Skip to content
Sign in
SWC 2026 FinalistGrand Final: Main Stage

Devol Robots Inc.

Force-based physical-AI 'world model' giving industrial robots human-like compliance for contact-rich assembly tasks.

Kuala Lumpur, Malaysia, AsiaRobotics & hardware, Physical AI – force-controlled robotic manipulation
Stage
UndeterminedAI at the core
Product
PilotPilots / LOIs
Funding raised
Undisclosedno funding
Founded
20233 years ago
Founders
2Sze, Elijah
Global score
4.8Grade DRank 48 of 151

Overview

Devol Robots is a physical AI startup developing a force-based "world model" for robotic manipulation and industrial assembly, deployed at US industrial customers. A US-incorporated company (Inc.) whose team (~18 people) is mostly based in Malaysia.

Traction

Pilots / LOIs

Deployment announced at US industrial customers, but no named customer; the website shows only demonstrations (conservative).

Market

AI in robotics (industrial manipulation)

Most factory robots excel at repetitive, position-controlled motion but fail at contact-rich tasks such as insertion, fastening or handling variable parts, which remain manual. Labour shortages, reshoring and the physical-AI wave are pushing manufacturers to automate these tasks. Buyers evaluate cycle time, yield and integration effort per robotic cell, usually through system integrators. Devol competes with heavily funded physical-AI players by focusing on force-based control.

Addressable market

2025
$25B
2030
$51.8B
+12.92% a year

AI in robotics (industrial manipulation): $25.0bn in 2025, +12.9%/year (Mordor Intelligence); Statista $22.5bn.

Customers

Manufacturers with contact-rich assembly (electronics, automotive components, motors, battery packs); robot system integrators.

Who signs the cheque

Manufacturing engineering and automation leads; integrators packaging robotic cells.

Geographies

  • United States
  • Malaysia (engineering base)

Drivers

  • Labour shortages and reshoring of manufacturing
  • Physical-AI models making robots adaptable
  • Demand to automate high-mix assembly

Headwinds

  • Customers not verified publicly
  • Massively funded competitors
  • Integration effort per cell

The AI in robotics market is estimated at $25.0bn in 2025 (+12.9%/year, $51.8bn in 2031) according to Mordor Intelligence; Statista values it at $22.5bn.

Scalability 7/10. Manipulation software model reusable across many robotic cells, but on-site industrial integration.

Market source: giiresearch.com

Technology

Category

Physical AI / robot control software

AI at the coreRobotics / physical AIMachine learning / predictiveComputer vision

Maturity

  1. R&D
  2. Prototype
  3. Pilot
  4. On the market
  5. Scaling

pilot

How it works

The model fuses joint torques and forces, stiffness parameters and vision so that the robot adapts its strategy in real time through force feedback, inspired by human muscle compliance. It targets manipulation and industrial assembly.

Moat

Potentially meaningful: the founders' mathematical foundations of force-based control and deployment data; weak until customer references are public.

Differentiation

Force/compliance-centred world model versus vision-only imitation-learning approaches that struggle with contact.

Integration points

  • Robot OEM controllers
  • Actuators and motors
  • Assembly lines (motors, battery packs, electronics)
  • Mobile robot platforms

Depends on

  • Robot arms with joint torque sensing
  • Integrator partnerships
  • Industrial safety standards

Use of AI

Force-based physical AI "world model" that adapts robotic manipulation and assembly in real time.

Product in detail

Model combining torques/forces measured at each joint, stiffness parameters and vision, allowing the robot to adapt its strategy in real time through force feedback (compliance control inspired by human muscles).

Business

Model
Not disclosed; likely software licences per robot cell plus integration services.
Revenue logic
Unknown; value measured in labour saved and yield per automated cell.
Go-to-market
Direct deployments at US industrial customers from a Malaysian engineering base.

Next milestones

  1. Named US reference customer with published cycle-time results
  2. First institutional round
  3. Partnership with a robot OEM or integrator

Key risks

  • Unverified traction
  • No disclosed funding
  • Competition from large physical-AI labs

Opportunities

5 found in the network

Concrete moves for Devol Robots Inc. inside the network: who to sell to, sell through, build with or raise from, each with the reasoning, the evidence and the first step.

  • Customers 2
  • Co-development 1
  • Investors 2

Customers2

Who could buy, or sell through

Automate contact-rich steps of motor production

EVR runs a coil factory in Manesar with capacity of 20,000 motors per month; winding, insertion and magnet placement are classic contact-rich tasks. A pilot cell using Devol's force-controlled manipulation could raise throughput as EVR scales with Indian OEMs.

25%confidence

Evidence from the records

  • “deployed at US industrial customers”
  • “a coil factory in Manesar (capacity 20,000 motors/month)”
If it works
A named industrial reference for Devol and higher productivity for EVR's plant.
First step
Ask EVR's operations team to list the three most manual assembly steps and share videos for feasibility.

Battery-pack assembly cell for EIC-funded pilot manufacturing

Leitspalt's EIC grant funds pilot-scale manufacturing validation of its integrated battery-pack component. Pack assembly involves precise insertion and fastening of cells and cooling structures; a Devol cell could be part of the pilot line, making Leitspalt's licensing package more attractive to OEMs.

20%confidence

Evidence from the records

  • “force-based "world model" for robotic manipulation and industrial assembly”
  • “the EIC grant (from June 2026) funds pilot-scale manufacturing validation and integration into EV platforms”
If it works
An automated assembly reference for Leitspalt licensees and a European pilot for Devol.
First step
Propose a remote feasibility review of Leitspalt's assembly sequence.

Co-development1

An integration or joint product

Precise ultrasonic actuators for compliant grippers and mobile manipulation

Piezo Sonic makes ultrasonic motors and all-terrain AMRs; Devol provides manipulation intelligence. Combining them could produce a mobile manipulator or compliant gripper, and Piezo Sonic's distributor network in Japan offers a sales channel.

20%confidence

Evidence from the records

  • “Model combining torques/forces measured at each joint, stiffness parameters and vision”
  • “Piezoelectric ultrasonic motors that operate in a vacuum and in strong magnetic fields, built into long-endurance all-terrain AMRs.”
If it works
A joint demonstrator and access to Japanese industrial distribution.
First step
Request Piezo Sonic motor specifications to evaluate torque-sensing compatibility.

Investors2

A fund whose thesis fits

Seed round from a robotics-focused early investor

Devol has no disclosed funding; Anorak invests at pre-seed and seed in AI, robotics and space and backed TRL11 in the network, offering a warm introduction.

30%confidence

Evidence from the records

  • “no public funding round found”
  • “Robotics & hardware”
  • “syndicate led by Boost VC, Wonder Ventures, Anorak Ventures”
If it works
A first institutional round to fund US deployments.
First step
Ask TRL11's founder Nicol Verheem for an introduction to Anorak.

Strategic robotics investor with manufacturing relevance

Toyota Ventures invests at Seed and Series A in Robotics & hardware and AI globally and backed ZymoChem in the network. Force-controlled assembly is directly relevant to automotive manufacturing.

25%confidence

Evidence from the records

  • “Robotics & hardware”
  • “Breakout Ventures (lead), lululemon, Toyota Ventures, GS Futures, KdT Ventures, Cavallo Ventures”
If it works
A strategic co-investor with potential access to automotive assembly use cases.
First step
Ask ZymoChem's CEO for an introduction to Toyota Ventures.

Analyst reading of 21 Nov 2026, from the public record of the network. Method: analyst-ai.

Investors that fit

Seed fund led by Greg Castle (Oculus, Flexport, Anduril) backing hard technical problems in AI, robotics and space.

Force-controlled robotic manipulation ("physical AI") is exactly a hard technical problem in robotics. Undetermined stage and no named customers, so Anorak would be an early lead.

50%confidence

Evidence from the records

  • “Physical AI – force-controlled robotic manipulation”
  • “Robotics & hardware”
Intro path
TRL11 (Anorak portfolio, network peer)
First step
Send a technical demo video and a seed deck to Anorak via TRL11.

Singapore accelerator investing up to $500K in twice-yearly batches of increasingly AI-focused founders from Southeast Asia and beyond.

Malaysian physical AI for force-controlled manipulation — SEA AI; hardware component lengthens cycles.

45%confidence

Evidence from the records

  • “"Southeast Asia"”
  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
Intro path
Founder-to-founder intro via Belli (network portfolio company of Iterative).
First step
Apply with a software-led milestone plan.

Singapore early-stage VC (~$300M AUM, 150+ companies) backing Southeast Asian and Middle Eastern category leaders like Ninja Van and Xendit.

Malaysian physical-AI robotic manipulation — SEA geography; hardware intensity is atypical for GGV.

30%confidence

Evidence from the records

  • “Southeast Asia”
  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
Intro path
Founder-to-founder intro via BioSapien (network portfolio company of Golden Gate Ventures).
First step
Test interest with an industrial-customer pipeline.

Toyota's Los Altos venture arm ($800M+ across Frontier and Climate funds) backing mobility, climate, robotics and AI at seed and Series A.

Force-controlled physical AI for robotic assembly: frontier robotics, no named customers yet.

30%confidence

Evidence from the records

  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
First step
Pitch once a customer is named.

Singapore-based early-stage VC backing Asian and Southeast Asian founders; led Baniql's US$1.6M seed.

Malaysian-linked physical-AI robotic manipulation: SEA robotics, but customers are US and unnamed.

25%confidence

Evidence from the records

  • “Physical AI – force-controlled robotic manipulation”
  • “Malaysia”
First step
Pitch with a named industrial pilot.

Japanese manufacturer that formed a 2022 capital and business alliance with Piezo Sonic.

Malaysian force-controlled manipulation — robotics.

25%confidence

Evidence from the records

  • “Piezo Sonic”
  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
Intro path
Founder-to-founder intro via Piezo Sonic Corporation (network portfolio company of Kanda Kogyo (神田工業)).
First step
Low priority.

Japanese VC holding a board seat at Piezo Sonic; otherwise little documented.

Malaysian robotic manipulation — robotics adjacency only.

15%confidence

Evidence from the records

  • “Piezo Sonic”
  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
Intro path
Founder-to-founder intro via Piezo Sonic Corporation (network portfolio company of Higin Capital (肥銀キャピタル)).
First step
Low priority.

Thai public agency supporting innovation-driven startups through grants and Startup Thailand programmes; backed MUI Robotics.

Malaysian robotics — needs Thai entity.

15%confidence

Evidence from the records

  • “Thailand”
  • “Physical AI – force-controlled robotic manipulation”
  • “Deployment announced at US industrial customers, but no named customer”
Intro path
Founder-to-founder intro via MUI Robotics (network portfolio company of National Innovation Agency (NIA) Thailand).
First step
Low priority.

Investment analysis

4.8D

Grade D

Rank 48 of 151

Disruption7.0/10
Business model3.0/10
Market6.0/10
Risk(lower is better)8.0/10

Global score = mean of disruption, model, market and (11 − risk). Last updated 21 Nov 2026.

Attractive technology thesis but commercially unproven: speculative.

Disruption

7/10

Force-based world model for robotic manipulation: a new approach in physical AI.

Business model

3/10

Deployments announced at US manufacturers but no named customer, only demonstrations.

Market

6/10

TAM $25.02bn → base 6; scalability 7/10 (neutral); no international presence: US deployments not confirmed → 6/10.

Risk

8/10

Lower is better

High risk: difficult technology, undetermined funding and heavily capitalized competition in physical AI.

Strengths

  • Differentiated technical approach
  • Booming physical AI market

Risks

  • Customers not verified
  • Massively funded competitors

Funding

Undisclosed, no funding

Amount used in the study. No public funding round found.

Note. no public funding round found

Website corrected: devolrobots.com redirects (301) to devolrobots.ai; legal entity Devol Robots Sdn Bhd (Malaysia) according to the LinkedIn link. The website shows only demos (plug insertion, hearing aid assembly, DIN rail, snap-fit) with no named customer. Not Malaysia 2026 champion (West: Atoz CarbonCycle; East: NEX EcoVentures) — qualification path not found.

Team

  • Sze Yuan Cheong

    Co-founder & CEO

    Developed with Elijah Ong the mathematical foundations of force-based robotic control commercialized by Devol.

  • Elijah Ong

    Co-founder & CTO

    Roboticist trained at Oregon State University (master's in reinforcement learning for robotic grasping).

Sources

Identification confidence
medium
  1. 1devolrobots.ai/
  2. 2ibtimes.co.uk/devol-robots-force-based-ai-automation-1781557
  3. 3sfplayground.vercel.app/success-stories/devol-robots-inc
  4. 4tegas.org.my/2026/08/startup-world-cup-malaysia-2026-east-chapter-final/

From the open study Startup World Cup 2026 by TIBOK, based on public sources, verified in November 2026. Scores are an analyst's judgment on an explicit grid, not the jury's.

Wall

Wall

No posts yet