Software-defined industrial automation: global opportunities and India's emerging role

Executive Summary

The global manufacturing landscape sits at a critical juncture. While companies worldwide invest some $34 billion annually in industrial robotics, with growth projected at 13.8 percent compound annual growth rate (CAGR) through 2034, reality on factory floors lags expectations.

Approximately 72 percent of manufacturing tasks globally still require human hands, and even advanced automotive plants automate less than 10 percent of operations. This gap exists not because automation technology lacks sophistication, but because traditional approaches require 15-36 months for design, validation, and deployment — timelines during which product specifications often change entirely.

Software-defined factories (SDFs) represent a reimagining of manufacturing systems as reconfigurable software platforms rather than rigid hardware installations. For Indian deep-tech entrepreneurs and investors, this shift opens a substantial opportunity: combining India's engineering talent depth and cost advantages with proximity to global research innovation and manufacturing customers to establish leadership in a market projected to reach as much as $84 billion by 2033.

CynLr's distributed model — research in Switzerland, execution in India, customers in America — provides a pragmatic template for how Indian startups can compete at the frontier of industrial automation.

POTENTIAL

Understanding the automation gap

Global manufacturing represents approximately 10 percent of world GDP and employs over 500 million workers across developed and developing economies. Despite technological advances over three decades, the sector remains remarkably labour-intensive. A report from Kearney notes that 72 percent of factory tasks are performed by humans, with even capital-intensive industries lagging significantly in automation adoption.

The report is from 2018, but remains one of the most cited findings on the topic. An automotive plant — the most mechanised manufacturing segment globally — typically has fewer than 10 percent of tasks automated, leaving 90 percent dependent on manual labour.

This paradox emerges from fundamental economic and engineering constraints. Traditional factory automation requires extensive customisation for each production environment. The typical timeline encompasses assessment and planning (8-12 weeks), design and engineering (8-24 weeks), procurement and installation (4-12 weeks), and commissioning and testing (6-24 weeks), with total timelines frequently extending to 15-36 months before production commences.

By this point, market demands have shifted, product designs have evolved, and the capital investment often fails to deliver promised flexibility.

The economics are compelling: Even a pre-Covid estimate, from a 2017 report by McKinsey Global Institute, reckons 64 percent of global manufacturing-related working hours (478 billion annually) are automatable, equivalent to $2.7 trillion in labour value. Yet the realisation of this potential requires a shift from hardware-centric to software-centric manufacturing systems.

Market size and growth trajectory

Despite the constraints mentioned above, the global industrial robotics market reflects underlying demand. Current market assessments vary by methodology, with estimates putting the market value at about $18 billion in 2024. It is projected to reach between $60 billion and $84 billion by 2033-2034, representing a CAGR of 9.9 percent to 13.8 percent.

Asia-Pacific dominates, capturing with about two-thirds of global revenue market share, with China, Japan, and emerging manufacturers in India and Southeast Asia driving adoption.

Within this market, material handling represents the largest application (42 percent of market share in 2024), followed by automotive manufacturing, electronics, and specialised industrial applications. The shift reflects not just automotive industry requirements, but growth in e-commerce and logistics automation, where companies like Amazon and Alibaba have deployed thousands of autonomous robots in warehouses and sorting facilities.

HUMANOIDS

Investment concentration

While industrial robotics delivers measurable results — improving throughput, reducing defects, enabling flexible manufacturing — investment capital increasingly flows toward humanoid robotics startups that remain in prototype and demonstration phases.

Figure AI's valuation reached $39 billion in 2024, representing a 15x increase from previous rounds, while the company is at fairly early commercial deployments.

This concentration of capital reflects broader venture capital behaviour: investors pursue narrative-driven investments in what they perceive as “platform” opportunities rather than following disciplined analysis of unit economics, revenue potential, or technological maturity.

Humanoid robotics holds out the allure of general-purpose automation — the promise that robots designed for human proportions will seamlessly integrate into human environments and perform diverse tasks without costly customisation. The narrative is compelling and captures venture capital momentum, particularly given the success of generative AI as an investment category.

Real-world constraints

Current humanoid systems face fundamental technical challenges that capital alone won’t solve. Leading humanoid platforms cost between $30,000 and $500,000 per unit, operate for only 2-4 hours on a single battery charge, move more slowly than human workers, and can lift 20-30 pounds comfortably. More critically, they struggle with inference (real-time decision-making), dexterity (precise object manipulation), and reliability in unstructured environments.

As Rodney Brooks, the roboticist who invented the Roomba, noted in 2025, current humanoid robots are unsafe for human interaction due to inadequate hand dexterity and gait control, with these limitations unlikely to resolve in the near term. CB Insights’ October 2025 report documented that while humanoid robotics startups secured 17 investment deals in a single quarter, these companies face “fundamental challenges with inference, dexterity, reliability, and cost, which limit initial use cases to structured environments like factories and warehouses with controlled and predictable sets of tasks.”

Critically, even industrial humanoid robots lack demonstrated commercial viability. Companies like Figure AI, Boston Dynamics, and others release demonstration videos showing robots performing tasks — running, boxing, handling objects — but these demonstrations occur in controlled environments with carefully prepared scenarios.

Real factory deployment requires systems that operate 16-24 hours daily, handle unexpected variations, maintain safety standards, integrate with existing factory IT infrastructure, and deliver measurable return on investment.

This gap between demonstration capability and operational deployment defines what investors call “timing risk’: the uncertainty about when or whether prototype-phase technologies transition to profitable mass deployment.

Industrial robots versus humanoid

Industry observers increasingly distinguish between industrial robotics (collaborative robots, mobile manipulators, material handling systems) and humanoid robotics. Industrial systems already generate revenue, deliver measurable productivity improvements, and integrate into manufacturing workflows. Humanoid systems remain almost entirely in prototype and pilot phases, with minimal revenue generation and unproven commercial economics.

As one venture capitalist observed, “While the same risks persist in humanoid robotics as in pre-revenue AI startups, many investors tend to overlook this.” Distinguishing between robotics and humanoid robotics is becoming important. Industrial and logistics robots already generate revenue and can deliver measurable results. Humanoids aren’t there yet.

SOFTWARE DEFINED FACTORIES

Core concept and architecture

Software-defined factories represent a fundamental departure from traditional manufacturing approaches. Rather than coupling control software tightly to specific hardware configurations, SDFs separate the software control layer from physical production hardware. This decoupling enables production processes to be modified through software updates rather than requiring expensive physical restructuring.

The technical foundation draws from principles pioneered in software-defined networking (SDN) and cloud computing. Manufacturing systems are reimagined as modular, hardware-agnostic platforms where production logic runs on standardised, reconfigurable hardware. A factory operator can modify production parameters — changing from screw-assembly to clip-assembly, for instance — through software configuration rather than replacing hardware, retraining robot positions, or redesigning mechanical systems.

TCS’s research on SDFs highlights the potential: “A Software Defined Factory (SDF) envisions a software layer that oversees machines, processes, workflows, and assets across the entire plant, enabling the execution of manufacturing processes with a single click.”

In collaborative software-defined ecosystems, manufacturers can boost productivity by 30-50 percent through rapid reconfiguration and optimised resource allocation.

Bosch Research, through the publicly funded SDM4FZI (Software-Defined Manufacturing for the vehicle and supplier industry) project, has demonstrated practical implementation of these concepts. Working at ARENA2036, a research campus in Stuttgart, Bosch engineers built a reconfigurable production module for automotive steering systems.

The system separates control logic from hardware configuration, allowing production process changes that previously required hardware redesign and extensive recalibration.

In one demonstration, changing a production step from screwing components to using clips required only software reconfiguration — no physical restructuring, no robot retraining, no control system replacement. This flexibility addresses a critical pain point for automotive suppliers managing product variants and demand variability.

As Bosch's Johannes Fisel noted, “The ARENA2036 ecosystem offers optimum conditions for the implementation of SDM.” The team transferred applications from traditional control systems (PLCs, DCS) to virtualised edge computing platforms, enabling central management of multiple machines through software rather than proprietary hardware controllers.

Industry 5.0 and market growth

The Industry 5.0 market, which encompasses software-defined manufacturing, AI-integrated systems, and human-machine collaboration, is projected to go from more than $71 billion in 2024 to nearly $890 billion by 2033, at a compound annual growth rate of 32.4 percent. This growth substantially exceeds traditional industrial automation forecasts, reflecting recognition that software-centric approaches unlock new capabilities beyond what hardware-locked systems can achieve.

India's factory automation market specifically was valued at $7.7 billion in 2024 and is projected to reach $13.7 billion by 2033 (7.4 percent CAGR), with automated assembly lines reaching $2.2 billion by 2033 (8.6 CAGR). More broadly, India's Industry 4.0 market – which differs from 5.0 in that it doesn’t incorporate human-machine collaboration – reached $5.5 billion in 2024 and is projected to reach $26.7 billion by 2033.

CynLr’s model

CynLr, founded in 2019 by Gokul NA and Nikhil Ramaswamy, emerged from a decade of direct manufacturing observation. Before founding the company, the cofounders spent four years consulting in machine vision and robotics, working alongside manufacturers to solve real automation challenges.

This ground-level experience yielded a crucial insight: the primary barrier to automation is not robot aesthetics, but robot capabilities — specifically, the ability to perceive and manipulate unfamiliar objects without exhaustive pre-programming.

CynLr's product philosophy prioritises utility over form factor. Rather than pursuing humanoid designs, the company developed a three-arm robotic platform specifically engineered for manufacturing tasks. The system learns in real time, handling objects of varying shapes, colours, and materials without requiring exhaustive pre-programming.

Partners and customers include Denso, which requires robots managing demand variability across different part types through hot-swappable station configurations, and General Motors, which needs a standardised platform capable of handling 22,000+ different parts for vehicle assembly.

In November 2024, CynLr secured $10 million in Series A funding led by Pavestone and Athera Venture Partners (formerly Inventus India), bringing total funding to $15.2 million. The capital infusion funded expansion across three dimensions: doubling the team from 60 to 120 engineers; expanding research, software development, business, and sales capabilities; and increasing manufacturing capacity with a target of deploying one robot system daily and achieving $22 million in revenue by 2027.

Notably, CynLr's funding discipline contrasts starkly with humanoid robotics venture valuations. The company raised modest capital relative to market opportunity focusing on solving concrete customer problems rather than pursuing speculative technology narratives. Apart from us, other existing investors have also participated, reinforcing confidence that the company's problem-first approach resonates with disciplined investors.

CynLr's strategic footprint across three geographies exemplifies how Indian deep-tech startups can compete at innovation frontiers while maintaining economic efficiency.

Bengaluru — execution and manufacturing:

The core engineering team, manufacturing operations, and integration work occur in Bengaluru, where CynLr manages a supply chain of 400+ parts sourced across 14 countries. This location decision reflects economic reality: software engineers in India cost approximately one-fifth to one-eighth of equivalent Swiss salaries, and manufacturing costs run 70-80 percent lower than Western countries. The Bengaluru lab housed 25 robotic systems in 2024, projected to grow to over 50 by 2026, providing testing grounds for customer pilots and production scaling.

Switzerland — advanced research and foundational innovation:

In September 2024, CynLr established a Design & Research Centre in Prilly, Switzerland, at the Unlimitrust Campus. This facility positions the company alongside EPFL's Learning Algorithms and Systems Laboratory (LASA) and CSEM (Centre Suisse d'Électronique et de Microtechnique).

LASA, directed by Aude Billard — President of the IEEE Robotics and Automation Society — conducts frontier research in robot learning from demonstration, dexterous manipulation, and human-robot interaction. CynLr hired Dr. Michael Bombile, who spent a decade at LASA under Billard's direction, to lead research initiatives in perception, vision algorithms, and advanced manipulation.

This Swiss presence serves multiple functions: anchoring CynLr within the world's leading robotics research ecosystem; providing access to researchers solving problems at the frontier of the field; facilitating collaboration with academic institutions and industrial partners across Europe; and signalling technological credibility to Western customers and partners.

United States — customer relationships and market access:

Co-founder Nikhil Ramaswamy relocated to America to establish customer-facing operations focused on the US automotive market — one of the company's largest and most demanding revenue sources. This presence enables direct engagement with customers like General Motors and Denso's US operations; proximity to American venture capital, corporate innovation teams, and technology partnerships; and positioning within the North American manufacturing ecosystem.

INDIA

Employment scale and labour

India's manufacturing sector employs 18.5 million workers in formally registered factories (FY 2022-23) according to the Annual Survey of Industries (ASI), growing 7.5 percent year-over-year—the highest growth rate in 12 years. Broader estimates including informal manufacturing and contract workers suggest the figure may exceed 45-70 million workers, depending on definitions.[26][27]

These workers are concentrated in manufacturing hubs: Tamil Nadu, Maharashtra, Gujarat, Uttar Pradesh, and Karnataka account for 55 percent of formal manufacturing employment. These same states host the automotive, electronics, pharmaceutical, and consumer goods facilities where software-defined automation would deliver greatest impact.

Notably, contract labour's share in manufacturing employment has doubled from 20 percent in 1999-2000 to 40.7 percent in 2022-23, reflecting growing operational flexibility pressures on manufacturers. This shift creates paradoxical incentives: on the one hand, easy availability of contract labour has reduced pressure for automation; on the other, labour cost increases and productivity demands are driving serious automation consideration, particularly in capital-intensive sectors like automotive.

Robotics startups in India

Indian robotics startups collectively raised $117 million across 41 deals in 2024, representing four-fold growth from $28.8 million in 2022 and more than doubling from $54 million in 2023. This trajectory mirrors global venture capital enthusiasm, but with distinctly different characteristics: Indian funding increasingly focuses on revenue-generating industrial applications rather than speculative technologies.

Notable fundings in 2024 included Ati Motors’ $20 million Series B (led by Walden Catalyst Ventures and NGP Capital) for autonomous material-handling robots; Niqo Robotics' $13 million raise for agricultural robotics; and CynLr's $10 million Series A. The diversity of applications — construction, material handling, agriculture, industrial manipulation — indicates market breadth rather than concentration in a single segment.

More importantly, Indian startups are competing on problem-solving rather than hype. As an Economic Times report notes, “Robot adoption is no longer about cost. That is an important realisation.” Cost structure advantages matter, but competitive differentiation increasingly derives from solving specific manufacturing challenges — demand variability, part complexity, integration with legacy systems — rather than claiming general-purpose capability.

Public support

The Indian government has explicitly prioritised deep-tech development. MeitY's Global Startup Bridge, launched in 2025, aims to connect Indian deep-tech startups with international R&D ecosystems, policymakers, and corporate partners, with chapters established in Brussels, Paris, and planned for other innovation hubs. The program facilitates startup exchange, innovation cooperation, and market access in Europe and beyond.

The Rs. 10,000 crore ($1.2 billion) Fund of Funds announced in April 2025 targets deep-tech companies working in quantum computing, robotics, semiconductors, and related domains, with explicit emphasis on patient capital and foundational thinking rather than trend-following. National policy documents establish a target of 25 percent manufacturing contribution to GDP by 2025 (up from current ~17 percent), requiring substantial productivity improvements potentially achievable through automation.

The National Manufacturing Policy, combined with the Production-Linked Incentive (PLI) scheme covering 14 key sectors, creates supportive infrastructure for companies building advanced manufacturing capabilities. Regional hubs like Bengaluru, Pune, Chennai, Hyderabad, and Mumbai now compete as innovation ecosystems, with companies like Bosch, ABB, Tata Consultancy Services, and Infosys deploying AI and automation solutions.

GOING GLOBAL

CynLr’s Swiss partnership

EPFL's Learning Algorithms and Systems Laboratory represents the global frontier of robot learning and manipulation research. Under Aude Billard's direction, LASA has pioneered methods enabling robots to learn control laws from human demonstration, acquire dexterous bimanual manipulation skills, and adapt to novel environments through dynamical systems approaches.

Billard's research on “programming by demonstration” allows robots to learn task execution from human examples, a capability fundamentally different from traditional reprogramming approaches. Her work on learning stable nonlinear dynamical systems has been cited over 900 times, indicating widespread academic adoption. Recent research on robot learning from demonstration and the application of learned controllers to new task variants directly addresses the flexibility challenges that CynLr targets.

This research ecosystem provides tangible competitive advantage: startups embedded within or adjacent to EPFL gain access to researchers solving decade-ahead problems, collaborate on problem formulation, and benefit from institutional credibility. For CynLr specifically, hiring Dr. Michael Bombile — an EPFL researcher with direct experience in manipulation and learning — enabled the company to remain at the frontier of perception and control research while maintaining commercial focus.

From brain drain to networks

Historically, India experienced significant “brain drain’ — highly educated professionals emigrating to developed countries for career opportunities and higher compensation. Over the past decade, this pattern has shifted towards more networked opportunities where talented individuals maintain connections to India while pursuing careers globally, or return to India while maintaining international networks.

India's Global Capabilities Centres (GCCs) — high-value research and development operations of multinational companies — grew from 1 million to 1.6 million jobs between 2018 and 2024, indicating that world-class R&D increasingly occurs in India. Simultaneously, Indian engineers and founders increasingly operate distributed, with dual homes or rotating presence across innovation hubs.

This shift benefits Indian startups: founders can hire talent from global universities, maintain ties to international research institutions, and build teams that span multiple geographies while retaining cost advantages of India-based operations. CynLr's model — hiring Dr. Bombile from EPFL while maintaining core teams in Bengaluru — exemplifies this pattern.

BARRIERS

Technical challenges

Despite progress, fundamental challenges remain unsolved. Real-time perception in unstructured manufacturing environments—the ability for robots to identify, classify, and grasp objects of varying shapes, materials, and colours without extensive pre-training—remains an active research frontier. Safe human-robot collaboration in existing factory layouts, where humans and robots share workspaces, requires advances in force sensing, compliance control, and real-time safety prediction.

Scaling real-time learning from limited robot experiences to diverse environments and task variations demands advances in transfer learning, meta-learning, and few-shot adaptation — problems that remain partially solved. Reliability in unstructured, noisy factory environments—where vibrations, dust, lighting variations, and material changes occur continuously — differs fundamentally from lab demonstrations.

Commercialisation and integration

Beyond research, commercialisation barriers are substantial. Integrating software-defined robotic systems with legacy factory IT infrastructure (ERPs, MES, sensors, PLCs) requires deep domain knowledge of specific manufacturing environments. Customers must be educated on new operational paradigms; many manufacturers remain conservative in adopting fundamentally different automation approaches.

Skilled technician shortages persist for deployment, maintenance, and troubleshooting of advanced robotic systems. While India produces 1.5 million STEM graduates annually, robotics-specific expertise remains concentrated in select institutions and companies. Training networks and maintenance ecosystems require years to develop.

Customer risk aversion compounds these challenges. Manufacturing executives are evaluated on operational continuity, not technology advancement. Deploying novel robotic systems introduces perceived risk: What if the system fails? How is it repaired? What if the software crashes? These questions—entirely reasonable from an operations perspective — slow adoption despite clear productivity upside.

Realistic adoption timeline

Despite enthusiasm, adoption will likely occur gradually. Pilot deployments and early customer reference installations should emerge during 2025-2027. Initial revenue generation will be limited to customers with acute automation needs, high product variability, and tolerance for emerging technology risks. These early customers will provide operational feedback, reference cases, and case studies enabling broader adoption.

Mainstream adoption in manufacturing — where software-defined systems become standard practice rather than innovation — likely occurs during 2028-2030, assuming continued progress in perception, manipulation, and control algorithms, and given policy tailwinds accelerating manufacturing automation in India, Europe, and North America.

Large-scale deployment reaching 50 percent+ of production facilities would extend to 2031-2035, still within reach of companies founded in the 2015-2020 period.

CONCLUSION

Software-defined factories represent neither revolution nor distant future. They constitute an incremental but consequential evolution from hardware-locked manufacturing to flexible, software-configurable systems. This evolution addresses genuine pain points: changing product designs in response to market demand, managing production complexity without extensive customisation, reducing time-to-production for new variants.

For Indian entrepreneurs, investors, and policymakers, the timing aligns unusually well with structural advantages:

Cost structure:

Indian software engineers, manufacturers, and systems integrators operate at 70-80 percent cost advantage relative to European and American equivalents, extending runway and enabling patient capital strategies that exceed venture norms.

Talent depth:

1.5 million STEM graduates annually, with growing concentrations in robotics, embedded systems, and machine learning, provide recruitment advantages global competitors lack.

Domestic market scale:

18.5-70 million manufacturing workers, distributed across automotive, electronics, pharmaceuticals, and consumer goods hubs, provide massive opportunities for automation adoption and local learning.

Geopolitical tailwinds:

Supply chain diversification away from China, “China+1” strategies, and friendly-shoring initiatives create demand for manufacturing alternatives in India, creating a runway for automation companies to establish reference customers and case studies.

Global research access:

MeitY's Global Startup Bridge, India-Japan Emerging Tech Corridor, and similar initiatives facilitate partnership with leading research institutions while maintaining India-based execution.

The future of Indian manufacturing automation is unlikely to belong to humanoid robots or general-purpose systems, but to companies solving specific problems rigorously: flexible part handling, demand variability management, integration with existing factory systems, cost-effective deployment at scale. CynLr is but one example.

For founders willing to pursue problem-solving over narrative, investors comfortable with distributed operations and patient capital, and policymakers supporting research partnerships and deep-tech development, the next decade offers India the possibility of establishing global leadership in software-defined manufacturing.

This would represent not a transformation of manufacturing, but real value-generating evolution from rigid systems toward flexible, learning-capable platforms that humans and machines operate in partnership.