Deep dive: Digital twins, simulation & synthetic data — the hidden trade-offs and how to manage them

The global digital twin market surged from $17.73 billion in 2024 to an estimated $24.48 billion in 2025—a 38% year-over-year leap that underscores an uncomfortable truth: organizations are deploying these technologies faster than they can measure their full environmental cost. Meanwhile, Gartner projects that by 2025, 60% of data used for AI and analytics projects will be synthetically generated, raising critical questions about the hidden carbon footprint of the very tools we deploy to reduce emissions. This deep dive examines the tension between operational sustainability gains and computational overhead, providing actionable frameworks for teams navigating these trade-offs.

Why It Matters

Digital twins, simulation platforms, and synthetic data generation represent a $25+ billion ecosystem poised to reshape sustainability strategy across manufacturing, energy, buildings, and supply chain management. The technology promises substantial environmental benefits: studies from the World Economic Forum and NVIDIA demonstrate that digital twins can reduce building energy consumption by 10–30%, cut maintenance costs by 20–40%, and enable predictive interventions that extend asset lifecycles by years.

Yet these gains come with a paradox. The computational infrastructure required to run real-time digital twins—continuous data ingestion, physics-based simulations, and AI inference—consumes significant energy. A single large-scale industrial digital twin processing data from thousands of IoT sensors can require GPU clusters that generate excess heat demanding energy-intensive cooling. When organizations deploy synthetic data generation to train AI models for climate scenario analysis, they often overlook that each training run has its own carbon footprint.

For sustainability practitioners, this creates a fundamental measurement challenge: how do you account for the emissions produced by the technology that measures emissions? The answer lies in understanding where net-positive impact occurs and where computational overhead erodes environmental gains.

The stakes are particularly high given regulatory momentum. The EU's Corporate Sustainability Reporting Directive (CSRD) and the SEC's climate disclosure rules require increasingly granular Scope 3 reporting. Digital twins offer a pathway to the real-time, auditable data these frameworks demand—but only if organizations can demonstrate that the tool itself aligns with decarbonization goals.

Key Concepts

Digital Twins are virtual replicas of physical assets, processes, or systems that synchronize with real-world data to enable simulation, monitoring, and optimization. In sustainability contexts, they range from building energy management systems tracking HVAC performance to industrial process twins modeling carbon-intensive manufacturing lines. The critical distinction from traditional simulation is bidirectionality: changes in the physical asset update the twin, and insights from the twin inform physical interventions.

Simulation Platforms extend digital twins by enabling "what-if" scenario analysis without physical experimentation. A wind farm operator might simulate blade configurations across 10,000 weather scenarios to optimize energy capture, or a city planner might model traffic patterns under various EV adoption rates. The sustainability value lies in avoiding physical prototyping waste—but each simulation run consumes compute resources.

Synthetic Data Generation creates artificial datasets that statistically mirror real-world data distributions while protecting privacy and filling data gaps. For climate applications, synthetic data enables stress-testing portfolios against extreme weather scenarios, training autonomous vehicle perception systems for edge cases, and augmenting sparse datasets for nature-based carbon credit validation. The market grew from $576 million in 2024 to approximately $765 million in 2025, with projections exceeding $6 billion by 2032.

Life Cycle Assessment (LCA) Integration represents the emerging practice of embedding environmental impact calculations directly into digital twin architectures. Rather than conducting periodic LCA studies, organizations can track product carbon footprints continuously as supply chain data flows through the twin.

Edge-Cloud Hybrid Architectures address computational trade-offs by distributing processing between on-site edge devices and centralized cloud infrastructure. This approach can reduce data transmission energy by 40–60% while maintaining real-time responsiveness for time-critical applications.

What's Working

Operational Energy Reduction at Scale

The evidence base for digital twins reducing operational energy consumption is now robust. IKEA's deployment across 37 stores in East Asia connected 7,000 data points to achieve 30% reductions in HVAC energy—savings that compound across their portfolio into millions of dollars and thousands of tons of avoided emissions annually. Foxconn's manufacturing digital twin delivered 30% annual energy reductions through robot training and process simulation, demonstrating that even energy-intensive industrial operations can achieve meaningful improvements.

Predictive Maintenance Extending Asset Lifecycles

GE's digital twin systems for industrial equipment have demonstrated 30–50% reductions in unplanned downtime and 20–40% decreases in maintenance costs. From a sustainability perspective, the lifecycle extension matters as much as operational efficiency: every year an asset remains in service without replacement represents avoided embodied carbon from manufacturing new equipment.

Synthetic Data Enabling Climate Risk Modeling

Financial institutions facing TCFD and emerging TNFD disclosure requirements are using synthetic data to stress-test portfolios against climate scenarios without exposing customer data. A bank can generate thousands of synthetic loan portfolios, each subjected to different temperature rise pathways and physical risk scenarios, to understand systemic exposure. This would be impossible with real data given privacy constraints and the impossibility of waiting for actual climate disasters to unfold.

Smart City Infrastructure Optimization

Istanbul's metro system achieved a 37.5% increase in operational efficiency alongside 25% reductions in energy and maintenance costs through digital twin deployment. The success factors included strong data integration across legacy systems, clear KPI definitions at the outset, and phased rollout that allowed teams to learn and adjust.

Sector	Primary KPI	Benchmark Range	Measurement Frequency
Buildings	Energy Use Intensity (kWh/m²)	10–30% reduction	Real-time
Manufacturing	Carbon per Unit Output (kg CO₂e/unit)	15–25% reduction	Hourly
Energy/Utilities	Grid Loss Rate (%)	5–15% improvement	Daily
Supply Chain	Scope 3 Visibility (% suppliers tracked)	40–80% coverage	Monthly
Transport	Fleet Fuel Efficiency (L/100km)	8–20% improvement	Weekly

What's Not Working

The Computational Carbon Paradox

Most organizations deploying digital twins for sustainability have not quantified the emissions from the digital twin infrastructure itself. GPU-powered AI servers generate substantial heat requiring energy-intensive cooling. Cloud-based architectures increase vulnerability and energy consumption compared to optimized on-premise deployments. A 2024 analysis in Nature found that training large AI models can emit as much carbon as five cars over their lifetimes—and digital twins increasingly rely on such models for predictive capabilities.

The irony is stark: organizations may report Scope 2 emissions reductions from HVAC optimization while failing to account for increased Scope 2 emissions from expanded data center usage. Without system boundary clarity, sustainability claims become greenwashing risk.

Data Quality and Integration Failures

Digital twins are only as reliable as their data inputs. Organizations frequently underestimate the effort required to integrate legacy sensors, reconcile data formats, and maintain synchronization. A 2024 industry survey found that 68% of digital twin pilots experienced delays due to data integration challenges, with 23% ultimately failing to achieve production deployment.

For synthetic data, the "garbage in, garbage out" principle is amplified. If the real-world data used to train generative models contains biases or measurement errors, synthetic data will reproduce and potentially amplify those flaws. Climate models trained on synthetic data derived from incomplete historical records may underestimate tail risks.

Over-Engineering for Limited Payoff

Not every application justifies a full digital twin. Organizations with small building portfolios, simple operational profiles, or limited maintenance complexity may find that the computational overhead exceeds the optimization potential. The rule of thumb emerging from industry practice: digital twins deliver positive ROI when applied to assets with annual energy costs exceeding $500,000 or maintenance costs above $1 million.

E-Waste from IoT Sensor Proliferation

The hardware layer enabling digital twins—sensors, gateways, edge compute devices—creates its own waste stream. A single smart building might deploy 1,000+ sensors with 5–7 year replacement cycles, generating electronic waste that rarely factors into sustainability assessments. Extended producer responsibility frameworks have not yet caught up with IoT-specific waste challenges.

Key Players

Established Leaders

Siemens operates one of the most mature digital twin ecosystems through its Xcelerator platform and Teamcenter software. Their partnership with NVIDIA to integrate Omniverse capabilities enables physics-accurate industrial simulation at scale. Siemens Gamesa's wind farm digital twins in Denmark represent a flagship sustainability deployment.

ANSYS provides simulation software used across aerospace, automotive, and energy sectors, with particular strength in structural and thermal analysis. Their digital twin offerings integrate with major PLM platforms and support LCA-embedded workflows.

Dassault Systèmes delivers the 3DEXPERIENCE platform powering virtual twins for product design, manufacturing, and urban planning. Their Virtual Singapore initiative remains a reference implementation for city-scale digital twins.

Microsoft Azure Digital Twins offers cloud-native infrastructure for building and managing digital twin graphs, with strong IoT Hub integration and Azure Sustainability Manager for emissions tracking.

PTC provides ThingWorx and Vuforia platforms connecting IoT data to digital representations, with particular strength in augmented reality overlays for field maintenance.

Emerging Startups

Vibrant Planet (California) models trillions of trees and vegetation parcels for wildfire resilience and land management, using NVIDIA-backed infrastructure. Their platform enables land managers to simulate intervention scenarios before committing resources.

WindBorne Systems deploys miniature weather balloons with AI-powered forecasting to provide 1000x more atmospheric data than existing systems, improving climate model accuracy for downstream digital twin applications.

Gretel Labs leads in privacy-preserving synthetic data generation, enabling organizations to create realistic datasets for AI training without exposing sensitive information. Their differential privacy guarantees are increasingly relevant for ESG data sharing.

MOSTLY AI specializes in synthetic data for financial services and healthcare, with emerging applications in climate risk scenario generation for TCFD compliance.

Tonic.ai focuses on developer-friendly synthetic data APIs, enabling teams to create representative test datasets for climate-focused applications without production data access.

Key Investors and Funders

Breakthrough Energy Ventures (Bill Gates) has invested across the digital infrastructure stack supporting climate applications, with portfolio companies leveraging digital twins for grid optimization and industrial decarbonization.

NVIDIA Inception supports 750+ climate and sustainability startups through its Sustainable Futures program, providing GPU compute credits and technical mentorship for digital twin development.

NIST allocated over $280 million in 2024 for digital twin standards development, recognizing the need for interoperability frameworks to unlock cross-sector sustainability benefits.

European Commission invested €50 million in digital twin programs for healthcare and urban resilience, with increasing emphasis on sustainability metrics in evaluation criteria.

Examples

1. IKEA East Asia: Portfolio-Scale Building Optimization

IKEA's digital twin deployment across 37 retail locations in East Asia connected 7,000 sensor points to a centralized platform for HVAC optimization. The implementation achieved 30% energy reduction in heating and cooling—the largest single energy expenditure for their retail footprint. Success factors included standardized sensor specifications across locations, pre-defined intervention protocols based on twin insights, and executive sponsorship ensuring operations teams acted on recommendations. The computational overhead was managed through edge processing at each location, with only aggregated data sent to cloud analytics.

2. Foxconn: Manufacturing Process Simulation

Foxconn's electronics manufacturing digital twin enables robot training and process simulation that delivered 30% annual energy reductions across production lines. The platform models material flows, equipment states, and energy consumption in real-time, identifying optimization opportunities that would be invisible to human operators. Critically, Foxconn quantified the emissions from their digital twin infrastructure and confirmed net-positive impact—the operational savings exceeded computational costs by a factor of 15x.

3. Istanbul Metro: Transit System Efficiency

Istanbul Metropolitan Municipality deployed digital twins across their metro network, achieving 37.5% operational efficiency gains and 25% reductions in energy and maintenance costs. The implementation required integration across ticketing, signaling, and rolling stock systems—a multi-year effort that tested data governance and change management capabilities. The success demonstrated that legacy infrastructure can be "digitally twinned" without wholesale replacement, extending asset lifecycles while improving performance.

Action Checklist

• Conduct a system boundary analysis defining which emissions sources the digital twin will track versus which sources it will create
• Establish baseline measurements for both operational energy consumption and digital infrastructure energy consumption before deployment
• Evaluate edge-cloud hybrid architectures to minimize data transmission and centralized compute requirements
• Implement data quality validation protocols including sensor calibration schedules and anomaly detection algorithms
• Define explicit KPIs with benchmark ranges before deployment to enable objective success evaluation
• Assess synthetic data generation needs against privacy requirements and determine whether differential privacy guarantees are necessary
• Create an IoT hardware lifecycle plan addressing sensor replacement, refurbishment, and end-of-life recycling
• Integrate LCA calculations into the digital twin architecture for continuous environmental impact tracking
• Develop a scaling decision framework identifying the asset size and complexity thresholds that justify digital twin investment
• Establish governance protocols for acting on digital twin insights to ensure recommendations translate to physical interventions

FAQ

Q: How do I calculate whether a digital twin deployment is net carbon positive?

A: Compare annualized emissions from digital infrastructure (cloud compute, edge devices, data transmission) against avoided emissions from operational improvements and extended asset lifecycles. Include embodied carbon from IoT hardware using standard LCA databases. A net-positive deployment typically shows a 3:1 or greater ratio of avoided to created emissions within three years of operation.

Q: What's the minimum asset scale where digital twins make environmental and economic sense?

A: Industry practice suggests annual energy costs exceeding $500,000 or maintenance costs above $1 million as reasonable thresholds. Below these levels, simpler monitoring solutions often provide 80% of the benefit at 20% of the computational cost. Multi-asset portfolios can achieve scale effects that justify deployment at smaller individual asset sizes.

Q: How should synthetic data factor into climate risk disclosures?

A: Disclose clearly when scenario analysis relies on synthetic data, specifying the generative methodology and validation approaches. Regulators increasingly expect transparency about model inputs. Synthetic data is appropriate for stress testing and sensitivity analysis but should be clearly labeled in governance and compliance documentation.

Q: What data quality metrics matter most for sustainability-focused digital twins?

A: Prioritize sensor accuracy (typically <5% measurement error for energy data), synchronization latency (<15 minutes for operational optimization, real-time for safety-critical applications), and completeness (>95% data availability over trailing 30 days). Establish automated alerts when quality metrics degrade below thresholds.

Q: How do edge-cloud architectures reduce the environmental footprint of digital twins?

A: Edge processing handles time-sensitive computations locally, reducing data transmission by 40–60% and avoiding round-trip latency to centralized servers. Only aggregated or anomalous data flows to cloud platforms for historical analysis and cross-site comparisons. This distributes heat generation across facilities with existing HVAC rather than concentrating load in data centers requiring dedicated cooling.

Sources

• Fortune Business Insights. "Digital Twin Market Size, Share & Growth Report 2025-2032." January 2025. Provides market sizing and segment analysis for digital twin deployments across industries.

• World Economic Forum. "Pairing AI and Digital Twin Technology to Cut Emissions." March 2024. Documents case studies including building energy optimization achieving 10–30% reductions.

• NVIDIA Blog. "Sustainable Manufacturing and Design: How Digital Twins Are Driving Efficiency." 2024. Details manufacturing sector implementations including the 1% carbon footprint reduction = 90M tons CO₂ equivalence for global manufacturing.

• MDPI Sustainability Journal. "Analysis of Digital Twin Applications in Energy Efficiency: A Systematic Review." Volume 17, Issue 8, 2025. Academic review of energy efficiency outcomes across 47 digital twin implementations.

• Frontiers in Artificial Intelligence. "Opportunities for Synthetic Data in Nature and Climate Finance." 2023. Examines synthetic data applications for ESG risk modeling and TCFD compliance.

• Mordor Intelligence. "Synthetic Data Market Size, Share, Trends & Research Report 2030." 2025. Market analysis projecting synthetic data growth from $510 million (2025) to $2.67 billion (2030).

• Nature Scientific Reports. "A Synergistic Approach Using Digital Twins and Statistical Machine Learning for Intelligent Residential Energy Modelling." January 2025. Presents methodology for residential building optimization achieving validated energy reductions.

• GM Insights. "Digital Twin Market Size & Share, Growth Analysis 2025-2034." 2025. Projects market growth to $428 billion by 2034 with 41.4% CAGR.