Recently, chip giant Broadcom announced a major financial and technological strategy: collaborating with Apollo and Blackstone to establish the AI XPV platform, aiming to build next-generation computing infrastructure supporting cutting-edge AI laboratories through large-scale investment.

This platform's establishment is truly impressive. As a key technology provider, Broadcom will provide its customized XPU and advanced networking solutions to meet the efficient computing power transmission needs for large model training. As initial anchor investors, Apollo and Blackstone's credit and insurance businesses will offer strong financial support, injecting an initial 3.5 billion US dollars.

The core objective of this investment is clear: to support the computing power expansion needs of leading AI laboratories such as Anthropic and OpenAI. According to the plan, the AI XPV platform aims to deploy more than 20 gigawatts of computing power by 2028. This figure not only demonstrates the vision of tech giants regarding computing power demand but also means that AI infrastructure has officially entered the industrialized construction phase of "energy-intensive" development.

Currently, the initial deployment is about to begin. It is reported that the relevant facilities will be implemented at Fluidstack sites starting in mid-2026, focusing on supporting Anthropic's previously announced expansion plan of over 1 gigawatt of computing infrastructure.

This initiative, combining chip manufacturers and financial capital, sends an important market signal: the AI industry is no longer just pursuing algorithm iteration but has entered the "deep water zone" of competing in underlying energy and hardware capacity. With this 3.5 billion dollar investment, the global AI computing power supply structure may undergo reshaping in the next two years.