What will be the highest critical temperature at ambient pressure of any non-cuprate superconductor that AI first proposed, as reported by 2030-12-31?

The 5.4 K Baseline The absolute floor for this forecast is 5.4 K, set by the intermetallic Mo20Re6Si4. A deep-learning network proposed this material before its synthesis, and its peer-reviewed publication demonstrates zero electrical resistance, a full Meissner signal, and a specific-heat jump confirming bulk superconductivity Deep Learning-Based Superconductivity Prediction and ... Deep Learning-Based Superconductivity Prediction and ... https://doi.org/10.1140/epjp/s13360-024-05947-w. No subsequent AI-proposed material has cleared this bar. Other confirmed AI-predicted superconductors have exhibited lower transition temperatures, such as the Be-Hf-Nb compounds around 3.2–4.2 K Developing a complete AI-accelerated workflow for ... Developing a complete AI-accelerated workflow for ... Developing a complete AI-accelerated workflow for ... and various kagome borides below 1 K Machine-learning-guided discovery of kagome ... - APS Journals AI just supercharged the race to find room temperature .... Higher-temperature claims have universally failed the stringent validation criteria—for instance, the ~9 K Zr–In–Ni candidate lacked zero-resistance measurements and phase isolation Closed-loop superconducting materials discovery https://www.nature.com/articles/s41524-023-01131-3, while in other generative studies, the physically synthesized samples completely failed to form the computationally predicted ordered structures https://www.nature.com/articles/s41524-026-02117-7 https://www.nature.com/articles/s41524-026-02117-7. The 10th percentile therefore remains at 5.4 K, allowing for the possibility that no qualifying discovery exceeds the existing record.

Routes to Modest Improvement A moderate improvement into the 7.2–20.0 K range represents the most likely outcome, with those values marking the 25th and 75th percentiles and a median of 11.5 K. Autonomous lab throughput and AI screening capabilities are advancing rapidly, with high-volume experimental efforts actively targeting novel conventional intermetallics, borides, and carbides. Human-in-the-loop acceleration also matters: researchers can use AI screening alongside experimental feedback to guide successive synthesis attempts, supporting modestly stronger central outcomes than screening throughput alone would suggest. This is consistent with the shared assumptions about human-in-the-loop acceleration underlying related forecasts. While the record has remained stagnant for nearly two years, the sheer volume of machine-learning screening—coupled with iterative physical synthesis campaigns—makes it highly probable that at least one new conventional material exceeding 5.4 K will be validated and published by the end of 2030. However, screening throughput does not translate directly into successful synthesis, limiting how far above the existing record the central outcomes are likely to reach.

Synthesis Bottlenecks and Physical Constraints Extrapolating AI's recent exponential gains in mathematical and coding capabilities directly to material critical temperatures is a mistake. Physical synthesis remains a severe bottleneck. Computational predictions routinely fail to survive the transition to the lab: in one large-scale workflow, candidates predicted to superconduct at 14.9 K yielded measured onsets of only 3.18 K Developing a complete AI-accelerated workflow for .... Highly touted theoretical predictions, such as quaternary hydrides computationally predicted to reach 68 K at ambient pressure Machine learning accelerated search for superconductors in ... Superconductivity in atom-intercalated quaternary hydrides ..., face massive thermodynamic instability and non-equilibrium synthesis barriers Machine learning driven exploration of hydride .... Furthermore, there is still no predictive theory for unconventional pairing, leaving AI systems without a reliable physical framework to intentionally design high-Tc unconventional superconductors. These constraints weigh especially heavily against the upper tail: identifying a promising computational candidate is insufficient unless it can be synthesized and its superconductivity rigorously confirmed within the forecast window.

The Upper Tail The best conventional superconductor at ambient pressure, MgB2, maxes out at 39 K, creating a soft physical ceiling for routine algorithmic discoveries. The 90th percentile of 34 K allows for a substantial conventional-material breakthrough while keeping the upper tail below that benchmark. Human-in-the-loop acceleration supports progress in the middle of the distribution more directly than it supports overcoming the physical barriers to exceptionally high transition temperatures. Pushing the record into the 40 K regime would likely require an AI system to successfully identify a metastable ambient-pressure hydride or propose a novel synthesizeable variant within known higher-Tc non-cuprate families, such as iron pnictides or nickelates. Given the strict requirement for both zero resistance and a Meissner effect, plus the typical 6-to-18-month lag for peer review, these higher-temperature scenarios remain realistic but constrained tail outcomes.