Investing.com -- Citi raised its price targets on the three major semiconductor equipment makers, lifting Applied Materials to $710 from $550, Lam Research to $450 from $315 and KLA Corporation to $290 from $206.40, after updating its forecasts for global wafer fab equipment (WFE) spending.

The Wall Street firm said it aligned its estimates for the three stocks to a revised WFE model and rolled forward its price targets based on calendar year 2028 earnings power.

Citi now projects bull-case WFE spending of approximately $145 billion in 2026, $200 billion in 2027 and $250 billion in 2028, the first time the firm has introduced a 2028 forecast. The figures are based on an updated hyperscaler capital expenditure model that assumes growth of 84%, 56% and 38% in those respective years.

Citi said it has become "more constructive on 2028 WFE given continued capacity constraints and expansion at both TSMC and memory makers, as well as recent progress at Intel and Samsung foundries," with the 2028 figure implying 25% growth over the prior year.

At the same time, the rise of agentic AI is driving a structural increase in NAND demand, Citi said, as memory requirements surge while DRAM supply tightens.

Multi-step inference workflows are expanding the footprint of KV cache, the intermediate memory state used in AI models, "well beyond what high-cost HBM and DRAM can efficiently support," the bank wrote, particularly given constrained DRAM supply and elevated pricing.

As evidence of the strain, Citi cited a report that Nvidia had cut SoCAMM2 DRAM capacity in its Vera Rubin NVL72 systems by roughly 50%, due to supply limitations and cost considerations.

“This underscores a widening gap between required and available memory, and we are seeing companies accelerating the adoption of complementary solutions as a result, such as KV cache offloading where intermediate model state is shifted to lower-cost, higher-capacity storage tiers,” it said.

Citi estimated that closing the gap would require significant new NAND manufacturing capacity. Based on the assumption that a modern 150,000 wafers-per-month NAND fab can produce about 15 exabytes of capacity annually, it calculated that the industry needs two to four new greenfield fabs, representing $20 billion to $40 billion in capital expenditure, or $15 billion to $30 billion in NAND-specific wafer fab equipment spending, to overcome the bottleneck.