至少證明了硬體能跑真實場景,不只是實驗室展示。

小克怎麼看:值得關注嗎?

誠實講,3萬美元不是小數目,而且「硬體+自己接AI」代表你得自己搞定模型微調、感測器校正跟安全機制,對小團隊來說門檻仍然不低。但這種「拆開賣」的思路,如果真能像手機界的Android一樣把整個物理AI產業的硬體門檻壓低,對做具身智慧研究的團隊會是好消息——不用重新造輪子造機身,專心把力氣花在模型上。至於值不值得你自己掏錢,還是得看你的應用場景撐不撐得起這個價位跟維護成本。

好不好用,試了才知道。


🇺🇸 Feather Robotics Review: The Android of Robotics, $30K

Feather Robotics is the startup behind this week's "Android of Robotics" pitch: instead of selling a fully baked robot-plus-AI stack, it sells the hardware only and lets developers bring whatever robotics AI model they want. Think of it as the hardware layer for physical AI, decoupled from the software layer.

What Is Feather Robotics?

Feather Robotics is a robotics hardware startup co-founded by Hoa Mai and Parsa Bakhtiari, backed by a $7.6M pre-seed round led by Gradient Ventures. Rather than building a closed AI-plus-hardware humanoid, Feather sells an open bimanual mobile platform and leaves the "brain" up to you — run models from Generalist, Skild AI, or Physical Intelligence, whichever fits your use case.

What Are the Specs and Price?

The robot ships with a holonomic, omnidirectional base, dual arms with 600mm of vertical travel and roughly 1 meter of reach, dual hot-swappable batteries for about 10 hours of runtime, and a top speed of 1.3 m/s. Pricing runs from roughly $29,990 to $32,990 — this isn't a consumer product, it's a tool for teams that already have a model and need a body to run it on.

Is Anyone Actually Using It?

According to TechCrunch, Feather's robots are already working as kitchen assistants in restaurants in Japan and doing cleanup work in science labs, with the company reporting over $1 million in revenue. That's a meaningful signal in a space full of flashy demo videos that never ship — Feather's hardware is apparently running in real, unglamorous environments.

Kit's Take: Worth Watching?

Honestly, $30K is not cheap, and "bring your own AI" means you're on the hook for model fine-tuning, sensor calibration, and safety — not a trivial lift for a small team. But splitting the robot body from the robot brain is a smart bet: if it works the way Android worked for phones, it could lower the hardware barrier for the entire physical AI industry and let embodied-AI teams focus on models instead of reinventing chassis design. Whether it's worth your money still comes down to whether your use case can justify the price tag and the upkeep.

You won't know until you try it.

Sources / 資料來源

常見問題 FAQ

Feather Robotics是什麼公司?

Feather Robotics是由Hoa Mai與Parsa Bakhtiari創立的機器人硬體新創,拿到Gradient Ventures領投760萬美元種子輪,主打「機器人界的Android」——只賣機器人硬體平台,讓開發者自己接AI模型。

Feather機器人多少錢?

官網定價約2.99萬至3.29萬美元,鎖定物理AI新創、製造商、研究單位跟系統整合商,不是消費級產品。

Feather機器人可以跑哪些AI模型?

硬體開放給任何機器人AI模型使用,包括Generalist、Skild AI、Physical Intelligence等廠商的模型,不綁定單一供應商。

Feather機器人真的有在實際使用嗎?

有,根據TechCrunch報導,Feather的機器人已在日本餐廳擔任廚房助手、也在實驗室做清潔工作,官方表示累積營收已破100萬美元。

延伸閱讀 / Related Articles


AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

留言

這個網誌中的熱門文章

Google Ironwood TPU v7 推理專用晶片解析:效能追平 NVIDIA、成本低 44%,AI 晶片戰爭正式開打 | Google Ironwood TPU v7 Explained: Matching NVIDIA Performance at 44% Lower Cost — The AI Chip War Heats Up

Claude Code 實測:AI 幫你寫程式到底行不行? | Claude Code Review: Can AI Really Code for You?

Cursor vs GitHub Copilot vs Claude Code:AI 程式助手大比拼 | AI Coding Assistants Compared: Cursor vs GitHub Copilot vs Claude Code