Industrial AI has moved from experimental pilots to boardroom strategy. The global market was worth $43.6 billion in 2024, and is growing at 23 percent per year. While most of the real value still lies in proven, production-grade use cases such as machine vision and predictive maintenance, generative AI is driving new interest, and new experiments with agentic AI are already in the works.
In sum – what to know:
Boardroom talk – once limited to ad-hoc pilots, AI is now embedded in manufacturing strategy, governance, and performance frameworks.
Practical value – automated optical inspection is the top industrial AI use case, delivering big ROIs for manufacturers like Renault and Georgia-Pacific.
New foundations – vendors such as Siemens are building domain-specific gen AI models, while early agentic AI systems hint at self-optimising factories.
Note: this article is the third in a short three-part series about the new tech foundations of Industry 4.0: industrial 5G (private wireless), industrial IoT, and industrial AI – linked but distinct technologies that form the connective, sensory, and cognitive layers of modern industry, driving digitalisation in hard-nosed industrial environments. Here is part three, based on figures from IoT Analytics, with some market-sizing about the state of the industrial AI market. Part one (5G) and part two (IoT) are available here and here.
Here is a quick run-through of an excellent (400-page) paper on the state of industrial AI by Germany analyst house IoT Analytics; it goes like this… The global industrial AI market was worth $43.6 billion in 2024, with compound growth (CAGR) pegged at 23 percent per annum through 2030 – when it is expected to be worth $153.9 billion. The new growth is because of the buzz about generative AI. But two things: industrial AI spending only represents 0.1 percent of corporate industrial revenue, but most manufacturing firms now have a CEO-driven AI strategy.
The average US manufacturer made $30.5 million in 2024, estimates IoT Analytics. All of them together spent over $10 billion on industrial AI in 2024. This translates to an average of roughly $40,000 per manufacturer, it says – which is about 0.1 percent of average revenue, three percent of average R&D spending ($1.56 million), and seven percent of average IT spending ($610,000). Larger companies spend more on AI than smaller companies. A significant portion of industrial AI spending is allocated to consulting and system integration services.
The top-earning AI services vendor is Ireland-based Accenture, which announced a $3 billion three-year investment in late 2023, and claimed 2,000-odd generative AI projects in 2024/2025 (fiscal; ending August 31). Other major suppliers – “in a fragmented services market” – are India-based Infosys and UK-based Deloitte. On the flipside, the top industrial AI user is Japanese auto manufacturer Toyota, which invested 1.7 trillion yen ($10.6 billion) during (its fiscal) 2025, including on front-line ML models, OT know-how digitisation, and analytics for safety and productivity.
When IoT Analytics polled execs at big manufacturers in 2021, AI was hardly on the radar, rarely appearing in more than “ad-hoc exploratory projects”. It is different today: most leading manufacturers have dedicated AI strategies, which are “vision-driven, supported by governance frameworks, performance targets, and integration with broader business objectives”, says IoT analytics. “This marks a significant cultural and structural shift, elevating AI from a peripheral technology investment to a top-of-mind discussion point for CEOs during earnings calls,” it notes.
But generative AI, so hyped-up, is way down the list; camera AI cases for quality inspections are way ahead in Industry 4.0. IoT Analytics states: “Of the 48 industrial AI use cases [we have] analyzed… automated optical inspection [is]s the leading one with a share of approximately 11 percent. For comparison, all the gen AI cases combined currently account for less than five percent of the market – with coding being the largest at one percent.” This is likely because the ROI proof (“nine-digits in savings and value gained”) is clearer with other AI cases.
IoT Analytics says: “While the financial community grapples with [the] AI bubble, and some outlets report 95 percent failure rates for enterprise AI pilots, many industrial AI projects have already proven their value through measurable cost savings, uptime improvements, and quality gains. In 2023, IoT Analytics noted that machine vision had the highest ROI and quickest amortization time of all Industry 4.0 technologies at that time, with AI-assisted flaw detection and process/operations optimization as the top rising machine vision applications at the time.”
It references a couple of case studies: car maker Renault SA saved €270 million on its energy and maintenance bills back in 2023 with predictive (AI/IoT) maintenance; pulp and paper company Georgia-Pacific claims to have saved (“annual value capture”) hundreds of millions through its AI projects – likely, mostly IoT analytics, but also via a generative AI document generation tool called ChatGP. And while generative AI represents only a scrap of industrial AI activity, the interest is palpable, and the projects are growing; IoT Analytics has a repository of 530 of them, it says.
These are being used for issue resolution (35 percent of projects), inquiry handling (34 percent), and post-sale support (19 percent). Marketing ( content-creation; 17 percent) and IT (development and coding; 15 percent) are also popular disciplines. It says: “In the manufacturing sector, issue resolution and coding support have become particularly important. Applications like these have helped gen AI… to become a leading industrial AI development.” Generative AI will comprise a quarter of industrial AI projects by 2030, from six percent in 2024.
IoT Analytics says: “Common use cases for gen AI in industry include operations and service support (documentation querying and troubleshooting) and code generation for OT and embedded assets. But it is also increasingly used across the entire manufacturing value chain, including in R&D (product discovery), design (generative design), engineering (gathering requirements), and field service (guided maintenance). At this point, manufacturing rollouts have largely been driven by industrial software vendors in the form of copilots in industrial software.”
It cites copilot integrations from Siemens, Rockwell Automation, and ABB.
There is an issue with industrial foundation models (LLMs), however. “Some manufacturers who have tried to build assistants and copilots with LLMs from the likes of OpenAI, Google, or Anthropic have seen limited understanding in industrial environments. Since many of the valuable industrial data points that are required to train an LLM do not reside on the public internet, some industrial tech vendors have started to build purpose-built industrial foundation models (IFMs) that aim to “speak the language of engineering,” and are trained on domain-specific data.
It cites examples from Siemens (Industrial Foundation Model), Google (Gemini Robotics), Nvidia Isaac GR00T N1), and others.
Beyond generative AI, talk about industrialised agentic AI is early. “While many industrial software vendors began prominently featuring the term agentic AI in their messaging in 2025, deployment is still in its infancy,” says IoT Analytics, with a review of its findings at Hannover Messe 2025, where most showcases demonstrated “only basic orchestration capabilities” – except for Accenture, whose ‘engineering orchestrator’, for modifying engineering designs using natural language, was the one bright spot (the “one promising showcase”.)
It says: “The agentic engineering chatbot serves as a control layer on top of existing tools, interpreting user prompts and executing design changes across multiple tools – like Siemens NX, Siemens Polarion, Altair Hypermesh, and Altair HyperView.” As well, it suggests dynamic AI agents will replace static rules in industrial manufacturing execution system (MES) setups, as offered by the likes Portugal-based MES vendor Critical Manufacturing – to “adapt, learn, and optimize production in real time”. More to come for sure, and fast.
