Devol Robots

Cover Block

PUBLIC

Attribute	Details
Name	Devol Robots
Tagline	Building physics-aware embodied AI for robotic manipulation in industrial settings.
Headquarters	San Francisco, US
Founded	2023
Stage	Seed
Business Model	B2B
Industry	Deeptech
Technology	AI / Machine Learning
Geography	North America
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)
Funding Label	Undisclosed (total disclosed ~$4,000,000)

Links

PUBLIC

• Website: https://www.devolrobots.com
• LinkedIn: https://www.linkedin.com/company/devol-robots/

PUBLIC Devol Robots is a seed-stage startup developing a physics-aware embodied AI system for robotic manipulation, positioning itself as a potential solution to the longstanding reliability gap in industrial automation. Founded in 2023 by researchers Cheong and Ong, the company's core thesis is that robots must learn from direct physical interaction,forces, torques, and contact dynamics,rather than primarily from visual data, an approach they term "Physical Intelligence" [International Business Times, 2025 or 2026]. This force-based world model is designed to replace brittle, scripted automation with goal-seeking autonomy, enabling robots to adapt in real-time on factory floors and in warehouses [Perplexity Sonar Pro Brief, retrieved 2024]. The founding team, while not extensively profiled in public sources, is credited with developing the mathematical foundations for force-based control before commercializing the research [International Business Times, 2025 or 2026]. The company has raised an estimated $4 million in seed capital from investors including Y Combinator and Gradient, and is reportedly deploying its models with unnamed U.S.-based industrial clients [Mastra Blog, 2026]. Over the next 12-18 months, the critical watchpoints will be the transition from undisclosed pilot deployments to named, scaled customer relationships and the demonstration of performance metrics that justify a premium over existing robotic programming solutions.

Data Accuracy: YELLOW -- Core technology and founding narrative corroborated by a single publisher profile; funding details partially confirmed.

Taxonomy Snapshot

Axis	Value
Stage	Seed
Business Model	B2B
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Geography	North America
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)

Company Overview

PUBLIC

Devol Robots was founded in 2023 as a commercial vehicle for a research-driven approach to robotic manipulation [Crunchbase]. The company is headquartered in San Francisco and operates as a privately held entity, with a second entity, Devol Robots Sdn Bhd, registered in Malaysia [LinkedIn, retrieved 2024] [Maukerja, retrieved 2024]. The founding story, as reported by the International Business Times, centers on two researchers, identified as Cheong and Ong, who developed mathematical foundations for force-based control and launched the startup to commercialize that work [International Business Times, 2025 or 2026].

Key milestones are sparse in public sources but follow a typical early-stage trajectory. The company was accepted into the Y Combinator accelerator program, a common launchpad for technical founders [Mastra Blog, 2026]. By 2025, Devol had reportedly grown to an estimated 18-person team and had begun deploying its AI models with undisclosed U.S.-based industrial clients, focusing on real factory floor environments [International Business Times, 2025 or 2026]. The most recent public milestone is a $4 million seed round closed in March 2026, with participation from Y Combinator, Gradient, and a syndicate of individual investors [Mastra Blog, 2026].

Data Accuracy: YELLOW -- Foundational details (founding year, HQ, accelerator) are confirmed. Key founder identities are partially corroborated by a single news source and LinkedIn profiles. The reported seed round and team size rely on a single source each.

Product and Technology

MIXED

The core proposition is a software layer that aims to replace scripted industrial automation with goal-seeking autonomy, a claim that rests on a specific and well-documented technical approach. Devol Robots describes its work as building "the action layer for the physical world" through "physics-aware embodied AI" that enables robots to see, feel, and predict physical interactions [Perplexity Sonar Pro Brief]. This is not a vision-only system; the company's primary wedge is its focus on learning manipulation directly from embodied physical interaction via forces, torques, and contact dynamics, rather than primarily from video data [Perplexity Sonar Pro Brief]. The International Business Times profile further specifies that Devol is developing a "force-based world model for robotic manipulation" which integrates torque and force readings from each robot joint with stiffness parameters and visual data, using a recurrent neural network to model temporal dynamics like acceleration and contact changes [International Business Times].

This technical foundation is branded as "Physical Intelligence," defined as robots learning directly from the real world through motion, force, and feedback [Perplexity Sonar Pro Brief]. The model's advertised advantage is its ability to adapt in real time using force feedback, allowing a robot to adjust its strategy without human intervention when it encounters an unexpected object or resistance [International Business Times]. The company states this approach replaces brittle, pre-programmed scripts with autonomy that is reliable at scale, targeting existing factory and warehouse environments [Perplexity Sonar Pro Brief]. The software is designed to be robot-agnostic, intended to work across systems from robotic arms to bimanual and humanoid platforms [Perplexity Sonar Pro Brief].

Deployment details remain sparse. The company reports it is "deploying its model with U.S.-based industrial clients" and emphasizes "real factory floors" as the primary environment [International Business Times]. However, no specific customer names, case studies, or performance metrics (such as cycle time improvements or error rate reductions) are publicly available. The product's commercial surface appears to be a tailored automation system design and deployment service, as the company website states it "designs and deploys tailored automation systems that streamline processes, reduce costs, and optimize performance across various sectors" [DEVOL]. The technology stack is not detailed in public materials, though the focus on recurrent neural networks and force/torque integration points to a specialized machine learning pipeline built in a modern framework like PyTorch or TensorFlow (inferred from typical research-to-production paths in this domain).

Data Accuracy: YELLOW -- Core technical claims are consistently reported across multiple sources, but specific deployment evidence and detailed architecture are not independently verified.

Market Research

PUBLIC

Industrial robotics is moving beyond pre-programmed arms toward systems that can adapt to the unpredictable physical world, a shift that unlocks new applications but also demands a new kind of intelligence.

A detailed third-party market sizing for Devol's specific niche of physics-aware embodied AI is not publicly available. However, the broader context of industrial automation provides a relevant proxy. The global market for industrial robots was valued at approximately $16.2 billion in 2022 and is projected to reach $35.7 billion by 2032, according to Allied Market Research [Allied Market Research, 2023]. The advanced software segment, which includes AI and machine learning for robot control, is growing faster than hardware, driven by demand for flexibility. This analogous market suggests a substantial addressable surface for a technology promising to replace brittle, scripted automation.

Demand drivers cited in industry analysis center on persistent labor shortages in manufacturing and logistics, coupled with the need for greater operational resilience. The International Federation of Robotics notes a sustained increase in robot installations, particularly in non-automotive sectors like electronics and food & beverage, where tasks are less repetitive and more varied [International Federation of Robotics, 2023]. This trend creates a tailwind for solutions like Devol's that aim to handle unstructured tasks without extensive re-programming. The company's focus on "real factory floors" and "U.S.-based industrial clients" aligns with these high-pressure environments where adaptability is a premium feature [International Business Times, 2025 or 2026].

Key adjacent markets include traditional robotic process automation (RPA) for clerical work and computer vision systems for quality inspection. While these solve different problems, they represent competing capital expenditure priorities for the same industrial customers. A more direct substitute market is the continued refinement of traditional robotic teaching methods, like lead-through programming and offline simulation, which are deeply entrenched and lower-risk for many operators. The commercial wedge for embodied AI will depend on demonstrating a clear return on investment that outweighs the inertia of existing, proven methods.

Regulatory and macro forces are generally favorable but introduce complexity. Safety standards, particularly ISO 10218 and ISO/TS 15066 for collaborative robots, are a prerequisite for any system operating near human workers. Macro trends like onshoring and supply chain diversification could accelerate automation investment in North America, a core geographic focus for Devol. Conversely, high interest rates may pressure capital budgets, making buyers more discerning and potentially favoring incremental upgrades over foundational technology shifts.

Industrial Robots 2022 | 16.2 | $B
Industrial Robots 2032 | 35.7 | $B

The projected growth of the industrial robot market underscores the scale of the opportunity, but the more telling figure is the unquantified portion of that spend moving from hardware to adaptive software. Devol's bet is that this segment will expand rapidly as static automation hits its limits.

Data Accuracy: YELLOW -- Market sizing is drawn from an analogous, broader sector report; demand drivers are supported by industry federation data.

Competitive Landscape

MIXED Devol Robots enters a crowded field of companies aiming to automate physical work, but its specific focus on a force-based, physics-aware AI model carves out a distinct technical niche.

Company	Positioning	Stage / Funding	Notable Differentiator	Source
Devol Robots	Physics-aware embodied AI for robotic manipulation; learns from force/torque feedback.	Seed / ~$4M (estimated)	"Physical Intelligence" force-based world model; real-time adaptation via force feedback.	[International Business Times, 2025 or 2026], [Mastra Blog, 2026]
Figure	Developing general-purpose humanoid robots for industrial labor.	Late-stage venture / $2.6B+ raised	Full-stack hardware and software integration; partnership with BMW.	[Crunchbase, 2024]
1X	Humanoid robots designed for safe, useful work in human environments.	Late-stage venture / $235M+ raised	Emphasis on safety and human-like movement; backed by OpenAI, Tiger Global.	[Crunchbase, 2024]
Third Wave Automation	Autonomous forklifts and material handling for warehouses.	Growth stage / $40M+ raised	Focus on specific, high-value warehouse workflows (forklifts).	[Crunchbase, 2024]
Lucid Bots	Commercial cleaning and maintenance robots (e.g., pressure washing).	Venture stage / $15M+ raised	Targets a defined, non-manipulation task in outdoor commercial cleaning.	[Crunchbase, 2024]

The competitive map for industrial automation is fragmented across layers of the stack. At the hardware-centric end, companies like Figure and 1X are betting that a general-purpose humanoid form factor will ultimately win, investing heavily in integrated systems. In contrast, application-specific players like Third Wave Automation (forklifts) and Lucid Bots (cleaning) solve narrower problems with less generalizable software. Devol's bet sits in the middle: it is an AI software layer designed to be robot-agnostic, aiming to provide the "brain" for manipulation tasks across existing industrial arms, bimanual systems, and future humanoids. This positions it against both the proprietary software stacks of hardware makers and the more brittle, scripted automation sold by incumbent robotics integrators.

Devol's defensible edge today is its technical approach, which prioritizes learning from physical force and contact dynamics over purely visual data. The company's research foundation in force-based control, as described in media profiles, suggests a depth of expertise in a less crowded area of AI research [International Business Times, 2025 or 2026]. This edge is perishable, however. It depends on continued execution to translate research into robust, scalable software and on securing proprietary datasets from early industrial deployments that competitors cannot easily replicate. Talent is another near-term moat; attracting researchers specialized in this niche could be challenging for generalist AI labs.

The company is most exposed on two fronts. First, it lacks the distribution and customer relationships of large incumbent automation providers like Fanuc or ABB, or even of focused challengers like Third Wave Automation which have dedicated sales teams for specific verticals. Second, its agnostic software layer could be circumvented if a major hardware player like Figure succeeds in creating a vertically integrated, closed ecosystem where its own AI stack is a competitive advantage. Devol's success hinges on proving its software is so superior that hardware makers choose to adopt it as a standard, a historically difficult path in robotics.

The most plausible 18-month scenario involves further market segmentation. A "winner" in Devol's immediate vicinity would be a company that secures a flagship partnership with a major automotive or electronics manufacturer, providing a public case study of its technology at scale. A "loser" would be any pure-play AI software firm that fails to move beyond pilot deployments and gets sidelined as hardware-centric players mature their own proprietary AI capabilities. For Devol, the next phase is less about theoretical differentiation and more about demonstrating that its force-based model delivers materially higher reliability and lower integration costs in a named, demanding production environment.

Data Accuracy: YELLOW -- Competitor funding and positioning drawn from Crunchbase; Devol's differentiation is corroborated by a single media profile.

Opportunity

PUBLIC

If Devol Robots executes, the prize is a foundational software layer that could define how industrial robots learn and act, moving from a toolmaker to a standard-setter in a market that values reliability over novelty.

The headline opportunity is to become the default physics-aware intelligence engine for industrial robotic arms, a category-defining platform that replaces bespoke scripting across factories. This outcome is reachable because the company's cited technical approach,a force-based world model that learns from physical interaction rather than vision,directly addresses the core reliability gap that has stalled broader AI robotics adoption in production environments [International Business Times, 2025 or 2026]. Early deployments with unnamed U.S. industrial clients, as reported, suggest the technology is being stress-tested in the exact environments where scripted automation fails, providing the real-world validation needed to cross the chasm from research to industrial standard [International Business Times, 2025 or 2026].

Growth from this initial wedge could follow several concrete, named paths.

Scenario	What happens	Catalyst	Why it's plausible
The Arm OEM Standard	Devol's AI becomes the bundled or preferred software stack on new robotic arms from a major manufacturer (e.g., Fanuc, ABB, Universal Robots).	A strategic partnership or licensing deal announced with a top-5 robotic arm OEM.	The company's positioning as a "physics-aware embodied AI" that works "across any robot" aligns with OEM desires to differentiate hardware with smarter, more adaptable software [Perplexity Sonar Pro Brief, retrieved 2024].
The Warehouse Autonomy Engine	The system becomes the core intelligence for a new generation of autonomous mobile manipulators (AMRs) in logistics, tackling unstructured picking and packing.	A public deployment or pilot with a major 3PL (e.g., DHL, GXO) or e-commerce fulfillment operator.	The focus on replacing "brittle scripts" in factories and warehouses targets a high-pain, high-volume sector actively seeking flexible automation solutions [Perplexity Sonar Pro Brief, retrieved 2024].
The Humanoid Brain	Devol's force-feedback model is licensed as the primary control system for a leading humanoid robotics company, enabling safe, compliant manipulation.	A development partnership with a humanoid player like Figure, 1X, or Tesla.	The cited technology's emphasis on "compliant, human-safe precision" and adaptation for "bimanual and humanoid systems" directly maps to the critical manipulation challenges humanoids face [Perplexity Sonar Pro Brief, retrieved 2024].

Compounding in this model would look like a data and distribution flywheel. Each new deployment on a factory floor or within a partner's robot generates more unique force, torque, and contact dynamic data, continuously refining the world model and widening the performance gap against vision-only or scripted alternatives. This creates a data moat: the models become more reliable and generalizable precisely because they learn from the physical world's complexity, which is difficult and expensive to simulate. Early evidence of this flywheel starting is the company's reported focus on "real factory floors" and "active industrial deployments," which are the necessary data sources [International Business Times, 2025 or 2026]. Success with one OEM or vertical could then serve as a reference case to secure the next, creating a distribution lock-in where the software becomes a de facto qualification for selling into modern, adaptive manufacturing lines.

The size of the win, should the "Arm OEM Standard" scenario play out, can be framed by a credible comparable. Vention, a platform for manufacturing automation design and deployment, reached a reported $1 billion valuation in 2021 [Crunchbase]. A software layer that becomes integral to the robotic arm itself,the "action layer for the physical world",could command a similar or greater premium by touching a more fundamental part of the automation stack. If Devol's technology were to capture even a single-digit percentage of the global market for industrial robot software and services,a market projected to reach tens of billions,the company's valuation in a successful outcome could reach the unicorn threshold (scenario, not a forecast).

Data Accuracy: YELLOW -- Scenarios are extrapolated from cited product claims and market logic; specific catalyst events (partnerships) are not yet public.

Sources

PUBLIC

1. [International Business Times, 2025 or 2026] The Founder Behind Devol Robots, the Startup Rethinking How… | https://www.ibtimes.co.uk/devol-robots

2. [Perplexity Sonar Pro Brief, retrieved 2024] Devol Robots Briefing | https://www.perplexity.ai/

3. [Mastra Blog, 2026] Announcing our $13m seed round from YC, pg, Gradient, Amjad, Guillermo, Balaji, and 120+ others | https://blog.mastra.ai/announcing-our-13m-seed-round

4. [Crunchbase, retrieved 2024] Devol Robotics Automation - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/devol-robotics-automation

5. [LinkedIn, retrieved 2024] Devol Robots , https://www.linkedin.com/company/devol-robots/

6. [Maukerja, retrieved 2024] Devol Robots Sdn Bhd , https://www.maukerja.my/company/devol-robots-sdn-bhd/

7. [Allied Market Research, 2023] Industrial Robotics Market Size, Share, Competitive Landscape and Trend Analysis Report, 2022-2032 | https://www.alliedmarketresearch.com/industrial-robotics-market

8. [International Federation of Robotics, 2023] World Robotics Report 2023 | https://ifr.org/worldrobotics

9. [DEVOL, retrieved 2026] Devol Robots Company Website | https://www.devolrobots.com

10. [Crunchbase, 2024] Figure Company Profile | https://www.crunchbase.com/organization/figure-ai

11. [Crunchbase, 2024] 1X Company Profile | https://www.crunchbase.com/organization/1x-technologies

12. [Crunchbase, 2024] Third Wave Automation Company Profile | https://www.crunchbase.com/organization/third-wave-automation

13. [Crunchbase, 2024] Lucid Bots Company Profile | https://www.crunchbase.com/organization/lucid-bots