AI 'Bottleneck Trades' May Be Coming to an End, Early SpaceX Investor Says Market Focus Will Shift t

On June 17, discussions around the popular trades related to the shortage in the AI supply chain may be nearing their conclusion, as some investors begin to shift from seeking the next shortage link to identifying companies that will maintain long-term competitiveness after the AI infrastructure cycle. Gavin Baker, managing partner at Altreides Management and an early investor in SpaceX, stated in an interview with TBPN that the market has been keen on chasing bottleneck assets in AI construction over the past year, including DRAM, storage chips, and key material suppliers. However, he believes that these 'AI bottleneck trades' are approaching their end. Baker pointed out that Ajinomoto in Japan has refused to raise the price of a key chip packaging insulation material, indicating that the pricing power of some supply chain bottlenecks may be starting to ease. This material is used to connect processors and the packaging layers of chips, which had previously drawn investor attention due to the surge in demand for AI chips. He noted that the past market game was to find the 'next bottleneck', but the more important question for the next phase is which companies will retain enduring franchise value after these bottlenecks dissipate. This perspective has cooled off the recent surge in storage and AI material stocks. Stocks like Micron and SanDisk have seen significant increases this year, driven by capital expenditures in AI data centers, HBM demand, and long-term procurement agreements leading to a reevaluation of supply and demand. However, as stock prices have risen rapidly, the market has begun to discuss whether the trades have become overcrowded. Baker also mentioned that the next main line of AI infrastructure may shift towards the actual capabilities of computing power and electricity. He highlighted SpaceX's potential in ground and orbital AI data centers, stating that the market will focus on the speed at which it can increase its 'gigawatts' level of computing power supply. This suggests that AI trades may transition from merely betting on shortages to a broader competition in infrastructure: those who can secure land, electricity, GPUs, and data center capabilities faster may find themselves in a more advantageous position in the next phase of the AI investment cycle.