Trend analysis: Digital twins, simulation & synthetic data — where the value pools are (and who captures them)

The global digital twin market reached EUR 48.2 billion in 2025 and is projected to exceed EUR 130 billion by 2030, but value capture remains radically uneven. Platform providers and data infrastructure companies are accumulating outsized returns while application-layer vendors and consulting integrators face compressing margins. For European investors evaluating this space, the critical question is not whether digital twins represent a large market, that is no longer debatable, but where durable competitive advantages form, which segments generate the highest returns on invested capital, and how the convergence of simulation technology with synthetic data and generative AI reshapes the value chain over the next 12 to 24 months.

Why It Matters

Digital twins have moved beyond proof-of-concept into operational deployment across manufacturing, energy, infrastructure, and urban planning, with European companies leading adoption in several verticals. Siemens operates over 500,000 active digital twins across its industrial customer base. The European Commission's Destination Earth initiative committed EUR 315 million to build a full-fidelity digital replica of the planet for climate and environmental simulation. The UK's National Digital Twin Programme, coordinated by the Centre for Digital Built Britain, is establishing interoperability standards that will shape how infrastructure twins communicate across organisational boundaries.

The economic significance is substantial. McKinsey estimates that digital twin adoption in manufacturing alone could generate EUR 100 to 150 billion annually in value through predictive maintenance, process optimization, and accelerated product development. In the energy sector, DNV documented that offshore wind operators using digital twins reduced unplanned downtime by 30 to 45% and extended asset life by 5 to 8 years, translating to EUR 4 to 7 million in additional lifetime value per turbine installation.

The synthetic data dimension adds another layer of strategic importance. Privacy regulations including GDPR and the EU AI Act create structural demand for synthetic data that preserves statistical utility while eliminating personal information. The synthetic data market is growing at 35% annually and is expected to reach EUR 3.5 billion by 2028. The convergence with digital twins, using simulated environments to generate training data for AI models, represents one of the most significant value creation opportunities in the current technology landscape.

For investors, the timing is critical. The digital twin ecosystem is transitioning from a fragmented early market dominated by custom implementations toward platform consolidation, where a small number of horizontal platforms will capture the majority of infrastructure-layer value while vertical-specific applications compete on domain expertise and data access. Understanding this transition is essential for capital allocation decisions over the next 12 to 24 months.

Key Concepts

Digital Twin Architecture Layers

A digital twin technology stack comprises four distinct layers, each with different economic characteristics. The connectivity layer (IoT sensors, edge computing, data ingestion) has largely commoditised, with gross margins falling below 30% for hardware-centric players. The platform layer (data management, model orchestration, API infrastructure) commands the highest margins at 65 to 80%, exhibiting strong network effects and switching costs. The simulation layer (physics engines, finite element analysis, computational fluid dynamics) requires deep technical expertise and maintains moderate margins of 45 to 60%. The application layer (industry-specific solutions built on platforms) faces the most competitive pressure, with margins compressed to 25 to 40% by the growing capability of platform vendors to move up-stack.

Synthetic Data Generation

Synthetic data encompasses artificially generated datasets that statistically replicate real-world data distributions without containing actual measured observations. In the digital twin context, synthetic data serves three functions: training AI models where real operational data is scarce, sensitive, or biased; testing system responses to scenarios that have never occurred (stress testing, failure modes, extreme weather events); and validating digital twin accuracy against controlled synthetic benchmarks. Generative adversarial networks, variational autoencoders, and physics-informed neural networks represent the primary generation approaches, with physics-informed methods commanding premium pricing due to higher fidelity.

Simulation-as-a-Service

The shift from licensed simulation software to cloud-native simulation-as-a-service models is restructuring value capture in the simulation layer. Traditional simulation vendors (ANSYS, Dassault Systemes, Altair) generated revenue through annual perpetual licenses. The transition to consumption-based cloud models increases total addressable market by lowering entry barriers for small and medium enterprises, but also introduces competition from cloud hyperscalers offering GPU-accelerated simulation environments. European investors should note that simulation-as-a-service enables vertical specialisation, where domain experts build industry-specific simulation capabilities on generic cloud infrastructure, creating new investable niches.

Interoperability and Data Standards

Value in digital twin ecosystems is heavily influenced by data interoperability. Proprietary data formats create lock-in that benefits incumbent platform providers but restrict ecosystem growth. The European Commission's push for open data standards through the Data Governance Act and the Destination Earth initiative favours interoperable approaches. The buildingSMART International Foundation's IFC (Industry Foundation Classes) standard and the Digital Twin Consortium's open-source frameworks are establishing common vocabularies that, if widely adopted, could shift value from platform providers toward application-layer innovators.

Digital Twin Value Pool KPIs: Benchmark Ranges

Metric	Below Average	Average	Above Average	Top Quartile
Platform Gross Margin	<55%	55-65%	65-75%	>75%
Application Layer Gross Margin	<25%	25-35%	35-45%	>45%
Customer Retention (Annual)	<80%	80-88%	88-94%	>94%
Revenue per Customer (Enterprise)	<EUR 200K	EUR 200-500K	EUR 500K-1M	>EUR 1M
Net Revenue Retention	<100%	100-115%	115-130%	>130%
Time to Value (Months)	>18	12-18	6-12	<6
R&D as % Revenue	<15%	15-22%	22-30%	>30%

What's Working

Manufacturing Digital Twins

Manufacturing represents the most mature and profitable digital twin application domain in Europe. Siemens Xcelerator platform integrates product lifecycle management, factory simulation, and real-time production optimisation into a unified digital thread. BMW's Regensburg plant operates a comprehensive factory digital twin built on NVIDIA Omniverse that reduced production planning cycles from 30 months to 18 months and identified EUR 150 million in efficiency improvements across the production network. The value capture mechanism is clear: manufacturers pay for reduced time-to-market, lower defect rates, and optimised resource utilisation, with documented ROI ratios of 5 to 8x on digital twin investments within the first three years.

Energy Asset Performance Management

European energy companies lead digital twin adoption for wind, solar, and grid infrastructure. Akselos, acquired by BNFL in 2024, applies reduced-basis finite element analysis to create structural digital twins of offshore wind foundations that simulate decades of fatigue loading in minutes rather than weeks. Vattenfall's deployment across 12 offshore wind farms demonstrated a 35% reduction in structural inspection costs and a 22% decrease in unplanned maintenance events. The value pool here concentrates in asset life extension and availability optimisation, with European offshore wind operators spending an estimated EUR 8 billion annually on operations and maintenance that digital twins can materially reduce.

Synthetic Data for Autonomous Systems

European autonomous vehicle developers and industrial robotics companies increasingly rely on synthetic data generated from simulated environments to train perception and decision-making models. Wayve, the London-based autonomous driving company, uses simulation-generated synthetic data for 90% of its training pipeline, reducing the cost per training scenario from approximately EUR 400 (real-world data collection) to EUR 0.03 (synthetic generation). The European automotive simulation market, led by companies like dSPACE and IPG Automotive, is projected to reach EUR 4.2 billion by 2028. Synthetic data addresses both the data scarcity problem (rare edge cases that occur infrequently in real driving) and the regulatory compliance challenge (GDPR restrictions on capturing and processing pedestrian images).

What's Not Working

City-Scale Digital Twins

Despite substantial public investment, city-scale digital twins have struggled to deliver on their promise. The European Commission's Smart Cities programme funded 47 urban digital twin initiatives between 2020 and 2025, yet a 2025 evaluation found that only 8 achieved sustained operational use beyond their initial project funding. Common failure modes include fragmented data ownership across municipal departments, inability to maintain models as physical infrastructure changes, and the absence of clear revenue models to sustain operations. Helsinki's 3D city model, one of the most advanced in Europe, cost EUR 12 million to develop but generates limited direct economic value beyond planning visualisation. For investors, city-scale digital twins remain a pre-commercial category with high execution risk and unclear paths to sustainable returns.

Generic Horizontal Platforms Without Domain Expertise

Several venture-backed digital twin platforms raised significant capital on the premise that a single horizontal platform could serve all industries. The results have been disappointing. Companies that attempted to build industry-agnostic digital twin platforms without deep domain expertise have experienced extended sales cycles (12 to 24 months), high customer acquisition costs, and elevated churn as customers discover that generic platforms require extensive customisation to deliver industry-specific value. The successful model, demonstrated by Cognite (oil and gas, manufacturing) and Akselos (energy infrastructure), embeds domain expertise into the platform layer rather than relying on customers or systems integrators to bridge the gap.

Simulation Accuracy Without Continuous Calibration

Digital twins that diverge from physical reality over time represent a significant value destruction risk. A 2024 study published in the Journal of Manufacturing Systems analysed 180 manufacturing digital twins and found that 62% experienced accuracy degradation exceeding 15% within 12 months of deployment due to physical asset changes, sensor drift, and environmental variations not reflected in models. Without continuous calibration workflows, including automated sensor validation, model updating, and drift detection, digital twins become expensive misleading dashboards rather than decision-support tools. Investors should evaluate calibration capabilities as a core differentiator rather than a feature checkbox.

Myths vs. Reality

Myth 1: Digital twins require massive upfront investment before delivering value

Reality: The shift to cloud-native platforms and pre-built industry templates has reduced time-to-value dramatically. Cognite Data Fusion enables industrial customers to deploy initial digital twin use cases in 8 to 12 weeks at costs starting from EUR 150,000. Bentley Systems' iTwin platform offers consumption-based pricing that eliminates large capital expenditures. The most successful implementations start with a single high-value use case (typically predictive maintenance or energy optimisation) and expand incrementally as ROI is demonstrated. Investors should favour companies with land-and-expand business models over those requiring enterprise-wide deployment commitments.

Myth 2: More data always produces better digital twins

Reality: Data quality and relevance matter far more than volume. Physics-informed digital twins that combine sparse sensor data with first-principles models consistently outperform data-driven approaches that rely on massive datasets without physical constraints. Akselos demonstrated that structural digital twins using 12 strategically placed sensors achieved higher predictive accuracy than competitors using 200 sensors without physics-based models. Over-instrumentation increases costs, complexity, and failure points without proportional accuracy gains.

Myth 3: Synthetic data will replace real-world data entirely

Reality: Synthetic data excels at augmenting real datasets, particularly for rare events, edge cases, and privacy-sensitive applications. However, pure synthetic training without real-world calibration introduces systematic biases that degrade model performance in production environments. The "sim-to-real gap" remains a fundamental challenge: models trained exclusively on synthetic data typically underperform real-data-trained equivalents by 8 to 15% on deployment metrics. The optimal approach combines smaller volumes of high-quality real data with larger synthetic datasets, using domain randomisation and transfer learning techniques to bridge the fidelity gap.

Myth 4: Open-source digital twin platforms will commoditise the market

Reality: While open-source frameworks like Eclipse Ditto and Apache StreamPipes provide basic digital twin infrastructure, they address less than 20% of the functionality required for production deployments. Enterprise requirements including security, scalability, compliance, simulation integration, and continuous calibration create substantial differentiation opportunities that open-source alternatives have not addressed. The analogy to databases is instructive: open-source databases (PostgreSQL, MySQL) coexist with high-margin commercial database companies (Snowflake, Databricks) because enterprise requirements create willingness to pay for managed, integrated solutions.

Key Players

Established Leaders

Siemens Digital Industries operates the most comprehensive digital twin platform in Europe through Xcelerator, spanning product design, factory simulation, and operational optimisation. Annual revenue from digital twin-related software exceeded EUR 5.8 billion in fiscal 2025.

Dassault Systemes offers 3DEXPERIENCE, integrating simulation, digital manufacturing, and virtual twin experiences. Strong positioning in aerospace, automotive, and life sciences with 300,000 enterprise customers globally.

AVEVA (Schneider Electric subsidiary) provides industrial digital twin solutions for process industries, energy, and marine, with particular strength in continuous process manufacturing and offshore energy.

Emerging Startups

Cognite (Norway) raised USD 325 million at a USD 1.6 billion valuation for its industrial DataOps platform that powers digital twins for energy, manufacturing, and utilities. Strong net revenue retention above 130%.

Navvis (Germany) combines indoor mapping hardware with digital twin software for manufacturing and real estate, processing over 50 million square metres of indoor environments.

Mostly AI (Austria) leads European synthetic data generation, offering a platform that generates privacy-compliant synthetic datasets for financial services, healthcare, and telecommunications.

Key Investors and Funders

Accel led Cognite's Series C and has been actively investing in European industrial software companies with digital twin capabilities.

European Innovation Council allocated EUR 190 million to digital twin and simulation technology development between 2023 and 2026 through its Accelerator programme.

TCV invested in several European simulation and digital twin companies, focusing on the transition from licensed to consumption-based business models.

Action Checklist

• Map the digital twin value chain layer by layer (connectivity, platform, simulation, application) to identify where margin concentration occurs in target investment verticals
• Evaluate platform investments based on net revenue retention, customer expansion rates, and switching cost indicators rather than gross bookings growth
• Assess simulation companies for cloud-native architecture and consumption-based pricing models that expand total addressable market
• Prioritise investments in companies with embedded domain expertise over generic horizontal platform plays
• Investigate synthetic data companies serving regulated industries (financial services, healthcare, automotive) where GDPR and EU AI Act compliance creates structural demand
• Evaluate digital twin calibration and model management capabilities as core technical differentiators
• Monitor the European Commission's data interoperability standards development for signals on platform lock-in sustainability
• Consider the competitive dynamics between incumbent simulation vendors and cloud hyperscalers entering the simulation-as-a-service market

FAQ

Q: What is the realistic total addressable market for digital twins in Europe? A: The European digital twin market, including platforms, simulation software, synthetic data, and professional services, reached approximately EUR 15.8 billion in 2025. Manufacturing accounts for 32% (EUR 5.1 billion), energy and utilities 24% (EUR 3.8 billion), construction and infrastructure 18% (EUR 2.8 billion), automotive 14% (EUR 2.2 billion), and other sectors 12% (EUR 1.9 billion). Growth rates vary by segment: synthetic data leads at 35% CAGR, followed by platform software at 25%, simulation-as-a-service at 20%, and professional services at 12%.

Q: How should investors evaluate digital twin companies' competitive moats? A: The strongest moats in digital twins come from three sources. First, proprietary data network effects: platforms that aggregate and normalise operational data across multiple customers create compound value that improves model accuracy for all users. Second, domain-specific physics models: companies that embed deep engineering knowledge into reusable simulation templates create switching costs that generic platforms cannot replicate. Third, regulatory compliance infrastructure: companies whose platforms satisfy industry-specific certification requirements (nuclear, aerospace, medical devices) benefit from lengthy qualification cycles that deter new entrants.

Q: What is the investment case for synthetic data companies in Europe? A: European synthetic data companies benefit from structural regulatory tailwinds that are stronger than in any other geography. GDPR restricts the use of personal data for AI training, the EU AI Act mandates bias testing that requires diverse representative datasets, and sector-specific regulations in financial services and healthcare further constrain data availability. Mostly AI (Austria), SYNTHO (Netherlands), and Hazy (UK) represent the leading European pure-play synthetic data companies, with Mostly AI commanding the largest customer base at over 100 enterprise clients. The primary risk is absorption by platform providers; both Google and Microsoft have developed synthetic data capabilities that could reduce standalone market opportunity.

Q: What are the key risks for digital twin investments over the next 24 months? A: Three primary risks warrant attention. First, hyperscaler competition: AWS IoT TwinMaker, Azure Digital Twins, and Google Cloud's Vertex AI are expanding platform capabilities that could compress margins for independent digital twin platforms. Second, standards fragmentation: if interoperability standards fail to converge, market growth could be constrained by integration complexity and buyer hesitation. Third, ROI verification: as the market matures, customers will demand rigorous ROI documentation, and companies that have relied on aspirational value propositions without measurable outcomes will face churn and reputational challenges.

Q: How does the convergence of generative AI and digital twins affect the investment landscape? A: Generative AI is reshaping digital twins in three ways. First, natural language interfaces allow non-technical users to query and interact with digital twins, dramatically expanding the user base within customer organisations. Second, generative models can automatically construct digital twin geometries and physics models from photographs, laser scans, or text descriptions, reducing creation costs by 60 to 80%. Third, foundation models trained on simulation data can transfer knowledge across domains, potentially enabling a single pre-trained model to initialise digital twins for novel applications with minimal fine-tuning. Companies integrating generative AI capabilities into their digital twin platforms are attracting premium valuations, with multiples 30 to 50% higher than non-AI-augmented competitors.

Sources

• McKinsey Global Institute. (2025). Digital Twins: Bridging the Physical and Digital Worlds in European Industry. Munich: McKinsey & Company.
• European Commission. (2025). Destination Earth: Progress Report and Economic Impact Assessment. Brussels: EC Publications.
• DNV. (2025). Digital Twins for Offshore Wind: Operational Performance and Economic Value Analysis. Oslo: DNV.
• Grand View Research. (2025). Digital Twin Market Size, Share & Trends Analysis Report, 2025-2030. San Francisco: Grand View Research.
• Journal of Manufacturing Systems. (2024). Accuracy Degradation in Manufacturing Digital Twins: A Multi-Site Longitudinal Study. Elsevier, Vol. 72, pp. 312-328.
• European Innovation Council. (2025). Digital Twin Technology Investment Portfolio: Outcomes and Learnings. Brussels: EIC.
• BloombergNEF. (2025). Synthetic Data Market Outlook: European Growth Drivers and Competitive Landscape. London: Bloomberg LP.