'Among the titles I like best is Arista Networks'

The supply chain for AI infrastructure is largely located in Asia, with countries such as Taiwan and Korea being the main suppliers of semiconductors. Tariffs on these products would undoubtedly increase costs for US buyers of AI infrastructure, but demand is relatively inelastic, as remaining highly competitive in AI development is critical for these companies. We expect the Trump administration to prioritise the reduction of tariffs on these key AI infrastructures, given the strategic importance of the sector.

Symbotic: provider of warehouse automation systems based on artificial intelligence. Unlike other automation solutions that focus on a single product or process, Symbotic's platform takes care of inventory management in a distribution centre in the round, without the presence of human staff in the main areas of the warehouse. Walmart, the largest retailer in the US, has contracted Symbotic to implement the Sym robotics platform in all its regional distribution centres over time. Vertiv: is a leader in thermal management (cooling) solutions in the data centre market, including advanced liquid cooling systems required by next-generation CPUs. The company has a partnership with Nvidia and is one of the main suppliers that Ia companies turn to when they want to build or expand an Ia data centre. Vertiv's other products (such as power supply) are also aimed at the data centre market, making it a pure beneficiary of the race for Ia hegemony. Arista Networks: is the leading supplier of data centre switches that turn independent servers into a single processing engine, transfer data to and from the Ia models as needed, and provide secure system access. Large language models and other forms of Ia require huge amounts of computing power and data access, and therefore extremely fast networks to minimise training time and costs. Reddit: artificial intelligence models are only as effective as the volume and quality of data they are trained on, and Reddit is one of the best sources of up-to-date textual content on the Internet. In just a few weeks, the platform is able to create as much trainable data as Wikipedia. Reddit has data licensing agreements with OpenAI and Google to provide training data for their large language models, and we anticipate further agreements in the future.