Energy and land emerge as primary AI compute constraints

Walter Goodwin identified a shift in the primary constraints for AI compute expansion, moving from GPU shortages to acute shortages in HBM memory, logic silicon, and ultimately land and power. He stated that grid connection queues in the US and UK are an order of magnitude larger than available energy capacity. Goodwin contrasted China’s regulatory environment, which he described as 'greenlight anything' and capable of moving millions of people for energy sites, with Western NIMBYism and environmental regulations that delay projects. He argued that China has 'stolen the march' on the West by leveraging mega hydro projects for AI data centers, where 15 gigawatts of power can be delivered to a single site.

Goodwin emphasized that because data center capacity is the fundamental bottleneck, there is intense pressure to deploy chips that maximize intelligence per megawatt. He noted that current HBM-based chips are inefficient for high-speed inference, creating a need for specialized silicon that can handle long-context, long-running agent workloads without excessive power consumption. In the US, he cited NIMBYism as a major obstacle. In the UK, he pointed to protections for bats and butterflies preventing sites of scientific or natural beauty from being developed. Goodwin argued that these Western regulatory hurdles place the US and UK at a disadvantage compared to China in securing the energy infrastructure required for AI dominance.