Export of bananas in America has been built on an extremely limited genetic basis: over 70 years, a single variety of bananas, Gros Michel, was almost the only one to be sold on the US markets. This variety produced large bunches which were resistant to transport and had a taste and rind which consumers in the USA identified as belonging to quality bananas. Meanwhile, Gros Michel has also shown itself very susceptible to a large number of fungal pathogens, including the Panama Mal and Sigatoka. The historical dynamics brought about by the spread of that fungus during the first half of the 20th century, accelerated the increase in the rate of run-off, destabilised rural livelihoods, increased the health risks for workers in the countryside, and limited the income of the main banana trading companies. These epidemics have pushed the UK Government and the United Fruit Company to set up phytoplating programmes during the 1920s with the aim of developing a banana for export that would be resistant to the Panama Mal. However, the creation of a hybrid that would be able both to thrive in tropical areas and to find acceptance on the US market has proved to be a difficult task to accomplish. The history of breeding programmes reveals one of the main contradictions of 20th century agriculture: the same massive production processes, which tend to reduce biological diversity at local and regional level, remained dependent on access to a “global” genetic bank in order to maintain profitable levels of production.