<p>A new preprint published on 24 September 2026 describes an AI-guided, human-supervised laboratory workflow for a difficult green-hydrogen materials problem. The work targets acidic oxygen evolution at the anode of proton-exchange-membrane water electrolysers, where iridium and ruthenium have been widely used but create concentrated and constrained material supply chains.</p> <p>The platform combined combinatorial sputter synthesis, high-throughput screening, machine-learning composition-property models, adaptive optimisation, and large-language-model reasoning. The authors report evaluating 2,942 catalysts across 53 material systems and 26 elements. The system surfaced complex palladium-oxide candidates that conventional design logic might not have prioritised, including compositions containing indium, manganese, nickel, tantalum, and palladium.</p> <p>The most important detail is that the model did not replace experiments. Candidate materials were synthesised, screened, and advanced to longer validation. The paper reports that one composition retained an overpotential below 0.5 volts over more than 1,000 hours under the stated test conditions. That is a research result, not a commercial electrolyser qualification, a guaranteed cost reduction, or proof that palladium can immediately replace incumbent materials at scale.</p> <p>For industrial readers, this is a compelling example of AI as a search-and-test accelerator. It also shows why material discovery must be connected to procurement reality. Abundance, price, processability, corrosion resistance, coating uniformity, recycling, qualification time, and supplier capacity all matter after a promising laboratory result. The strongest workflow is one that records both the performance evidence and the material dependencies that could constrain deployment.</p> <p><strong>Source:</strong> <a href="https://arxiv.org/abs/2609.30133" target="_blank" rel="noopener noreferrer">Jenewein et al., AI-guided high-throughput discovery of iridium- and ruthenium-free palladium-oxide catalysts</a>.</p> <p><strong>Further reading:</strong> <a href="https://supplybridge.com/automotive-should-costing">should-costing and evidence-aware technical assumptions</a>.</p>