EN
Back to the archive

The archive · Hardware & Devices · Strategic decision · 2024–2025

K-Scale Labs' open-source humanoid bet: $6.25M raised, refunds, shutdown Nov 2025

YC W24 open-source humanoid startup raised $6.25M, took 100+ pre-orders worth $2M+, then shut down in Nov 2025, refunding deposits and open-sourcing its IP.

K-Scale Labs

The betThat an open-source ~$9K humanoid from off-the-shelf parts could win developers, and pre-orders would fund mass production before Series A, without China's supply chain.No longer exists

What the business is

K-Scale Labs built the K-Bot, a 1.4m open-source bipedal humanoid robot with a base price of $8,999, plus an open-source software stack for training and deploying robot AI, aimed at developers and researchers.

Starting capital$6.25M total: ~$4M seed at a $50M valuation led by Fellows Fund, $250K from Nat Friedman and Daniel Gross (Apr 2025), and a $2M strategic investment from Zhejiang Taotao Vehicle Industry (Jun 2025); 36Kr also reports 2024 revenue of $2.9M

How it started

Benjamin Bolte, an ex-Meta AI researcher and ex-Tesla senior Autopilot engineer, started the project in a Palo Alto garage in 2024 because no cheap, hackable humanoid existed; K-Scale went through YC W24 and had a standing, waving robot by Demo Day.

What happened

In May 2025 K-Scale open-sourced its software suite and opened K-Bot pre-orders at $8,999 with a 'Full Autonomy' upsell. June 2025 brought a $2M strategic round from Zhejiang Taotao Vehicle Industry for supply-chain support. About 10 engineers built roughly 10 prototypes costing over $100,000 each. More than 100 pre-orders worth $2M+ came in, but the Q3 2025 attempt to raise $10–15M failed; on one admired VC's advice the team had killed the cheaper Z-Bot and bet everything on the high-end K-Bot.

How it ended up

On Nov 4–5, 2025 Bolte emailed customers and investors that K-Scale was shutting down: the team was laid off, deposits were refunded in full, and all proprietary hardware and software was released open-source (CERN-OHL-S-2.0 / MIT) with under $400K and less than a month of runway left; acquisition talks with 1X and The Bot Co. went nowhere.

Background

K-Scale Labs was founded in 2024 by Benjamin Bolte, an ex-Meta AI researcher and former Tesla senior Autopilot engineer, to build the cheap, hackable humanoid he couldn't find: an open-source, 1.4m bipedal robot sold at $8,999 when comparable US research humanoids cost over $50,000. The company went through YC's W24 batch and had a standing, waving robot by Demo Day.

K-Scale raised about $6.25M in total — a ~$4M seed at a $50M valuation led by Fellows Fund, $250K from Nat Friedman and Daniel Gross, and a $2M strategic investment from Zhejiang Taotao Vehicle Industry — and opened K-Bot pre-orders in summer 2025. From May to early November it took more than 100 pre-orders worth over $2M, including one from OpenAI's robotics head Caitlin Kalinowski. Traction was real: its Launch HN drew 233 points and 95 comments in July 2025.

The economics never closed. A team of about 10 engineers built only about 10 prototypes, each costing over $100,000, and moving from 3D printing to die-casting and cold forging required tooling investment no one would fund. The Q3 2025 attempt to raise $10–15M failed, and acquisition talks with 1X and The Bot Co. produced only offers to hire a few team members. By early November 2025 the company had under $400,000 and less than a month of runway.

On Nov 4–5, 2025 Bolte emailed customers that K-Scale was shutting down: deposits refunded in full, the team laid off, and all proprietary hardware and software released open-source (CERN-OHL-S-2.0 hardware, MIT software). In his postmortem he blamed failed financing and a US supply-chain gap, arguing Chinese competitors such as Unitree, EngineAI, Booster and Noetix had funded large-scale production while Western firms shipped almost nothing.

What has to be true

  • Killing the cheaper Z-Bot to bet everything on the $8,999 K-Bot, on one admired VC's advice, left a single capital-hungry product line (36Kr).
  • Mass-production economics never closed: about 10 prototypes at over $100K each, with die-casting and cold-forging tooling costs no one would fund (36Kr).
  • The $10–15M Series A failed in Q3 2025 and acquirers only wanted to hire team members, leaving cash at ~$400K (36Kr).
  • Bolte's farewell email blamed the US's missing small-batch manufacturing ecosystem versus Unitree, EngineAI, Booster and Noetix, who had raised mass-production capital (via Mike Kalil).

What can be applied

Pre-order demand can't substitute for tooling capital: K-Scale took $2M+ in K-Bot pre-orders, but no Series A came for its $100K+ prototypes — production never began.

Aftermath

As of 2026-09-02 K-Scale Labs is defunct: deposits were refunded, staff laid off, and its IP released under CERN-OHL-S-2.0 and MIT licenses, with the kscalelabs GitHub organization (130+ repos) still public. Co-founder Jingxiang Mo left in Sept 2025 and founded Gradient Robotics, which inherited part of the team. The shutdown became a widely cited example of US humanoid startups losing the manufacturing race to China's supply chain.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases