AI TRENDS | Atreides’ Gavin Baker Says DRAM Remains the Key Bottleneck in the AI Race

Atreides Management founder and chief investment officer Gavin Baker said DRAM remains the most important bottleneck in the AI race, citing memory capacity and bandwidth as the foundation of AI model performance. According to Odaily, Baker made the remarks on a recent episode of the All-In podcast, where he discussed what he described as bottleneck issues in AI development. Baker said some people on social media focus on identifying bottlenecks by using Claude search and pointing to lesser-known Japanese materials companies, but he argued that DRAM is the only bottleneck that matters. He added that the development of companies such as ChangXin Memory Technologies could become an alternative for Apple to address high costs and could partially meet market demand for consumer-grade DRAM. However, Baker said only three companies can produce the DRAM required for AI servers: SK hynix, Samsung, and Micron. He also said DRAM could account for 30% to 40% of total hyperscale data center capital expenditures in 2027.