Moët Hennessy Uses AI Sensors to Catch Fresh Mushroom Aroma Early

Moët Hennessy, Analog Devices and researchers at the University of California, Davis are using sensors and machine learning to detect early chemical signs of wine defects, a step they say could let wineries spot quality problems sooner and at lower cost than many conventional testing methods.

Details released Tuesday describe a system that analyzes the chemical signatures emitted by grapes and wine and uses artificial intelligence to flag patterns linked to future quality risks. The work builds on earlier research that began at the Massachusetts Institute of Technology and is now being applied to wine production through the partnership among the luxury drinks group, the chipmaker and the California university.

The first major target for the project is a defect known as Fresh Mushroom Aroma, or FMA, an unwanted smell that can affect wine quality and can be difficult to identify before it becomes obvious. According to the collaborators, researchers at Moët Hennessy’s Robert-Jean de Vogüé Research Center assembled a library of samples and related data to train machine learning models to recognize the chemical patterns associated with that risk.

Using Analog Devices’ sensing and machine learning platform, the system processes a broad range of chemical data rather than looking for a single marker. The platform then identifies which signals matter most and searches for links that may not be visible through standard analysis. The companies said that approach allows the model to keep learning from complex biological and chemical information instead of checking only for preset conditions.

In the current study, the technology identified samples that showed an elevated risk of developing Fresh Mushroom Aroma at an early stage, before the defect would typically be found through traditional methods, according to the information released Tuesday. That result is important because wine defects are often easier to manage before they become fixed in the product, when corrective options may be limited or expensive.

Manuel Reman, a member of Moët Hennessy’s executive committee who oversees research and development, said the project could change how producers understand grapes and wine during production. In comments released with the announcement, he said the ability to detect and interpret the “chemical language” of grapes and wine could help wineries anticipate risks before they become visible and improve resilience over time.

The project reflects a broader push to bring advanced sensing and AI into food and beverage production, where quality can shift quickly because of biological activity, weather, storage conditions and fermentation. In wine, those changes can begin in the vineyard and continue through crushing, fermentation, aging and bottling. Finding reliable chemical signatures early in that process has been a long-standing challenge because the compounds involved can be subtle and may interact in complex ways.

The partners said their system is designed to address that problem by combining a large sample base with pattern recognition tools that can sort through many variables at once. Instead of depending only on human tasting panels or slower laboratory confirmation after a defect has started to appear, producers could use earlier readings to decide whether to separate lots, adjust handling or monitor certain batches more closely.

For the beverage industry, the potential value goes beyond wine. Earlier detection of off-notes or instability could help producers reduce waste, avoid sending flawed product further down the line and shift quality control from late-stage correction toward prevention during production. In a sector where margins can be damaged by rejected batches, product downgrades and brand risk, even small gains in early warning can matter.

The announcement did not provide commercial rollout dates or say when the system might be used at scale across Moët Hennessy operations. It also did not disclose performance figures, the size of the training data set or whether the same method is being tested for other wine defects. Even so, the reported results suggest that chemical sensing paired with machine learning is moving from research into more practical use inside wineries.

UC Davis has long been a major center for wine and viticulture research, and its role in the project adds academic support to the industry effort. Analog Devices brings sensor and signal-processing expertise, while Moët Hennessy contributes access to production knowledge and sample data. That combination is meant to close a gap that has limited earlier attempts to use digital tools in winemaking: having enough real-world, well-characterized samples to train models that can perform under production conditions.

The collaborators said the broader goal is not only to find one specific defect but to build a deeper understanding of the biological activity taking place in grapes and wine as they move through production. If that approach proves reliable, it could give wineries a more continuous picture of emerging risks and help them intervene earlier, before sensory problems become obvious in the finished wine.