Will a material discovered primarily by AI systems be deployed in a commercially sold battery, solar cell, or catalyst product before 2031?

TL;DR

I forecast a 69% chance of YES. SES AI is the strongest route: it has named AI-discovered drone-battery electrolytes and a converted production line, but its latest authoritative disclosure described drone shipments as samples for qualification rather than qualifying commercial sales (SES materials catalogue; Q1 2026 shareholder letter). The estimate is below the team's 71–79% range because electrolyte sales alone may not count as battery-product sales, public attribution is a real bottleneck, and the fallback routes remain early.

Context

I found no example that clearly satisfied all five conditions by the forecast cutoff. The closest candidates have crossed parts of the chain—AI discovery, synthesis, product integration, customer testing, or initial sales—but none has publicly tied a specific AI-primary material to an ordinary commercial product line with at least $1 million in qualifying revenue.

The question is easier than mass-market adoption. It needs one product anywhere in the world, and $1 million is a small threshold for batteries or industrial catalysts. But it is harder than the usual AI-materials headline because the resolver must be able to trace the discovery, the incorporated material, the commercial product, and the revenue.

Evidence

The historical backbone is slow. The National Academies found that a viable catalyst can take 10–15 years to reach commercial plant start-up (National Academies). The IEA says AI can shorten materials search, but that most energy-related successes remain concentrated in the earliest innovation stages and still face laboratory and scale-up bottlenecks (IEA). A recent perovskite review reaches the same conclusion: commercialization is now constrained more by manufacturing, reliability, standards, and product economics than by finding another promising laboratory material (Nature Reviews Clean Technology).

This reference class is not fully controlling. A drop-in electrolyte, additive, catalyst coating, enzyme, or interface material can use an existing product architecture. It can reach $1 million without an automotive launch or a new chemical plant. The fastest route is therefore likely to look mundane: a proprietary formulation inserted into a niche drone battery or an existing catalyst component, not a celebrated GNoME crystal becoming a new mass-market chemistry.

SES is the clearest battery route. Its fiscal-year 2025 filing says Molecular Universe discovered six novel electrolyte materials that were being tested by more than 40 potential customers, with commercial-scale manufacture through Hisun planned for the second half of 2026 (SES 2025 Form 10-K). Its catalogue gives the materials distinct identifiers and maps SES-S8.62 to high-silicon/NCM811 drone cells (SES materials catalogue). By the first quarter of 2026, SES had converted its Korean line, planned capacity above one million drone cells a year, shipped samples to defense and commercial prospects, and moved about six materials customers into second-phase testing (SES Q1 2026 shareholder letter).

That is strong pipeline evidence, not resolution evidence. SES described the cells as samples for evaluation and qualification, said first-quarter results were driven mainly by its acquired conventional energy-storage business, and provided no revenue breakout proving $1 million of ordinary sales from batteries containing SES-S8.62 or another named AI material (SES Q1 2026 shareholder letter). I assign SES a 48% standalone chance. This includes its own drone cells and the chance that a customer sells a battery containing an SES material with enough public documentation.

Calicat is the best independent catalyst route. The company says AI begins and ends its catalyst-development process, that AI-predicted catalysts are physically tested and synthesized, and that it has brought non-iridium oxygen-evolution catalysts and coatings to market (Calicat). Its Amplifier product is a catalyst-coated membrane sold as an iridium replacement for PEM electrolyzers (Amplifier). The missing pieces are public sales figures and a clean technical document linking Amplifier's particular catalyst composition to an AI-primary discovery. I give this route 18%.

Aionics adds a smaller battery route. It says it has developed nonflammable and low-temperature electrolyte formulations with aerospace and cell-manufacturing partners and intends to move these drop-in formulations into commercial production through manufacturing partners (Aionics). The customers, formulations, deployed products, and revenue are mostly confidential. I assign Aionics and other non-SES battery programs 13% collectively.

Solar has scale but weak attribution. XtalPi and JinkoSolar formed a closed-loop AI and robotics venture for perovskite–silicon tandem cells, targeting about 1,000 experiments per day and projecting mass production in roughly three years (XtalPi). JinkoSolar's filing confirms AI-driven R&D and commercialization work but does not identify a specific AI-discovered absorber, interface material, or additive (JinkoSolar 2025 Form 20-F). If a qualifying material reaches a Jinko product, the revenue threshold becomes easy; the hard part is ensuring the eventual advance is a material discovery rather than process optimization. I assign solar 10%.

Two weaker edges add diversification. Arzeda says it has launched AI-designed enzyme products and is producing formulation-ready designer enzymes for home and personal care (Arzeda milestones; home-care programme). An enzyme sold as a catalyst or present in a commercial formulation could qualify, but an AI-designed enzyme used only to manufacture a separate sweetener would not. CuspAI has raised $450 million and assembled more than 45 laboratory and industrial partners, including battery, photovoltaic, and catalyst companies (Reuters; CuspAI); its most detailed industrial project nevertheless remains at the stage of roughly 20 PFAS-removal candidates moving into development and testing (Kemira). I assign 12% to enzymes and 12% to future or otherwise unmodelled programmes.

My route assumptions are therefore 48% for SES, 18% for Calicat, 13% for other batteries, 10% for solar, 12% for enzymes, and 12% for future or other programmes. Treating them as independent gives 74.1%. A Gaussian-copula aggregation with 0.15 shared correlation—representing common scale-up, attribution, disclosure, and resolver-interpretation failures—gives 68.6%, which I report as 69% in prose.

What's non-obvious

The main correction to the bullish view is semantic but consequential. Selling an AI-discovered electrolyte through Hisun does not necessarily satisfy the question. The resolution language requires a commercially sold battery that incorporates it. That adds downstream product qualification, battery sales, and public product-material linkage to the SES materials route.

The evidence bottleneck may be harder than the revenue bottleneck. One defense-drone order or electrolyzer-component contract can clear $1 million, but manufacturers often hide exact formulations and private companies rarely publish product-level revenue. I also excluded a widely cited study claiming large materials-discovery and product gains among 1,018 scientists because MIT later raised data-integrity concerns and recommended that the paper be withdrawn (The Atlantic editor's note).

Limitations

The largest unknowns are private-company sales, proprietary chemistry, and future disclosure quality. Calicat, Aionics, and Arzeda could already have more commercial traction than public sources reveal; silence is not evidence that they do not. Conversely, manufacturer descriptions such as AI-discovered or AI-designed may later prove to mean AI-assisted optimization rather than primary candidate identification. The enzyme route also depends on whether a resolver treats an enzyme product as a qualifying catalyst product and requires the enzyme to remain present in the sold formulation. My reasonable interpretation-sensitive range is about 55% to 82%.