ArcelorMittal makes Azure its primary cloud as it rebuilds around AI

ArcelorMittal has named Microsoft Azure its primary cloud platform, deepening a partnership the steelmaker is using to rebuild its digital operations around artificial intelligence.

The company announced the expanded collaboration on Monday, casting it as the technical spine of a ‘Cloud First, Data Centric’ strategy that spans a business with steelmaking in 14 countries and a presence in around 60.

The agreement makes Azure the default home for ArcelorMittal’s data and applications, alongside Microsoft Fabric for data integration and Microsoft Purview for governance.

The steelmaker also plans to build on Microsoft Foundry, the platform Microsoft sells for developing and running AI models and agents.

The move arrives as the real cost of enterprise AI comes under fresh scrutiny, with large customers weighing the price of running models at scale against the efficiency they are promised. ArcelorMittal’s argument is that standardising on one cloud reduces its dependence on legacy systems and makes those gains easier to reach.

‘Technology is central to how we compete, a driver of performance, innovation and lasting advantage,’ said Nik Puri, ArcelorMittal’s chief information officer. The framing is deliberate, positioning steel less as a heavy-industry laggard and more as a data business that happens to run blast furnaces.

Samer Abu-Ltaif, Microsoft’s president for Europe, the Middle East and Africa, described the aim as building business ‘around AI from the ground up’ rather than simply bolting AI onto existing processes. It is the kind of sentence that turns up in most cloud announcements now, though the reach of ArcelorMittal’s footprint gives it more weight than usual.

The stated priorities are familiar for a migration of this scale: scaling digital innovation, strengthening cybersecurity and system reliability, and wringing more efficiency out of operations.

Neither company disclosed a value for the deal, the number of sites moving to Azure, or a timeline for the work.

For Microsoft, an industrial anchor customer is useful evidence that its cloud can carry more than software firms and banks.

The company has spent the past year pressing enterprise AI on several fronts, from funding Mistral’s European expansion to merging its scattered assistants into a single Copilot super app.

Cost has been the recurring tension in those efforts. Microsoft has reportedly weighed putting cheaper models, including China’s DeepSeek, inside Copilot to tame its AI bills, a reminder that even the vendor selling the platform is hunting for savings.

ArcelorMittal is not becoming a single-vendor shop. The steelmaker struck a separate collaboration with Amazon Web Services earlier this year on industrial automation and lower-carbon construction, which makes ‘primary’ the operative word rather than ‘only’.

The context is an industry told to decarbonise and cut costs at the same time. European steelmakers face high energy prices and tightening carbon rules, and any tool that trims waste or downtime has an obvious appeal when margins are this thin.

Microsoft, for its part, has spent the past year courting heavy industry as the next frontier for Azure and its Foundry stack, having already saturated financial services and retail.

Landing a company of ArcelorMittal’s size, and being named its primary rather than secondary cloud, is exactly the reference win its sales teams will lean on.

The industrial logic is plain enough. Steel plants throw off enormous volumes of sensor data, and predictive maintenance, energy optimisation, and quality control are exactly the problems machine learning is suited to, provided the underlying data is clean and sits in one place.

That last condition is the hard part, and it is where the Fabric and Purview pieces matter more than the AI headline.

Decades of plant-level systems, each with its own formats and owners, have to be consolidated before any model can do useful work on top of them.

What the announcement does not settle is how quickly any of this reaches the shop floor. Cloud-first strategies are easy to declare and slow to deliver, and the value of ArcelorMittal’s bet will be read in migrated workloads and lower legacy costs long before it shows up in the language of a press release.