Barrick Ties AI Partner Avathon’s Fees to Mine Performance

Barrick Mining has selected Avathon’s Autonomy Platform as the AI-native operating system for its entire North American business, a deployment already running at the Nevada Gold Mines complex and built around a five-year agreement that ties Avathon’s fees to actual operational outcomes.

This is not a pilot. Barrick is presenting the arrangement as a foundational change to how it plans, operates, and manages assets across the full mining value chain, from exploration and mine planning through processing, maintenance, and supply-chain management.

For anyone tracking how major gold producers are responding to cost pressure, declining ore grades, and mounting ESG obligations, the structural choice Barrick has made here carries read-through for the whole sector.

Here is what Barrick is actually deploying, why it is committing now, and what the deal’s structure and industry precedents reveal about where large-scale mining is heading.

A five-year deal that pays Avathon only when Barrick’s operations improve

The most unusual feature of this agreement is not the technology. It is how Avathon gets paid.

The five-year AI cooperation agreement, announced on 23 September 2026, is structured as a base fee for using the Avathon platform plus additional performance-based fees tied to the operational outcomes actually achieved. No hard dollar values have been publicly disclosed.

That structure tells you something about how Barrick has approached the risk. By linking a portion of Avathon’s compensation to measurable operational improvement, Barrick has built accountability directly into the contract, reducing the odds that this becomes a large technology spend with ambiguous returns. Performance-linked terms at this scale are uncommon, which is precisely why the structure warrants attention.

The scope fills out the picture. The partnership connects data, operational knowledge, and AI intelligence across Barrick’s North American business, with initial applications spanning:

• Safety: computer-vision monitoring to flag hazardous conditions and unsafe behaviour
• Production and recovery: linking ore flow, processing decisions, and operating constraints to lift throughput and recovery
• Asset reliability: predictive monitoring of equipment health and maintenance coordination
• Supply-chain intelligence: connecting asset health, inventories, supplier performance, and demand forecasts
• Mine planning: AI-enhanced planning and scheduling
• Exploration and growth: machine learning applied to geological and production data

The two companies began working together in December 2025. Avathon’s software has been operational since Q2 2026 at several of Barrick’s U.S. sites, including Nevada Gold Mines, and full deployment across the Nevada complex is expected within one to two quarters of the announcement. The next planned rollout is at Barrick’s operations in the Dominican Republic, signalling the partnership is already being scoped beyond North America.

“Barrick chief executive Mark Hill expressed intent to link the entire operational value chain through the platform,” according to reporting by James Stent in Mining Magazine.

What Avathon’s platform actually does inside a mine

Enterprise AI announcements tend to arrive wrapped in abstraction. The more useful question is which specific decisions are being handed to the software, and what physical systems it is watching.

What the platform monitors

Avathon’s Autonomy Platform is built on approximately 200 AI models that map operational processes, drawing on cameras and sensors to read real-time site conditions. It functions as a common intelligence layer across systems that have traditionally operated in isolation.

Inside a mine, that monitoring spans several concrete domains:

• Safety: computer-vision systems that keep workers separated from heavy haul equipment
• Processing: real-time tuning of processing parameters using ore-quality data
• Asset reliability: detecting crusher and conveyor anomalies before circuits halt
• Supply chain: tracking inventories, supplier performance, and materials readiness

The Draslovka partnership, announced in February 2026, shows how the platform consumes outside data. It integrates Draslovka’s real-time mineral and slurry monitoring into the Avathon system, feeding chemical and process measurements straight into the decision layer.

From sensor data to autonomous action

Sensing is only the first half. Camera feeds, sensor readings, and operational data flow into the AI models, which generate recommended responses as conditions shift on site.

For routine decisions, the platform is designed to act without waiting for an operator to initiate. Avathon describes it as enabling “end-to-end planning, decision intelligence, and autonomous actions across the mining value chain.”

What separates this from conventional process automation is coordination. A standard system watches individual machines in isolation; a platform running roughly 200 models against live sensor data is attempting to reconcile decisions across the entire mine at once. For investors, that distinction matters, because genuine value-chain optimisation is a different proposition to rebranded automation, and it is the difference that determines whether Barrick’s cost structure actually moves.

What separates Avathon’s approach from conventional automation is the move toward agentic AI systems that coordinate decisions across an entire site rather than optimising individual machines in isolation, a distinction that determines whether cost-per-ton metrics actually shift at the portfolio level.

Why Barrick is making this move now, and what it is trying to solve

Committing an entire North American business to a single autonomy platform is a large decision. It reads as a rational response to specific operational economics rather than trend-chasing when you look at the pressures Barrick faces.

Three forces are converging. Ore grades are declining across the industry, which makes tighter processing optimisation the difference between margin and loss. Labour constraints and safety obligations are pushing routine supervisory tasks toward AI. And ESG reporting increasingly demands data-rich operational traceability that fragmented, human-led systems struggle to provide.

The breadth of AI applications in mining now extends well beyond haul-truck autonomy into geological modelling, water management, and emissions accounting, which explains why Barrick structured the Avathon agreement to span exploration and supply-chain intelligence rather than limiting it to processing optimisation.

The performance-fee model addresses a fourth issue: incentive alignment. Because Avathon earns more only when outcomes improve, the structure reduces the principal-agent risk that plagues technology deployments where the vendor is paid regardless of results.

Avathon’s stated value proposition Higher yield, lower cost per ton, and more consistent, more predictable performance across operations.

The vendor’s reach extends well beyond mining, into aerospace, defence, manufacturing, and energy, which supports the argument that the platform is a mature industrial product rather than an experiment. It also arrives amid an aggressive commercial push. Barrick was the third major Avathon announcement inside a single month.

That clustering tells you Avathon is expanding fast, and it is worth factoring into how you assess the vendor’s bandwidth and prioritisation. The relevant question for your read on Barrick is whether the partnership is tackling durable structural challenges or reacting to cyclical cost pressure, because only the former holds up across different gold-price environments. On the evidence, the pressures Barrick is targeting are structural, not cyclical.

What the precedents say about how these programmes actually unfold

Barrick is not the first major miner to bet on large-scale autonomy, and the earlier programmes offer a grounded sense of timeline and risk that neither dismisses the ambition nor takes the vendor claims at face value.

Rio Tinto’s AutoHaul programme in the Pilbara is the most comparable precedent. It built one of the world’s largest autonomous heavy-haul rail systems alongside autonomous trucks and remote-operations centres, and it took roughly a decade to move from pilot to full deployment. Company case studies and coverage in Mining Magazine and S&P Global Commodity Insights report higher utilisation, fewer safety incidents, and lower cost per ton, achieved alongside complex software integration and extensive organisational change.

BHP’s Jimblebar iron-ore mine tells a similar story on a staged rollout, with productivity and safety gains earned through significant change management. Newmont and Freeport-McMoRan have both invested in remote-operations centres and AI-assisted control rooms, pointing to a clear sector trajectory rather than an isolated Barrick experiment.

Rio Tinto and BHP both reached comparable scale through staged programmes that took the better part of a decade, and the broader pattern of autonomous mining adoption across major producers shows that integration complexity, not technology maturity, is consistently the rate-limiting factor.

Barrick’s deployment carries four implementation risks worth watching:

• Cybersecurity exposure: unifying sensors, cameras, and control systems expands the attack surface for ransomware or control-system compromise
• Integration complexity: Nevada Gold Mines runs haul trucks, shovels, drills, and DCS/SCADA control systems built over decades, which are difficult to knit into one intelligence layer
• Workforce displacement: AI agents handling safety supervision, maintenance planning, and procurement can reshape or reduce traditional roles
• Vendor lock-in: a single-platform dependency limits future bargaining power and complicates any later migration

Proponents counter that standardised autonomy architectures simplify support, enable shared innovation, and produce data-network effects, with better models emerging as more sites feed the system.

The read for investors is straightforward. The precedents show the question is not whether industrial autonomy works at this scale, but whether Barrick and Avathon can navigate the integration and change-management challenges that slowed every comparable programme. The five-year duration looks like a realistic acknowledgement of that complexity, which means material benefits are more likely in the 2028 to 2030 window than in near-term results.

What this deal signals, and what investors should watch next

The technology thesis is set. What comes next is evidence, and there are a handful of specific indicators that will show whether the partnership is delivering.

The clearest sector signal is directional: major gold producers are shifting from discrete automation projects toward integrated AI operating models that span the whole value chain. Barrick’s decision to use a dedicated industrial autonomy vendor, rather than a general cloud-provider AI suite, is itself a commercial choice worth tracking, because it bets on mining-specific capability over broad platform reach.

Three checkpoints will tell you whether the programme is working:

1. Full Nevada Gold Mines deployment, expected within one to two quarters of the announcement, is the first execution milestone to confirm.
2. Barrick’s next quarterly operational reports, covering Q4 2026 and Q1 2027, are the earliest windows in which throughput and recovery gains at Nevada could surface in disclosed data.
3. Expansion beyond the Dominican Republic, the only named next-step location, would signal Barrick sees results worth scaling.

Keep the horizon honest. This is a five-year programme at a single major producer, and its trajectory over the next two to four quarters will be the first genuine test of whether the industrial autonomy thesis delivers at gold-mining scale.

The AI deployment at Nevada Gold Mines sits alongside a separate set of corporate questions about Nevada Gold Mines strategic options that Barrick has been navigating through 2026, and investors tracking the AI programme should hold both threads simultaneously when reading quarterly results.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results, and forward-looking statements are speculative and subject to change based on market developments and company performance.