Reviewed August 2026 against Rio Tinto’s 2025 full-year and H1 2026 results releases, its Q4 2025 production report, and Hitachi Review’s AutoHaul engineering paper.

Jump to the calculator →

Rio Tinto Supply Chain Digitalization: The Real Numbers

By Farmonaut Research & Content Team · Published 19 September 2025 · Updated 28 September 2026

Rio Tinto’s supply chain digitalization is not a strategy deck. It is a 1,700 km heavy-haul railway in Western Australia on which 221 locomotives run with no driver on board, a processing plant with a full digital twin, and — under an agreement dated 15 January 2026 — a copper mine in Arizona whose leach chemistry is being modelled on Amazon Web Services infrastructure. In the twelve months to 31 December 2025 that machine shipped 326.2 million tonnes of Pilbara iron ore and produced 883 kt of mined copper, per the company’s Q4 2025 production results released 21 January 2026.

This page collects the figures Rio Tinto has actually disclosed — with the date each was disclosed and the filing it came from — rather than the adjectives that usually stand in for them. Where a number is a target rather than an outcome, it says whose target and for what period. Every source is linked so you can pull a fresher figure than the one printed here.

On this page

1. The pit-to-port chain: what is actually digitalized
2. Cloud transformation strategy: where the compute sits
3. Digital workflow automation and the Safe Production System
4. Enhancing customer experience with AI and the START label
5. The Rio Tinto digital twin, precisely defined
6. Initiative-by-initiative table with verification sources
7. Benchmark your site against Gudai-Darri (calculator)
8. Metals and mining digital transformation: a five-step verification method
9. The satellite layer: where Farmonaut fits
10. Frequently asked questions
11. What would change this picture

---

The pit-to-port chain: what is actually digitalized

Rio Tinto’s supply chain digitalization began with the Mine of the Future programme announced in 2008, whose central bet was that a mine 1,500 km away could be run from an operations centre in Perth. That centre exists and, per Rio Tinto’s own description of the Gudai-Darri opening on 21 June 2022, sits roughly 1,500 kilometres from the ore body it supervises.

The rail leg is the part with the hardest engineering numbers attached, because Hitachi published them. According to Hitachi Review Vol. 69 No. 6 (2020), the AutoHaul system covers a 1,700 km network, equips 221 heavy-haul locomotives, and runs trains of 240 wagons at 106 tonnes each — about 28,000 tonnes and 2.5 km of train. The Office of the National Rail Safety Regulator granted accreditation on 16 May 2018; the first autonomous loaded run followed on 10 July 2018; full deployment completed in December 2018. By the time of that paper, 98% of the AutoHaul-equipped fleet was operating autonomously and the system had logged over 11 million autonomous kilometres across 42 protected level crossings, built on ETCS Level 2 signalling with automatic train operation.
Beyond the railway, the company’s wider programme, from autonomous haulage to back-office systems, is laid out in Rio Tinto’s broader AI push.

Those are the mechanics. The output is the number that matters commercially: 327.3 Mt produced and 326.2 Mt shipped from the Pilbara in calendar 2025. Rio Tinto’s guidance for calendar 2026, issued with the same 21 January 2026 release, is 323–338 Mt, with the company flagging the lower end. That framing is worth holding onto — a decade of automation has bought consistency and cost, not a step change in tonnage.

Cloud transformation strategy: where the compute sits

The clearest public statement of Rio Tinto’s cloud posture is not an IT press release — it is a copper deal. On 15 January 2026 the company announced a two-year collaboration with Amazon Web Services in which AWS became the first customer for Nuton bioleach copper from the Johnson Camp mine in Arizona, and in the same agreement supplies cloud-based data and analytics capacity used to simulate heap-leach performance and feed Nuton’s decision systems.

That structure is the interesting part for anyone benchmarking a cloud transformation strategy. Compute is not being bought as overhead; it is being traded against offtake. The disclosed production target is approximately 30,000 tonnes of refined copper over four years — about 16,000 tonnes from the run-of-mine leach pad and about 14,000 tonnes attributed to the Nuton technology. Rio Tinto also published intensity figures for that copper: 2.82 kgCO₂e per kg of copper on a full-scope basis including Scopes 1, 2 and 3, and 71 litres of water per kilogram, with 100% of site electricity matched by renewable energy certificates. AWS’s stated use is physical: cables and busbars, transformer and motor windings, printed circuit boards, and processor heat sinks in US data centres.

Two things follow. First, the modelling workload — heap-leach recovery prediction, acid and water optimisation — is genuinely cloud-shaped: irregular, simulation-heavy, and worthless without long historical series. Second, the copper is going into the AI build-out that is driving the compute demand in the first place. If you are writing a business case for cloud in metals, this is the closest thing to a public reference architecture with numbers attached to both sides of the trade.

Digital workflow automation and the Safe Production System

Rio Tinto’s digital workflow automation programme has a name and a start date. The Safe Production System (SPS) began rolling out in July 2021 and, per Rio Tinto’s own account of the programme, reached 80% of sites, built on four elements: great people, strong stable assets, customer focus, and effective simple processes. The company’s Chief Technical Officer Mark Davies described the intent as “transitioning from being an organisation of ‘fixers’ to one of ‘problem solvers’.”

The site-level evidence is specific. Amrun, the bauxite operation at Weipa in Queensland, reached its nameplate capacity of 22.8 million tonnes a year for the first time in 2024, a 12% year-on-year production uplift, with plant flow improvements of 6% and shutdown losses eliminated from March 2024. Employee satisfaction scores across deployed sites rose by an average of two points.

The programme was extended on 1 July 2026, when Rio Tinto deployed a Management Operating System (MOS) focused on safety standards and integrated performance management, disclosed in its H1 2026 results release of 29 July 2026. That release also gives the running total on the productivity programme: $870 million of productivity benefits banked in the first half of 2026, against a stated target of $1.8 billion by year-end. The comparable figure a year earlier, from the full-year 2025 results of 19 February 2026, was $650 million of annualised productivity benefits identified by Q1 2026, alongside a 5% reduction in operating unit costs in 2024 real terms.

For context on how much room a productivity programme has, set $1.8 billion against the $25.4 billion of underlying EBITDA Rio Tinto reported for 2025 — roughly 7%. Against $12.3 billion of 2025 capital expenditure, it is about 15%. Neither ratio is transformational on its own, which is precisely why the company frames it as a run-rate rather than a windfall.

Enhancing customer experience with AI and the START label

The customer-facing side of Rio Tinto’s digitalization is narrower than the marketing language suggests, and clearer for it. On 3 February 2021 the company launched START, described as a “nutrition label” for responsible aluminium: a blockchain-backed digital label carrying ten criteria — carbon footprint, water use, recycled content, energy sources, community investment, safety performance, diversity in leadership, business integrity, regulatory compliance, and transparency — issued per production site and made available for aluminium from Rio Tinto’s managed operations globally. The same release states Rio Tinto’s aluminium operations run at a greenhouse gas emissions intensity 60% below the industry average, and notes RenewAl, launched in 2016, as the company’s earlier certified low-CO₂ primary aluminium brand.

What START does for a buyer is replace an annual PDF audit with a per-shipment data object. For a US automotive or packaging purchaser that has to substantiate embodied-carbon claims to a customer of its own, that is the difference between an assertion and an artefact. The analytics and machine-learning layer sits behind it — reconciling smelter-level emissions and energy data into a label — rather than in a chatbot facing the customer.

If you are evaluating a metals supplier’s digital services, the practical test is whether the sustainability data arrives attached to the metal or attached to the corporate report. A buyer can verify Rio Tinto’s side by requesting a START label for a specific lot and checking that the site named on it matches the shipping documents.

The Rio Tinto digital twin, precisely defined

Searches for “Rio Tinto digital twin” usually return generic explainers. The concrete answer: the digital twin Rio Tinto has publicly described is a full digital replica of the processing plant at Gudai-Darri in the Pilbara, used for monitoring and for virtual-reality training. It is not a twin of the whole mine, and it is not a twin of the rail network.

The surrounding automation at that site, from the 21 June 2022 opening release: 23 Caterpillar 793F autonomous haul trucks, 3 Caterpillar MD6310 autonomous drills, the world’s first fully autonomous water carts, a 166 km rail spur, a 34 MW solar farm of roughly 83,000 panels expected to supply about one-third of average electricity demand, and around 600 permanent roles against 43 million tonnes a year of nameplate capacity and a mine life of over 40 years. Construction consumed more than 14 million workhours and A$3.2 billion of goods and services sourced in Western Australia.

Those ratios are the useful export. Gudai-Darri runs at roughly 1.87 Mt per autonomous truck per year and roughly 14 permanent roles per Mt of annual capacity. Any site claiming a comparable level of digital maturity should be able to state its own two numbers. The calculator below does that arithmetic against these disclosed benchmarks.

Initiative-by-initiative table with verification sources

Benchmark your site against Gudai-Darri

Enter your own operation’s four numbers and the tool compares them with the ratios Rio Tinto disclosed for Gudai-Darri on 21 June 2022.

Assumptions: benchmarks are Gudai-Darri's 43 Mt/a nameplate capacity, 23 autonomous haul trucks and ~600 permanent roles as disclosed on 21 June 2022; nameplate is treated as throughput. It excludes haul distance, truck payload class, ore hardness, strip ratio, contractor headcount, and the rail and port legs — all of which move these ratios materially. Treat it as a conversation-starter with your own operations team, not a valuation.

Metals and mining digital transformation: a five-step verification method

Most digital transformation initiatives in the mining and metals industry are announced, then never quantified again. This checklist survives whichever numbers are in fashion, and works on any operator's claim, not just Rio Tinto's.

1. Find the denominator. A "30% efficiency gain" means nothing without the base. Rio Tinto's $1.8 billion FY2026 productivity target reads differently once you set it beside $25.4 billion of FY2025 underlying EBITDA.
2. Separate outcome from target. The $870 million banked in H1 2026 is an outcome. The $1.8 billion is guidance. Company releases distinguish the two; secondary coverage frequently does not.
3. Check the scope word. "Digital twin" at Gudai-Darri means one processing plant. Vendors routinely stretch the same phrase to cover an entire operation.
4. Pull the regulator's file. For a US site, MSHA's Mine Data Retrieval System holds mine-by-mine records filed on an ongoing basis; look up the mine ID and read employment, hours and inspection history directly rather than accepting a company's headcount claim. For commodity-level context on US production and import reliance, the USGS Mineral Commodity Summaries is republished annually in the first quarter.
5. Wait one reporting cycle. Any automation claim worth believing shows up in the next quarterly production report as tonnes, unit cost, or both. If it does not appear within two cycles, treat it as a pilot.

Applied to the biggest names, this method is unforgiving and useful. It is also the reason the useful comparison in metals industry digital services is not vendor-versus-vendor but disclosure-versus-disclosure: an operator that publishes site-level ratios is auditable, and one that publishes only percentages is not. For a longer view of how these programmes sit inside the wider sector, see our coverage of Rio Tinto and IOC mining trends.

The satellite layer: where Farmonaut fits

Everything above depends on data an operator collects inside its own fence line. The independent layer — what a site looks like from orbit — is where Farmonaut works, and it is deliberately the part a mining company cannot mark its own homework on.

• Satellite monitoring of disturbance and stockpiles: repeat imagery over a lease shows waste dump growth, pit advance and land-use change without site access, which is what makes it usable by a lender, an offtaker or a regulator.
• Emissions and footprint tracking: our carbon footprinting tools attach a measured spatial record to disclosures of the kind START publishes per site.
• Chain-of-custody: blockchain product traceability for material moving from pit to processor to buyer.
• Fleet and equipment visibility: fleet management for mining vehicles, the operational cousin of the truck ratios in the calculator above.
• Advisory and optimisation: Jeevn AI delivers site-specific analysis; the same logic we apply to production optimization strategies applies to hard-rock operations.
• Integration: the Farmonaut satellite API pushes geospatial and environmental data straight into your own systems — see the developer documentation for endpoints and payloads.

Farmonaut runs on a subscription model that scales from a single lease to enterprise and government portfolios, so the entry cost is a monthly line item rather than a capital programme.

Frequently asked questions

What would change this picture

Three things would move these numbers, and each has a published checkpoint. First, the productivity programme: if the FY2026 result lands short of $1.8 billion, the run-rate story weakens, and the next half-year release settles it. Second, fleet electrification — Rio Tinto disclosed battery-electric haul truck trials at multiple operations in its H1 2026 results, and a commercial order would change the truck-ratio arithmetic in the calculator above. Third, the AWS agreement runs two years from January 2026; whether it is renewed, expanded beyond Nuton, or quietly lapses is the single clearest signal of how central cloud is to the company's operating model rather than its marketing.

The durable point is the method, not the figures. Take any operator's digital claim, find the denominator, split outcome from target, check the scope word, pull the regulator's file, and wait one reporting cycle. Applied consistently, it separates the mines that are genuinely digitalized from the ones that have merely said so — and it will still work when every number on this page has been superseded.

To layer independent, satellite-verified data over your own operations, explore Farmonaut's fleet management, carbon footprinting and traceability tools.