Livium Advances AI Battery Recycling Collaboration with Oscorp Energy

Livium Brings AI to the Battery Recycling Line in Collaboration with Oscorp Energy

Livium Ltd (ASX: LIT) has confirmed progress on its artificial intelligence battery sorting programme with Oscorp Energy, following the August 2025 Memorandum of Understanding (MOU). The Livium battery recycling AI collaboration with Oscorp Energy marks a significant step forward, with hyperspectral vision sensors now operating at Livium's Envirostream facility in Derrimut, Victoria. An AI-powered battery intelligence platform and dashboard are expected to be delivered in the coming weeks.

For Livium Ltd, which operates Envirostream as an established and profitable battery recycling business, this collaboration represents a move toward more data-driven, automated processing while maintaining current production.

"Deploying Oscorp Energy's hyperspectral vision technology at Derrimut and training AI models against our live operating data is a meaningful step toward smarter, more automated battery recycling. The intelligence platform is expected to give Envirostream greater visibility over its feedstock and material flows, supporting better planning, improved operational control and a foundation for autonomous sorting capability over time."

— Simon Linge, CEO and Managing Director, Livium Ltd

From MOU to Live Data: Phase 1 Deployment Completed

The Oscorp Energy collaboration is being progressed under a four-phase development roadmap that aims to take an AI-driven battery sorting module from proof of concept toward potential commercial deployment at Envirostream.

According to Livium Ltd, Phase 1 concentrated on a software-only pilot:

• Hyperspectral vision sensors were installed above Envirostream's existing feed conveyor at Derrimut.
• Batteries continued to move through the normal production line, with no mechanical diversion or interruption.
• The sensors recorded real-world images and spectral data from mixed battery streams.

This initial phase was intended to gather high-quality operating data without affecting plant throughput. The collected data is now being used by Oscorp Energy to train AI models to recognise different types of batteries within live, mixed streams.

The project has now moved from:

• Sensor deployment and data capture

to

• AI model development and operational analytics capability.

Phase 1 is reported as completed. The next stage is delivery of an AI-enabled battery intelligence platform and operational dashboard to Envirostream as part of Phase 2.

AI-Powered Battery Intelligence Platform: Role and Expected Capabilities

The upcoming AI-powered intelligence platform is described by Livium Ltd as a core tool to improve visibility over material flows at Envirostream's Derrimut facility.

According to the ASX announcement, the platform is designed to:

• Provide monitoring and analysis of material flow, battery categories and object counts across the processing line.
• Enable classification of battery chemistries and form factors directly from the live conveyor feed.
• Deliver trend analysis, including identification of operational patterns and exception items over time.
• Form a data foundation for improved resource planning, feedstock forecasting and process optimisation.

Battery recycling operations typically receive mixed battery streams with varying chemistries, sizes and formats. Real-time knowledge of what is arriving and moving through the plant is often limited.

Furthermore, Livium Ltd reports that this intelligence platform is intended to:

• Give Envirostream a more detailed understanding of incoming stock.
• Support more accurate planning of processing runs and resource allocation.
• Provide structured data that can underpin later automation and robotics.

Educational Section: Understanding Hyperspectral Vision and AI in Recycling

What Is Hyperspectral Vision?

Standard industrial cameras generally capture light in three colour bands:

• Red
• Green
• Blue

A hyperspectral camera, however, captures light across many more, narrower bands, including wavelengths outside the range of human eyesight.

Different materials reflect and absorb light differently at each of these wavelengths. This creates a unique "spectral fingerprint" for each material. By analysing this fingerprint, software can distinguish between materials that may look very similar in normal visible light.

In battery recycling, a hyperspectral camera placed above a conveyor can:

• Detect differences in battery chemistry (for example, lithium-ion, alkaline, nickel-metal hydride).
• Distinguish form factors, such as cylindrical cells, button cells or prismatic cells.
• Operate at line speed without physical contact with the batteries.

This is important for safety and efficiency, as it removes the need for manual sorting based purely on visual inspection and labelling.

How Does AI Model Training Work in This Context?

Oscorp Energy is using the real-world data from Envirostream's Derrimut conveyor to train AI models. In simple terms:

1. The hyperspectral cameras capture images and spectral data of thousands of batteries.
2. Each item of data is labelled by type (chemistry, size, format) to create a training dataset.
3. AI algorithms learn to recognise patterns in these data, allowing them to identify battery types automatically in new images.

Over time, as more data is collected and the models are refined, classification accuracy is expected to improve. This process is central to enabling automated, high-volume sorting in later phases of the project.

Key Terms Explained

For investors, understanding these fundamentals helps in assessing how technology could influence processing capacity, cost, and safety outcomes at Envirostream over time, subject to successful development and validation.

Robotics Roadmap: From Intelligence Layer to Autonomous Handling

According to Livium Ltd, the AI and analytics platform is the first stage in a planned robotics-enabled recycling system with Oscorp Energy.

The roadmap set out in the MOU includes four phases:

Once the vision and intelligence layer is operating and validated, the roadmap contemplates integration of a robotics layer capable of:

• Autonomous sorting of batteries and battery-related materials.
• Automated handling that may reduce manual contact with potentially hazardous items.

According to the announcement, if the pilot phases are completed successfully and agreed benchmarks are met, this progression is expected to:

• Support improved processing efficiency at Envirostream.
• Reduce manual handling requirements, with potential implications for worker safety and labour allocation.
• Create a pathway toward higher-volume autonomous operations.

For investors, the conditional nature of Phase 4 is important. The robotics deployment is explicitly described as being subject to performance results from the earlier phases, rather than being assumed.

Investment Perspective: Why This Matters for Livium Ltd

1. Technology Layered onto an Existing Business

Envirostream, a wholly owned subsidiary of Livium Ltd, is described as a profitable business that recovers valuable materials from end-of-life batteries. The Oscorp Energy collaboration is being implemented on top of this existing platform rather than as a stand-alone experiment.

This means:

• AI and automation are being built into a real operating environment.
• Outcomes can be measured against existing throughput, cost structure and safety metrics.

2. Potential Operational Leverage

Battery recycling margins are often sensitive to sorting accuracy, handling costs, and line downtime. If the intelligence platform and later robotics phase perform as planned, they may:

• Allow higher throughput without a proportional increase in staff.
• Improve planning based on better feedstock visibility.
• Reduce manual interventions on the line.

These are factors investors in Livium Ltd are likely to monitor as indicators of potential operational leverage.

3. Use of Real Operating Data, Not Synthetic Data

The ASX announcement emphasises that AI models are being trained on real operating data from Envirostream, including actual mixed battery streams, live conveyor conditions, and real-time lighting and environmental variables.

In AI systems, training on real-world data generally improves the probability that models perform effectively under real operating conditions. Consequently, this approach may reduce the risk that models behave differently when moved from a laboratory setting into a production environment.

4. Alignment with Broader Growth Areas

Livium Ltd has stated that it is expanding into adjacent areas, including recycling of rare earth elements, solar panel recycling, and black mass processing.

Improvements in vision, AI and automation at Envirostream could, if successful, provide a technology base that may be adaptable to these adjacent waste streams, as well as a more scalable foundation for higher-volume processing in future. This alignment between current operations and expansion areas is a key context point for investors assessing the long-term role of technology at Livium Ltd.

What Investors Should Watch Next

The ASX announcement outlines several near-term and medium-term milestones that may act as reference points for the market:

• Platform delivery and commissioning — actual delivery of the AI-powered battery intelligence platform and dashboard to Envirostream in the coming weeks, and successful commissioning and integration with existing operations.
• Initial analytics and operational data — early performance of the platform in monitoring material flows and classifying batteries, as well as the quality and usefulness of insights for planning and process control.
• Phase 3 validation outcomes — the extent to which the platform meets agreed performance benchmarks, and any disclosure from Livium Ltd on classification accuracy, throughput impacts or operational improvements.
• Decision on Phase 4 robotics deployment — whether the conditions are satisfied for progression to the physical robotics layer, and any indicative timelines or scopes if the project moves ahead.

Investors tracking Livium Ltd may also consider how this AI initiative fits alongside other company announcements on recycling capacity, new feedstock sources and adjacent technology programmes.

Summary: Livium's Path Toward Data-Driven Battery Recycling

The Livium battery recycling AI collaboration with Oscorp Energy, progressed through Envirostream, is developing an AI-based battery sorting capability under a structured four-phase roadmap. Phase 1, involving deployment of hyperspectral sensors and data capture at the Derrimut facility, is reported as complete. Phase 2, centred on an AI-powered intelligence platform and operational dashboard, is expected to reach a key milestone with platform delivery in the near term.

By focusing first on visibility and analytics, the collaboration aims to build a verified data and AI layer before considering robotics-enabled autonomous sorting. The project remains conditional at each stage, with further progression subject to performance against agreed benchmarks.

For investors, the initiative presents a technology-focused development within an already-operating and profitable battery recycling business. The key areas to watch are actual platform performance, validation results, and any subsequent decisions regarding robotics, as these will inform the potential operational and financial impacts on Livium Ltd over time.

Want to Learn More About Livium's AI-Driven Battery Recycling Business?

Livium Ltd (ASX: LIT) is advancing a structured, four-phase AI and robotics programme at its profitable Envirostream facility — built on real operating data, not laboratory conditions. With Phase 1 complete and an AI-powered battery intelligence platform imminent, the company is laying the groundwork for a more automated, data-driven recycling operation. Investors looking to follow this programme's development and understand what it could mean for Livium's broader growth strategy can find out more by visiting the Livium investor hub.