Deep Tide TechFlow message, September 20: UBS currently estimates that in 2026, global AI capital expenditures will approach $1 trillion, and in 2027 will rise further to about $1.4 trillion. The main reason behind this growth is a sharp increase in memory costs. UBS estimates that memory-related spending will jump from $71 billion in 2025 to $367 billion this year, reaching $923 billion in 2027. Other AI-related costs are expected to be $631 billion in 2026 and decline to $525 billion in 2027. This means that this year, the increase in memory costs accounts for about 60% of the incremental AI capital spending. And by 2027, as spending on other components declines, the growth in memory costs will even exceed the entire net increase in AI capital expenditures. (Jin10)