Inference in industrials: enhancing efficiency through AI adoption

Rapid progress in ‘training’ artificial intelligence models paves the way for AI’s deployment in a spectrum of real-world applications. As trained models become capable of independently making decisions or predictions based on new data in real time – the process known as ‘inference’ – we see huge opportunities for software providers that successfully leverage AI to deliver efficiencies across industries, from construction to logistics.

The growing powers of AI inference

Increasingly sophisticated algorithms, growing volumes of data and rising computational power have driven stunning advances in AI. As explained in our recent blog, innovation has driven down the cost of large language models 10-fold each year since 2022.¹

Advanced AI-based technologies are rapidly becoming less prohibitive for companies to explore as costs tumble. In a virtuous circle, lower costs enable greater adoption which should drive innovation.

Within the AI industry, a shift in focus is underway from training to inference.² According to BofA Global Research forecasts, global revenues from AI model inference will overtake those from model training in 2026.³ Growing inference capabilities open the door to business-critical use cases with strict latency and reliability requirements.

Source: BofA Global Research, August 2024. Forecast data for 2024 to 2027.

Applications within industrial software

Industrial software companies have been among the most eager adopters of AI inference, which is particularly well suited to the tasks of automating workflows, analysing patterns in data and optimising processes.

We observe applications of AI within four distinctive markets for industrial software.

First is product lifecycle management (PLM) software, which underpins the so-called ‘digital thread’ that follows products’ journey from development to manufacturing to service. McKinsey has estimated that up to US$2.1tn in economic value could be added by AI in factory automation.⁴ US-listed PTC, a leading PLM software supplier, uses AI agents to interact with service technicians, help automate tasks and provide predictive maintenance recommendations for its clients’ products, thereby reducing downtime.⁵

Second, and related, is computer-aided design (CAD) and engineering (CAE) software. AI is being used by the likes of US-listed Ansys within software that allows engineers to optimise product performance simulation in various environments. For example, Ansys has demonstrated how forecast quality for modelling the rotational stiffness of a turbine impeller could be improved from 81% to 98% by training a neural network.⁶

Third is within construction, one of the least digitised sectors in the global economy.⁷ Software developed by the likes of US-listed Bentley Systems digitises how buildings are designed, built and managed. Their products employ AI to measure building materials’ carbon footprint and impact at the design phase, and to generate insights around structural conditions for predictive maintenance. By leveraging its digital twins of infrastructure assets and AI-powered inspections, Bentley reports that engineering clients have saved up to 30% in labour costs and up to 15% in repair costs.⁸

Fourth is within logistics. AI is being used in route planning algorithms developed by the likes of Canada-listed Descartes Systems to optimise logistics companies’ route productivity, driving cost and emission reductions. The World Economic Forum estimates that, by leveraging real-time data and predictive analysis, AI software could reduce the industry’s greenhouse gas emissions by up to 7% through route optimisation and more efficient asset management.⁹

According to one estimate, AI adoption is expected to drive a doubling in the size of the global industrial software market by 2030.¹⁰

Source: IoT Analytics, December 2024: Industrial Software Landscape 2024-2030. Forecast data for 2024 to 2030.

Enhancing the customer value proposition

Industrial software companies are in the business of unlocking productivity gains. Through making processes more efficient, they enable companies to accomplish more with fewer resources and so deliver a tangible return on their clients’ investment.

Those that successfully integrate the potential of AI can aim to heighten their value proposition for clients, deepening the long-term relationships at the heart of their business models.

Over decades, industrial software companies have become deeply integrated into companies’ complex workflows. They have collected domain-specific data on which they can train their AI models, providing end users with highly relevant and actionable information.

These proprietary insights, combined with the advantages of incumbency in often mission-critical applications, support a competitive advantage that is very challenging for generic AI models to erode. By embracing AI, we believe leading industrial software companies can continue to successfully partner with industry as the powers of inference become more sophisticated.

¹ Andreessen Horowitz, January 2025. Analysis based on the cheapest LLM with a minimum MMLU score of 42. Measuring Massive Multitask Language Understanding (MMLU) is a benchmark for evaluating the capabilities of large language models (LLMs). The MMLU score represents the percentage of multiple-choice questions answered correctly by the LLM. A minimum MMLU score of 42 means that at least 42% of questions were answered correctly.
² Oracle, 2024: What Is AI Inference?
³ BofA Global Research, August 2024: Data Centers – Asia-Pacific APAC data centers: Set to see strong growth and benefit from AI uplift
⁴ McKinsey, 2023: The economic potential of generative AI: The next productivity frontier
⁵ PTC, 2024: How PTC Uses AI to Create Value for Customers
⁶ Ansys, 2021: Artificial Intelligence and Machine Learning Applied in Computer Aided Engineering
⁷ McKinsey, 2023: From start-up to scale-up: Accelerating growth in construction technology
⁸ Bentley Systems, March 2025: Fixing America’s Bridges: AI and The Digital Revolution in Infrastructure Design and Maintenance
⁹ World Economic Forum, January 2025: Intelligent Transport, Greener Future: AI as a Catalyst to Decarbonize Global Logistics
¹⁰ IoT Analytics, December 2024: Industrial Software Landscape 2024-2030

References to specific securities are for illustrative purposes only and should not be considered as a recommendation to buy or sell. Nothing presented herein is intended to constitute investment advice and no investment decision should be made solely based on this information. Nothing presented should be construed as a recommendation to purchase or sell a particular type of security or follow any investment technique or strategy. Information presented herein reflects Impax Asset Management’s views at a particular time. Such views are subject to change at any point and Impax Asset Management shall not be obligated to provide any notice. Any forward-looking statements or forecasts are based on assumptions and actual results are expected to vary. While Impax Asset Management has used reasonable efforts to obtain information from reliable sources, we make no representations or warranties as to the accuracy, reliability or completeness of third-party information presented herein. No guarantee of investment performance is being provided and no inference to the contrary should be made.