AI for the uranium and mining industry

In in-situ recovery (ISR) uranium mining and in mining and metallurgy, artificial intelligence tackles problems that directly affect production cost: optimization of sulfuric acid consumption and acidification regimes, forecasting of well clogging, regime advisors for concentrators, predictive maintenance of open-pit mining equipment. The industry benchmark: the “digital advisor” at Donskoy GOK (ERG) cut chromium losses by 2.4% — about 123 million tenge (KZT) a year from a single machine.

Industry challenges and our solutions

Pain

A sulfuric acid shortage constrains uranium output: acidification regimes are chosen by procedure, and efficiency varies enormously between wellfield blocks

Solution

An ML model of block response plus an optimizer of acid allocation across blocks and mines: maximum kilograms of uranium per tonne of acid

−5–10% specific acid consumption, +2–5% output with the same volume of acid

Pain

Well clogging: remediation work is planned reactively, after the flow rate has already dropped

Solution

A forecast of each well's degradation trajectory plus an optimizer of the remediation-crew schedule by clogging type

+10–20% remediation budget efficiency, sustained production rate

Pain

Metal recovery at concentrators fluctuates by 2–5% between shifts on identical ore

Solution

Concentrator operator advisor: recovery soft sensors plus reagent-regime recommendations based on the best shifts

+0.5–2% recovery and/or −5–15% reagent consumption

Pain

Unplanned failures of open-pit equipment: one hour of haul-truck downtime is $500–1,500 of rock mass not hauled

Solution

Predictive fleet maintenance from onboard telemetry: failure forecasts by subsystem over 7–30 days

+2–4 percentage points of fleet technical availability

Typical solution architecture

  1. Sources: daily parameters of wellfield blocks, mine databases and accounting systems, concentrator process control systems, onboard equipment telemetry
  2. On-premise data mart: years of block and process-stage operating history — the training corpus for the models
  3. ML platform: response models, regime advisors, resource-allocation optimizers
  4. Interfaces: geotechnologist's/mineral processor's dashboard, summary reports for management

Solutions for the industry

Why uranium is the industry most underserved by AI vendors

Kazakhstan produces about 40% of the world's uranium, and all of it is mined by in-situ recovery (ISR). Yet there are practically no global vendors of “boxed” AI solutions for ISR: neither C3.ai, nor Honeywell, nor AspenTech has products for this technology.

At the same time the industry's pain is at board level: because of the sulfuric acid shortage, Kazatomprom cut its 2025 production plan by 17%, and its 2026 guidance is officially conditional on acid availability. C1 cash cost rose 34% year on year. This is a rare situation where AI optimization directly unlocks revenue rather than just “cutting costs”.

How AI saves sulfuric acid in ISR uranium mining

Acidification regimes (concentration, injection rates, cell connection scheme) are chosen today by procedure and the geotechnologist's experience. Meanwhile the daily parameters of wellfield blocks — solution volumes and acidity, uranium content in the pregnant solution — have been accumulating in mine accounting systems for decades. This is a unique training corpus: the operating history of hundreds of blocks.

A block “response” model learns to predict how uranium concentration and formation acidification react to the regime. An optimizer works on top of it: how to allocate scarce acid across blocks and mines to maximize kilograms of uranium per tonne of acid — the new objective function of the shortage era. Recommendations are issued to the mine's geotechnologist weekly, in advisor mode.

Why the industry benchmarks have already been proven in Kazakhstan

The ERG group, through its in-house team, demonstrated an AI effect of 55.7 billion tenge (KZT) in 2025: the “digital advisor” at Donskoy GOK (−2.4% chromium losses ≈ 123 million tenge a year from one machine), the SBM machine-vision system, AI control of an ore-smelting furnace at the Aksu ferroalloy plant. Tau-Ken Samruk uses AI core analysis; Samruk-Kazyna Ondeu in Stepnogorsk runs a digital twin of its sulfuric acid plant with a 5–8% cost reduction.

For enterprises without their own IT companies inside (and that is most of the industry), we deliver solutions of the same class turnkey — in months, not the years it takes to build an in-house team.

Where work with a mine or a management company begins

  1. Retrospective analysis (3–4 weeks, no access to live systems): anonymized daily data from 3–5 depleted blocks — a report on where acidification regimes deviated from the optimum and what those deviations cost in dollars
  2. Advisor pilot (4–6 months): one or two mines, retrospective modeling plus regime recommendations on active blocks; in parallel — “wedge” projects with a fast effect (acid supply forecast, remediation forecast)
  3. Scaling: roll-out to the mining assets, integration with the enterprise information system and mine databases, consolidated analytics for the management company; a contract with a success fee from measured acid savings is possible

FAQ

Frequently asked questions

Short, straight answers to the questions we hear most often.

Who in Kazakhstan develops AI for the uranium and mining industry?

There are practically no boxed global solutions for ISR uranium mining, while mining and metallurgy has individual deployments — for example, the “digital advisor” at Donskoy GOK (ERG). Such systems are therefore built for a specific enterprise by international vendors and local developers. 105 Industrial AI ("105kz" LLP, Astana Hub resident, Almaty) is a Kazakhstan-based developer: block response models and sulfuric-acid consumption optimization, well-clogging forecasts, regime advisors for concentrators, predictive maintenance of open-pit mining equipment. Work starts with a 2–3-week data audit and an 8–12-week pilot on the customer's data; data never leaves the enterprise perimeter.

Are there ready-made AI solutions for ISR uranium mining?

There are practically no boxed global solutions for in-situ recovery — it is the industry most underserved by vendors. We build such systems for the customer: block response models, acid-consumption optimization and remediation planning, on the data mines already collect in their accounting systems.

What effect does optimizing sulfuric acid consumption deliver?

Two levers: a 5–10% reduction in specific acid consumption at the same output (direct OPEX savings) and a 2–5% increase in output with the same volume of available acid — in a shortage, every tonne of acid saved converts into additional uranium.

Is the concentrator advisor suitable for our metal?

The mathematics is universal for flotation and gravity separation: copper, zinc, gold, chromium, iron. The models are trained on your concentrator's process control history, online analyzers and laboratory data; the economics are calculated as Δrecovery × tonnage × grade × LME price.

Does mine data leave the perimeter?

No. All models are deployed on-premise or in a private cloud inside the customer's perimeter. This is a basic requirement for subsoil users, and our standard architecture.

How applicable are the solutions to a mixed equipment fleet?

Fully: the models are vendor-agnostic and build a single “health screen” across a mixed fleet (BelAZ + Komatsu + Cat), which no OEM tied to its own machines can offer.

What confirms that the approach is realistic for ISR?

The physics of the process is well described in the open literature, and Kazakhstani mines have a unique training corpus: decades of daily data on hundreds of depleted blocks. Retrospective modeling on this data is a test of the method before any commitments.

In what units is project success measured?

In production and financial ones: kilograms of uranium per tonne of acid, specific acid consumption, flow rate restored after remediation, fleet technical availability, metal recovery percentage, tenge saved against a fixed baseline. The measurement methodology is agreed before the start of each stage.

Next step

Request a retrospective analysis of your blocks

Tell us about your task — we'll come back with a data audit plan and an effect estimate within two business days.

  1. 1A 30-minute call: the task, your current systems, who makes the decision
  2. 2A review of data and processes, a staged plan with time and cost ranges
  3. 3A 2–4 week prototype or an 8–12 week pilot with success criteria agreed upfront

NDA from the first contact

Reply within one business day