Interview summary says SpaceX and xAI are building out their own AI supply chain as PCB and optical modules emerge as bottlenecks

ABMedia, citing a Substack roundup by WK of an internal expert interview involving SpaceX and xAI, reported that the group is targeting a sharp expansion in AI compute capacity over the next two years. The interview said current capacity is about 1.3GW, with plans to reach 2.3GW by December 2026, 4GW to 5GW by April to May 2027, and 10GW to 12GW by November to December 2027. A longer-term target of 20GW in 2028 was also outlined. The reported buildout is centered in the US, mainly in Memphis and Southaven, while land and permits in Texas have already been secured but construction has not started. Power plans mentioned in the interview include a mix of grid supply, gas turbines, Tesla solar, and Megapack backup. Of the first 10GW, around 60% is expected to be used internally by xAI, SpaceX, and Tesla, with the remaining 40% allocated to outside customers including Anthropic and Google. The interview also identified PCB, optical modules, transformers, and cooling hardware as the most difficult supply-chain constraints. It said the company has a 43-person supply-chain team and keeps a 20% to 40% inventory buffer for key hardware such as GPUs, while also exploring deeper vertical integration across cables, fiber, racks, and cooling systems.