X-ray structures of reactant and product states, as well as theoretical calculations, have helped uncover the mechanism of Escherichia coli phosphofructokinase 2.

Optimization of catalytic stereoselectivity for new substrates often requires a time consuming experimental process, and high-accuracy molecular modelling remains intractable for comprehensive virtual screening. Now, highly enantioselective rhodium hydrogenation catalysts have been identified using a rapid computational transition-state analysis protocol and then experimentally verified.

Models play a significant role in the development of catalysts. However, they are constructed using a reductionist approach and this poses the question of their relevance for the comprehension of physical phenomenon.

Databases of computational results hold high promise for accelerating catalysis research. Still, many challenges remain and consensus on facets such as metadata, reliability and curation is crucial to transform the hype into an attractive technology.

Typically, catalysts are discovered through trial and error coupled with chemical intuition. Now, an automatic machine-learning framework has been developed that can guide itself to find intermetallic surfaces with desired catalytic properties.

Catalysis research has immensely benefited from the use of high-performance computing facilities. On the occasion of the twenty-fifth anniversary of the first Top500 list, we briefly revisit its content and evolution and the impact that supercomputers have had in catalysis.