UBS raises AI capex forecast: approaching $1 trillion by 2026, with 90% of the increase coming from memory price hikes

BlockBeats news, September 19 — UBS's latest estimates show that global AI capital expenditure will reach $998 billion in 2026, nearly doubling from $506 billion in 2025, and further rising to $1.447 trillion in 2027. This substantial upward revision is mainly driven by the rapid rise in memory prices rather than an overall expansion in the scale of infrastructure investment. Memory spending will jump from $71 billion in 2025 to $367 billion in 2026, and further expand to $923 billion in 2027; over the same period, non-memory AI capital expenditure will be approximately $631 billion in 2026, but will fall back to $525 billion in 2027.

Memory's share of AI capital expenditure will rise from about 14% in 2025 to 37% in 2026, and further surge to 64% in 2027. UBS calculates that about 60% of the year-on-year increase in AI capital expenditure in 2026 will come from rising memory costs, and in 2027 the increase in memory spending will even exceed the overall net increase in AI capital expenditure, because spending on other components will decline over the same period; of the nearly $1 trillion increase between 2025 and 2027, about 90% will come from rising memory spending.

UBS specifically points out that if incremental spending is mainly driven by price increases, the boost to U.S. real GDP will be quite limited, and will be more reflected in the transfer of income and profits to Asian memory producers, thereby making a more direct positive contribution to the GDP of the relevant economies. The world's major memory producers are concentrated in Asia, which means that as the AI investment cycle deepens, the Asian memory supply chain may benefit to a greater extent than previously expected by the market. This structural shift indicates that the economic driver of the AI investment cycle is shifting from pure infrastructure scale expansion to the rising cost of key components needed to support more powerful computing systems, and investors' framework for tracking AI capital expenditure will also need to be adjusted accordingly.