- The Rundown Robotics
- Posts
- Rivian spins off robotics startup
Rivian spins off robotics startup
PLUS: Xpeng debuts female-form humanoid
Good morning, robotics enthusiasts. EV maker Rivian just spun out its second startup of the year — Mind Robotics, an industrial AI venture backed by $115M and a vague promise to turn factory-floor data into a “robotics data flywheel.”
While details are under wraps, could Rivian be quietly building the next big leap in factory intelligence, or something even more disruptive?
In today’s robotics rundown:
Rivian launches a robotics spinoff
Xpeng unveils next-gen robotaxis and humanoid
Humanoid startup K-Scale calls it quits
MIT teaches robots to map faster
Quick hits on other robotics news
LATEST DEVELOPMENTS
RIVIAN

Image source: Wikimedia Commons / Richard Truesdell
The Rundown: Rivian just launched Mind Robotics, its second spinoff this year — an industrial AI venture aimed at turning the EV maker’s factory-floor data into a commercial “robotics data flywheel.” So far, what that means is anyone’s guess.
The details:
Funding is already lined up, with roughly $115M in external seed capital to kickstart the new venture.
Rivian CEO RJ Scaringe announced the spinoff on an earnings call but offered limited details on what it will actually build or sell.
The company didn’t confirm employee transfers, though its shareholder letter nods to a “strong bench of technology talent,” hinting at internal hires.
In March, the company carved out its skunkworks micromobility team into a new venture called Also, which just debuted its first e-bike.
Why it matters: Rivian is hush about its robotics plans, but U.S. automakers are racing to monetize their factory AI — GM announced similar plans in October, and Tesla’s pitching its Optimus for manufacturing. With $115M committed, investors are betting that Rivian’s learned something worth commercializing beyond the assembly line.
XPENG

Image source: Xpeng
The Rundown: Xpeng just unveiled three new robotaxis, engineered for full self-driving and powered by its own AI chips, as well as its sleek next-gen humanoid, in male and female versions.
The details:
At its Xpeng AI Day, the company unveiled Iron, its next-gen humanoid with fluid, humanlike movement, slated for mass production by late 2026.
The new robotaxis run a vision‑only stack powered by four in‑house Turing AI chips per car, avoiding lidar and HD maps to cut hardware cost.
Xpeng said its VLA 2.0 model will power the robotaxis, humanoids, and its flying-car projects, unifying autonomy and embodied AI under one system.
Alibaba is partnering on the robotaxi rollout through its AutoNavi mapping unit and Amap ride-hailing app, with trials starting in Chinese cities next year.
Why it matters: Xpeng is clearly taking a page from Tesla’s playbook, skipping driver-assist tech to jump straight into driverless rides, controlling both the AI brains and the hardware while cutting costs with vision-only systems. If it works, the company runs its own ride network and mass-produces humanoids on the same tech stack.
K-SCALE

Image source: K-Scale / X
The Rundown: K‑Scale Labs, the year-old Palo Alto startup promising a low‑cost open‑source humanoid, is pulling the plug on K‑Bot preorders and returning deposits after failing to raise the cash to build at scale.
The details:
The startup’s K-Bot was pitched as a low-cost, open-source humanoid for developers, priced under $10K to undercut commercial rivals.
Founder Benjamin Bolte told customers that the company has laid off most of its staff and has “less than a month of runway.”
The Information reports that K-Scale pursued potential deals with 1X and Bot Co., but those talks fell through.
The core engineering team has now launched Gradient Robots, a new startup aiming to be “the open‑source Unitree for America.”
Why it matters: Bolte drew a sharp comparison between U.S. and Chinese capital markets, saying he expected a deeper U.S. appetite for a cost‑competitive domestic humanoid. In a final nod to its ethos, K‑Scale is open‑sourcing K‑Bot and Zeroth Bot so others can keep building.
MIT

Image source: MIT
The Rundown: MIT researchers developed a system that lets robots rapidly map large, unpredictable environments by creating and stitching together smaller 3D submaps on the fly — solving a major bottleneck in robotic navigation.
The details:
The MIT system lets robots map large spaces by stitching together smaller 3D submaps, bypassing AI vision models’ 60-image limit.
The breakthrough blends vintage computer vision math with modern AI to correct the distortions machine-learning models introduce into submaps.
It generates close-to-real-time 3D maps of complex spaces using just smartphone video, with less than 5 cm of error.
The system can improve search-and-rescue robots, extended reality apps for VR headsets, or help warehouse robots accurately locate and move inventory.
Why it matters: Current robot mapping systems can’t process the thousands of images required to navigate disaster zones in real time. MIT’s approach looks to make fast, accurate 3D mapping practical for search-and-rescue missions, warehouse automation, and VR headsets, with no special cameras or expert tuning required.
QUICK HITS
Norway’s Physical Robotics, makers of π humanoid and founded by Halodi Robotics’ Phuong Nguyen, has come out of stealth and announced $4M in fresh funding.
Volkswagen is building its own smart‑driving chip in China with Horizon Robotics, a homegrown partner, to power future VW models sold there.
Hullbot, an Australian ocean robotics firm, secured $16M in funding to develop autonomous underwater robots to clean and inspect ship hulls.
China is extending its EV‑and‑battery playbook into humanoids, reusable rockets, and LEO satellites in a push to win on scale and cost, The Information reports.
Elon Musk said on the All‑In podcast that Tesla aims to scale its robotaxi pilot to 500 cars in Austin and 1K in the Bay Area by year‑end.
China’s Leju Robotics unveiled Kuavo 5, a modular humanoid that can switch between walking and wheels, swap hands for different jobs, and work for hours.
Infravision raised $91M to scale its drone‑powered TX System for building and maintaining power lines, cutting costs on helicopter methods.
DJI launched its Neo 2 selfie drone, a sequel to last year’s model that adds forward lidar plus downward IR sensing for obstacle avoidance and safer follow‑me flight.
Adaptronics raised $3.6M to roll out its next‑gen electrostatic robotic grippers across Europe, moving from pilots to broader factory and logistics deployments.
COMMUNITY
Read our last AI newsletter: Apple taps Gemini for Siri overhaul
Read our last Tech newsletter: This startup wants to edit embryos
Read our last Robotics newsletter: Why LLMs aren’t ‘robot-ready’
Today’s AI tool guide: Use AI to search patents & find opportunities
RSVP to next workshop @ 4PM EST Friday: AI foundations for marketers
That's it for today!Before you go we'd love to know what you thought of today's newsletter to help us improve The Rundown experience for you. |
See you soon,
Rowan, Jennifer, and Joey—The Rundown’s editorial team
