文章指出随着AI计算密度增加，数据移动、冷却和光学互连成为影响性能和扩展的关键瓶颈。
As AI compute densifies, data movement, cooling and optical interconnects are becoming critical to performance and scale.