How chemical, mining and steel companies deploy AI
Chemical, mining and steel companies run AI on 2 tracks: machine learning inside plant and mine controls, and language models and agents in sales and finance. The plant track reports gains in physical units, such as a 3% mill throughput gain from AI blast design at BHP Group's Escondida mine, while Freeport-McMoRan's AI agents process about $500 million of sales invoices. At Linde, the chief human resources officer oversees AI rather than IT, and the board's Human Capital Committee oversees its AI policy.
Verified See methodology
How chemical, mining and steel companies have deployed AI
- Experimenting
Air Products
Industrial gases
Air Products has named no internal AI platform; C3 AI listed new and expanded agreements with Air Products in its May 28, 2025 and Dec 3, 2025 results releases, without saying which applications they cover.
None
Return credited to AI since May 2025
- Experimenting
BHP Group
Mining (copper, iron ore, steelmaking coal, potash)
BHP Group builds its own operational AI with its assets on Microsoft Azure Machine Learning and AWS, opened an Industry AI Hub in Singapore in May 2025, and is building a Central Data Platform for exploration data (Mar 2026).
~US$500M
EBITDA run rate from technology
- Experimenting
Dow
Chemicals and materials science
Dow buys its AI tools and runs them on its own Integrated Data Hub, built on Microsoft Azure and Databricks, which Chief Data & Analytics Officer Chris Bruman called the biggest change to Dow's AI strategy (Apr 2025).
50+
Steam cracker furnace operations on C3 AI
- Experimenting
Freeport-McMoRan
Copper, gold and molybdenum mining
Freeport-McMoRan builds its own process models, starting with TROI, a concentrator model first trained at the Bagdad mine in 2018 with McKinsey, and adds patented predictive leach models and AI agents in about 10 areas (Sep 2026).
~10
Business areas with AI agents
- Experimenting
Linde
Industrial gases and engineering
Linde builds most of its AI in-house through a global AI team and regional AI & Digital teams, tracks every AI project on the internal platform it uses for productivity projects (Oct 2025), and sells its own AI plant control software, AOPS AIPC, to other plant operators.
~300+
AI use cases
- Experimenting
Nucor
Steel and steel products manufacturing
Nucor runs C3 AI's demand forecasting, inventory optimization and production schedule optimization applications on the C3 Agentic AI Platform across multiple facilities, according to C3 AI (Sep 2025); Nucor Business Technology also builds its own software, and Jeffery Roberts's Business Technology Integration team set up an AI Center of Excellence for AI adoption and education (AWMI, Sep 2026).
3+
C3 AI applications live at multiple Nucor sites
- Experimenting
Rio Tinto
Metals and mining
Rio Tinto builds AI applications on Palantir Foundry, whose enterprise agreement Rio Tinto extended in Nov 2024 to include Palantir's AI Platform (AIP), and on AWS services such as Amazon Bedrock and SageMaker; ReconAI, Rio Tinto's rail fault classifier, reads data collected in Foundry and shows its results there (Jun 2025).
96%
ReconAI accuracy classifying locomotive faults
- Experimenting
Sherwin-Williams
Paints and coatings manufacturing and retail
Sherwin-Williams embeds Ask Henry, an AI assistant built on GPT-5 and its own knowledge base, in its Insight and sales platforms (Sep 2026), and a Jul 2026 job posting names Microsoft Copilot, Copilot Studio and Azure AI Foundry among its enterprise AI platforms.
~3,500
Field reps with AI tools on sales calls
Findings
What the materials profiles show
The sector's largest gains credit AI together with other technology and give AI no share: BHP Group's about US$150 million of FY2026 EBITDA and almost US$500 million exit run rate from technology initiatives (Aug and Oct 2026) also cover autonomous haulage, leaching and rail technology, Linde credited digital and AI solutions with about 31% to 32% of its productivity efforts (May 2025), Dow's roughly $450 million of Operating EBITDA improvement from Transform to Outperform in the first 9 months of 2026 works through AI and automation alongside role cuts, sourcing and pricing, and Freeport-McMoRan's 214 million pounds of extra leach copper in 2025 is credited to new applications, technologies and data analytics together with additives and operating tactics.
Executives at several materials companies describe AI savings without a figure or put returns ahead of scale: Freeport-McMoRan's Bert Odinet said on Sep 29, 2026 that AI agents processing about $500 million of sales invoices had delivered significant cost savings, Dow's Jim Fitterling cited big AI savings in legal patent and discovery work (Jan 2026), Sherwin-Williams CFO Ben Meisenzahl said in Sep 2026 that the company wants to see returns before it scales AI further, and Air Products' Eduardo Menezes said on Nov 6, 2025 that the staffing goal comes before further productivity from initiatives like AI.
Mine and plant AI results are reported as operating measures, pilots or vendor claims rather than company returns: BHP Group's Johan van Jaarsveld put oversize detection at Western Australia Iron Ore at about $50 million a year (Jan 2026), but BHP Group called the system a pilot in Mar 2026 and the figure a target at scale in May 2026; Escondida's 3% mill throughput gain from AI blast design (Jan 2026) is a throughput measure; Rio Tinto reports 96% classification accuracy for ReconAI (Jun 2025), and the more than doubled scheduler productivity at Rio Tinto is a BCG X claim; Linde's 2% to 3% product yield gain from AI plant control (Feb 2026) comes from a product pitch.
No materials company has reported a company-wide count of employees using AI tools since May 2025: Dow's nearly 60% Copilot figure comes from a Microsoft blog in Mar 2025, Sherwin-Williams puts AI tools with about 3,500 field reps out of 64,249 employees (Sep 2026), Air Products said in Jul 2025 that it is investing to bring AI tools to the majority of employees, and BHP Group, Linde and Freeport-McMoRan count AI work instead of users, with about 200 point solutions at BHP Group (Jan 2026), about 300 use cases at Linde (Oct 2025) and agents in about 10 areas at Freeport-McMoRan (Sep 2026). Nucor, Rio Tinto and Air Products name no executive who owns AI, and Sherwin-Williams has not named the holder of the SVP, Global AI Transformation role in its Jul 2026 job posting.
BHP Group, Rio Tinto, Freeport-McMoRan and Linde build their own operational AI, while Nucor, Dow and Air Products buy applications from C3 AI: BHP Group runs in-house models on Microsoft Azure Machine Learning and AWS and opened an Industry AI Hub in Singapore in May 2025, Rio Tinto builds on Palantir Foundry and Amazon Bedrock, Freeport-McMoRan trains its own TROI concentrator and leach models, and Linde builds most of its AI in-house and sells its AOPS AIPC plant control software to other operators. Nucor runs C3 AI's demand forecasting, inventory and scheduling applications across multiple facilities (C3 AI, Sep 2025), Dow buys its AI tools, C3 AI Reliability among them, and runs them on its own Integrated Data Hub on Azure and Databricks, and C3 AI listed agreements with Air Products in May and Dec 2025.
Stages
Where chemical, mining and steel companies stand
Experimenting
Air Products, BHP Group, Dow, Freeport-McMoRan, Linde, Nucor, Rio Tinto, Sherwin-Williams
Experimenting means the company has reported neither a business return it credits to AI nor a company-wide AI rollout with a user count. Pilots, partnerships, plans and spending alone keep a company at Experimenting.
Compare
Materials side by side
| Company | Ownership | Platform | Models | Hosting | Reach | Stage |
|---|---|---|---|---|---|---|
| Air Products | No named AI owner; EVP and CIO Brian Galovich leads digital strategy | Bought: No platform named; C3 AI agreements (May and Dec 2025) | Not disclosed | Not disclosed; Air Products said in 2021 it had moved most infrastructure to the cloud, without naming providers | Not disclosed; AI tools planned for most employees (Jul 2025) | Experimenting |
| BHP Group | Chief Technical Officer Johan van Jaarsveld, on the Executive Leadership Team | Hybrid: AI built in-house with assets on Azure and AWS | In-house machine learning and computer vision; no LLM provider named | Microsoft Azure (Azure Machine Learning at Escondida) and AWS for analytics and ML; detection models wired into site process control | No employee usage count published | Experimenting |
| Dow | Chief Information & Digital Officer Debra Bauler, under CFO Jeff Tate | Bought: Bought tools on an in-house data hub (Azure, Databricks) | Microsoft Copilot, OpenAI-assisted chat, C3 AI Reliability | Integrated Data Hub on Microsoft Azure and Databricks; Copilot in Microsoft's cloud; Kyndryl runs Dow's IT infrastructure; where the OpenAI chat and C3 AI run is not disclosed | Nearly 60% on Copilot (Microsoft, Mar 2025) | Experimenting |
| Freeport-McMoRan | Bert Odinet, SVP and Chief Innovation Officer; the CIO and CISO report to him | Hybrid: Process models built in-house; agent vendors not disclosed; autonomy bought | Not disclosed for agents; in-house machine learning process models | Microsoft Azure hybrid cloud (Accenture contract, 2015) and Nutanix infrastructure at sites; where AI models run is not disclosed | No user count; agents in about 10 areas (Sep 2026) | Experimenting |
| Linde | SVP and CHRO Desiree Bacher oversees AI and digitalization; Henning Tomforde leads Global AI & Digital | Hybrid: AI built in-house by a global AI team, with outside builders and Azure | Not disclosed; in-house machine learning and reinforcement learning models | Not disclosed by Linde; vendor case studies put a plant model pipeline and a knowledge assistant on Microsoft Azure | Not disclosed; upskilling program open to all 65,177 employees | Experimenting |
| Nucor | No AI executive named; AI sits with Jeffery Roberts's team in Nucor Business Technology, led by Sushma Walker | Hybrid: C3 Agentic AI Platform for planning apps (per C3 AI), plus in-house builds | Not disclosed; C3 Generative AI agreement (Sep 2025), LLM skills in job postings | Not disclosed; Oct 2026 job postings name Azure as the preferred cloud and Oracle Cloud Infrastructure as an option | Not disclosed; C3 AI apps at multiple facilities, AI courses open to teammates | Experimenting |
| Rio Tinto | No named AI executive; IS&T under CIO Daniel Evans, data science in the product groups | Hybrid: Palantir Foundry and AIP plus AWS services; ReconAI and other tools built in-house | Anthropic Claude via Amazon Bedrock, Meta Llama 3.1 8B, Microsoft Copilot | AWS (Bedrock, SageMaker, EC2, S3) and Palantir Foundry on AWS; Microsoft and ServiceNow cloud services; own data centers not disclosed | Not disclosed; finance, rail maintenance and aluminium engineering teams | Experimenting |
| Sherwin-Williams | SVP, Global AI Transformation, with an AI Center of Excellence (Jul 2026 job posting); holder not named by Sherwin-Williams | Hybrid: Ask Henry on GPT-5 with in-house data; Microsoft Copilot and Azure AI Foundry | OpenAI GPT-5 in Ask Henry; Microsoft Copilot; Azure OpenAI per a vendor | Not disclosed; a Jul 2026 job posting points to Microsoft platforms, and vendor Improving says it built on Azure OpenAI | Not disclosed; AI tools with about 3,500 field reps on calls (Sep 2026) | Experimenting |
Questions
AI in materials, answered
- Which chemical, mining and steel companies are furthest along with AI?
- In the Enterprise AI diffusion report, Air Products, BHP Group, Dow, Freeport-McMoRan, Linde, Nucor, Rio Tinto, and Sherwin-Williams are at Experimenting. A stage follows from the returns and rollouts a company has publicly reported in the last 18 months, not from its spending or plans.
- How do mining companies use AI?
- Mining companies use AI mostly in mine and plant operations. BHP Group detects oversize ore at Western Australia Iron Ore, a system BHP Group called a pilot in Mar 2026, designs blasts with AI at Escondida for a 3% mill throughput gain (Jan 2026) and runs about 200 AI point solutions (Jan 2026). Rio Tinto classifies rail faults with ReconAI on Palantir Foundry (Jun 2025) and uses Anthropic Claude through Amazon Bedrock. Freeport-McMoRan runs its own TROI concentrator model, predictive leach models and AI agents in about 10 areas, including sales invoice processing (Sep 2026). BHP Group, Rio Tinto and Freeport-McMoRan are all at Experimenting, with no business return credited to AI alone since May 2025.
- How do chemical companies use AI?
- Chemical, industrial gas and coatings companies use AI in plant operations, maintenance, planning and sales support. Dow uses C3 AI Reliability and Microsoft Copilot and keeps its data in an Integrated Data Hub on Azure and Databricks, Linde reports about 300 AI use cases (Oct 2025) and ValveWatch at more than 40 facilities, Sherwin-Williams gives about 3,500 field reps its Ask Henry assistant built on GPT-5 (Sep 2026), and Air Products says it uses a lot of AI in administration and engineering (Jan 2026). Dow, Linde, Sherwin-Williams and Air Products are all at Experimenting, and none credits a cost, headcount or revenue figure to AI alone.
- What return have chemical, mining and steel companies reported from AI?
- Chemical, mining and steel companies have reported no business return they credit to AI alone since May 2025. BHP Group's about US$150 million of FY2026 EBITDA from technology initiatives (Aug 2026), Linde's 36% of 2025 productivity savings from digital initiatives (Jun 2026) and Dow's roughly $450 million of Transform to Outperform improvement in the first 9 months of 2026 each combine AI with other technology or cost work. Freeport-McMoRan reports significant savings from AI invoice agents without a figure (Sep 2026). Nucor's only figure on its own record, about 80% fewer man-hours for production scheduling at the Memphis SBQ mill, dates from Jul 2024 and has no cost or headcount equivalent.