Presentation Description: Core chips within a server installation have traditionally been powered by multiphase DC/DC converters surrounding the target and its heatsink, held up by a large array of capacitors underneath. Modern CMOS power density and Artificial Intelligence (AI)/Machine Learning (ML) power demands are such that this lateral power delivery scheme is no longer capable of holding up the supply voltage properly with changes in compute load. This scheme is limiting the maximum compute power of the server itself. Moving the DC/DC converters themselves to the space directly underneath the target allows that supply voltage to be supported properly and enables higher compute power. It also removes the lateral transmission loss associated with the traditional scheme, improving efficiency and reducing heat generation. This vertical topology requires fast converters and co-designed silicon, inductance, and packaging to achieve the power density needed to fit in that space while still remaining thin enough to allow an effective cooling solution.