Stratum AI is now being linked to a publicly reported mineral resource estimate at St Barbara’s operating Simberi gold mine in Papua New Guinea, a development that matters less because it is flashy and more because it sits inside a formal JORC Code reporting framework. The claim, as reported by International Mining and echoed by Stratum AI, is that the company’s weathering and material-type models are helping underpin the resource estimate for an active mine, not just a conceptual study or a desktop exercise. For investors, that places the discussion on execution, reconciliation, and operating discipline rather than on abstract AI branding.
The relevant models classify material into oxide, transitional, and sulphide domains. That sounds technical, but the business logic is straightforward: if a miner can better tell one type of ore from another, it can route material more efficiently through the plant and avoid sending unsuitable ore into a circuit that is not designed for it. In this case, St Barbara has been using Stratum AI’s deep-learning work over multiple years at Simberi, and the models were first tested in pilot programs before being applied to operational data. That progression matters because it suggests the system was not adopted on speculation alone.
According to the source material, the models were validated against grade-control reconciliation before St Barbara adopted them as its preferred approach for material-type classification. That is a meaningful step. Grade-control reconciliation is where a mine checks whether the modelled understanding of orebody performance matches what is actually being mined. If the model survives that test, it becomes more relevant to mine planning and reserve conversion. If it does not, it remains a theory. The reported adoption implies St Barbara concluded the AI approach improved enough on conventional methods to earn a place in the operating workflow.
The economic point is not that artificial intelligence is automatically valuable. It is that operational mines generate large volumes of geological and production data, and the value of that data is only realized if it can be turned into better decisions. At Simberi, the reported benefit was better classification of ore that could be treated through the existing carbon-in-leach plant for oxide ores, while sulphide material could be stockpiled for later treatment. That is a practical mine-planning issue, not a marketing slogan. Better classification can reduce waste, lower costs associated with handling material that does not belong in the current circuit, and reduce wear and tear on the mill.
The source material says this AI-assisted classification helped St Barbara identify additional treatable material and extend the oxide processing life at Simberi by 17 months. It also says the modelling helped build Simberi’s 2024 resource and reserve update and the 2025 feasibility study. Those are not trivial milestones. Resource updates and feasibility studies are the points where geological interpretation meets capital allocation and operating assumptions. If the model meaningfully changes those inputs, it can influence the mine plan, the cost profile, and the timing of future capital decisions.
What sets this story apart is the claim that Stratum AI’s models are underpinning the publicly reported mineral resource estimate for an operating mine. That is the part investors should focus on, because it moves the discussion beyond a pilot tool. International Mining describes this as one of the first cases of its kind in mining, but that framing should be treated carefully. The evidence pack does not independently benchmark that claim against the broader industry, so it should be read as a company-led characterization rather than a proven market fact. Even so, the underlying point remains: the models are being used inside a formal reporting pathway tied to the JORC Code.
That reporting framework matters because JORC is incorporated into listing guidelines companies must meet to remain listed on the Australian Securities Exchange. In other words, this is not just an internal planning aid. It sits within a disclosure regime that is meant to enforce consistency, transparency, and technical discipline. For a junior or mid-tier producer, having a data model influence a JORC-aligned resource estimate can be significant if the work holds up under scrutiny. But it also raises the bar. Anything used in that process must be defensible, reproducible, and credible to auditors and investors alike.
There are a few hard datapoints in the story that help frame the opportunity, and the limits. In May 2024, St Barbara said it had identified an additional 143,000 oz of treatable material at Simberi, extending the mine’s current oxide processing life by 17 months. A month later, the company said Stratum’s modelling had helped increase proved and probable ore reserves by 40%. Those are material improvements on paper, but they are still company-reported outcomes tied to a specific operating context. Investors should not automatically generalize them to other mines, ore bodies, or processing setups.
The transaction backdrop also matters. St Barbara released the study on December 10, 2025, the same day it announced a strategic transaction with Lingbao Gold Group that values 100% of the New Simberi gold project at approximately A$800 million, or $562 million. The evidence pack does not provide enough detail to judge how much of that valuation reflects geology, mine life, capital intensity, gold price assumptions, or strategic optionality. But the coincidence of a feasibility study release and a transaction announcement tells investors that Simberi is being positioned as a value-bearing asset in a live corporate process, not as an isolated technical case study.
There are also clear limitations. The evidence does not show a new filing date or vote tied specifically to the AI milestone, and it does not provide independent verification that the models are uniquely innovative versus other geological classification tools. The “one of the first cases of its kind” framing is promotional and not independently confirmed. More broadly, using AI in mine planning does not eliminate geological uncertainty. Ore boundaries remain uncertain, reconciliation can change with more drilling, and operating performance can still be affected by plant availability, dilution, mining selectivity, and commodity price swings.
Another point worth stressing is that this is a collaboration, not a standalone AI takeover of the mine plan. Stratum AI worked with St Barbara for over four years, and the models were developed and refined through pilot programs, operational data analysis, and validation against mine performance. That is a gradual implementation path, which is usually what you want to see in mining technology. It also means the value of the models depends on ongoing data quality and operational discipline. If the inputs deteriorate, the outputs will too. The technology may improve decisions, but it does not replace geology.
For investors tracking either St Barbara or mining-tech adoption more broadly, the next issue is whether Simberi’s feasibility and expansion pathway continues to convert technical gains into operating and financial gains. The evidence pack identifies the Simberi expansion and feasibility-study pathway as the key next catalyst, but it does not provide a new filing date or a vote specifically linked to this AI milestone. That means the immediate investment question is not whether AI was used, but whether the mine continues to produce better resource confidence, better ore routing, and a stronger operating plan.
If those outcomes persist, Simberi becomes a useful example of AI embedded inside a live mining asset with JORC reporting implications. If they do not, the story becomes a reminder that even useful models must prove themselves under changing ore conditions and real operating constraints. For now, the evidence supports a cautious reading: Stratum AI’s models appear to have moved from pilot work into a meaningful mine-planning role, but the real test is still whether that work continues to improve Simberi’s economics in the field.
