Artificial Intelligence at BHP

BHP is reported to be the world’s largest mining company by market capitalization, according to Wikipedia, citing 2025 data. It has more than 80,000 employees and contractors working across operations in Australia, Chile, Peru, Brazil, Canada, and the United States. The company posted US$51.3 billion in revenue for fiscal year 2025, on record production of 2,017 kilotonnes of copper and 263 million tonnes of iron ore.​

That scale sits against a hard physical constraint, not simply a growth target: BHP projects that the world will need to double the copper produced over the next 30 years, relative to the past 30, to keep pace with decarbonization technology, even as ore grades decline at existing mines and discoveries become rarer — a claim corroborated by EY, which finds that meeting the world’s electrification goals will require 115% more copper mined over the next 30 years than has been mined throughout all of human history. ​

Rising demand against a shrinking margin of accessible ore is not a problem more labor or capital alone can solve; it is why BHP’s AI investment is aimed at extracting more value from mines it already operates rather than simply expanding faster.​

BHP’s response has been to formalize that investment rather than leave it to individual sites; it opened its first Industry AI Hub in Singapore in May 2025, a dedicated center intended to accelerate AI adoption across its mining and resources operations rather than leaving each site to develop its own tools independently.​

This article examines two AI use cases that show how BHP applies that investment inside its own operations:​

We begin by examining how BHP applies AI-driven analytics to address processing efficiency at its flagship copper operation.

Escondida, BHP’s copper operation in Chile’s Atacama Desert, is the world’s largest copper mine by production volume, producing over one million metric tons of copper annually. Like most mature mines, it faces declining ore grades over time, meaning more rock has to be processed to extract the same amount of copper.

Screenshot: Escondida’s resource scale and grade as part of Chile’s significant share of BHP’s copper resources. (Source: mining.com)​

Because Escondida sits in one of the driest places on Earth, every gain in processing efficiency also has to account for water and energy use, not just output — a constraint BHP has described as central to why it is turning to AI, machine learning, and data analytics to “unlock more production and value from our existing mines” rather than relying solely on discoveries.​

BHP partnered with Microsoft in May 2023 to bring AI into Escondida’s concentrator circuit — the stage where crushed and milled ore is floated and separated into copper concentrate and waste. The system runs on Azure Machine Learning, Azure Synapse Analytics, and Azure Data Lake Storage, drawing on real-time plant data from the concentrators to generate hourly predictions.

BHP’s deployment of Microsoft’s AI stack at Escondida has introduced several operational changes across the concentrator and supporting infrastructure: ​

The workflow change is a shift from periodic adjustment to continuous, AI-informed decision-making. A concentrator operator who once relied on scheduled reviews of plant performance can now see and act on an hourly, machine-generated recommendation as ore characteristics and conditions change within a single shift — turning a process historically managed in daily or weekly cycles into one adjusted closer to real time.​

BHP has not published a comprehensive figure for how much the partnership has improved overall copper recovery rates industry-wide; the original 2023 announcement was framed around expected gains rather than delivered ones.

What the company has since quantified is the resource side of the same system, under what it calls the Energy and Fresh Water Sustainability Program: BHP Chief Executive Officer Mike Henry said that AI at Escondida’s processing plants has helped save more than three gigalitres of water and 118 gigawatt hours of energy since fiscal year 2022, a figure independently reported at 3.5 gigalitres of water over the same period. ​

BHP Chief Technical Officer Laura Tyler framed the broader ambition directly: “We expect the next big wave in mining to come from the advanced use of digital technologies.” The company is expanding the technology to a second concentrator at the site, evidence that this has moved past a single pilot circuit — even though the headline metric most people would ask about, copper recovery itself, remains a claimed rather than a disclosed figure.​

Screenshot: BHP–Microsoft infographic illustrating how AI and cloud computing are used to optimize copper recovery at Escondida. (Source: Linkedin:Mining Down Under)

Materials-handling infrastructure — conveyors, crushers, and other fixed processing plants — runs continuously across mine sites that can span tens of kilometers, and problems like spillage, oversized material, or a foreign object on a conveyor belt can cause both safety incidents and unplanned downtime.

Human inspection teams cannot watch every meter of that infrastructure at once, which makes continuous, automated monitoring a natural target for AI investment ahead of a failure or injury, rather than a response to one after the fact — a shift already visible in coal-mining research, where manual sorting is described as unable to meet the demands of intelligent, crewless operations. ​

BHP has deployed computer vision systems across its operations in Chile and Western Australia that run on existing camera infrastructure, rather than requiring new hardware to be installed at every site. BHP’s computer vision rollout has introduced several changes to how material‑handling risks are monitored and managed: ​

For a maintenance or operations team, this replaces scheduled or incident-triggered inspection with continuous, exception-based monitoring. Rather than walking a conveyor line on a fixed schedule, staff are alerted only when the system flags something that needs attention. In some cases the corrective step happens automatically before anyone is dispatched at all — a shift BHP has described in its own operations, where a real-time monitoring system detects objects, alerts controllers, and can automatically stop the conveyor, narrowing the gap between when a hazard. ​

BHP frames the underlying goal in workforce terms as much as in efficiency terms: the system helps “keep material moving safely and consistently, while reducing the need for teams to work in higher-risk situations” — fewer people physically inspecting live conveyor lines and crusher circuits during routine monitoring.​

BHP describes this as one of several computer vision applications now running across its portfolio, alongside similar systems used for environmental monitoring and digital quality control, which suggests a platform being extended across use cases rather than a single-site pilot. Unlike the Escondida processing work, this use case comes with a disclosed before-and-after outcome. ​

Screenshot: AI‑driven conveyor monitoring system, showing automated belt alignment, rip detection, and computer‑vision‑based hazard identification. (Source: Discovery Alert)

At BHP’s Western Australia Iron Ore operations, where the same computer vision approach — cameras and machine learning integrated directly into the process control system — monitors conveyor points for oversized rocks and foreign objects, S&P Global reported that the system unlocked roughly 1 million tonnes of additional annual iron ore production, worth an estimated $50 million, after cutting crusher downtime by 20% and related disruptions by up to 60% since its 2025 deployment, with no subsequent incidents from oversized or foreign objects. ​

BHP has separately described the disruption events the system targets as having historically contributed to more than 1,000 hours of downtime across the system — the baseline against which the improvement is measured. S&P Global’s reporting also credits the result as much to how the system was integrated directly into BHP’s existing process-control systems as to the underlying technology itself, which is part of why it scaled into an operational capability rather than staying an isolated pilot. ​

This analysis examines the following lessons enterprise leaders can draw from BHP’s AI adoption:​