Big Tech’s latest capital expenditure (capex) projections have sparked investor excitement, but a closer look reveals a more nuanced picture. According to new research from RBC Capital Markets, the apparent surge in spending may be significantly overstated due to a sharp spike in memory prices. Amazon, Google, Meta, and Microsoft are collectively expected to spend nearly $600 billion this year on data centers, chips, networking equipment, and other infrastructure to support the growing demands of artificial intelligence. On the surface, this represents a massive acceleration in investment. However, RBC analysts argue that a large portion of this growth is driven not by increased hardware volume, but by soaring component prices—particularly for data center memory. The report highlights that memory chips, including DRAM, high-bandwidth memory (HBM), and NAND flash, could account for about 45% of the projected dollar growth in cloud capex by 2026. While demand for AI infrastructure continues to rise, much of the spending increase stems from companies paying far more for the same quantities of memory rather than buying more units. RBC estimates that memory spending across the top 10 hyperscalers will soar from around $107 billion in 2025 to nearly $237 billion in 2026—a $130 billion jump. Of that increase, about $98 billion, or roughly 75%, is attributed solely to higher prices, not higher unit volumes. The price surge is unprecedented. According to TrendForce data cited by RBC, DRAM prices are expected to more than double in 2026, while NAND flash prices could rise by over 85%. This inflation is driven by intense demand from AI-driven data centers, which rely heavily on high-performance memory to support advanced GPUs and massive storage needs. When memory costs are excluded from the analysis, the picture changes dramatically. RBC projects that capex growth for Big Tech would slow to about 40% in 2026, down from nearly 80% in 2025. While still substantial, this represents a notable deceleration in the pace of expansion. The analysts stress that this isn’t necessarily a sign of weakening AI investment. Instead, they frame the memory price spike as a major market distortion that’s inflating capex numbers without reflecting proportional growth in infrastructure buildout. They caution that memory pricing has now become the biggest uncertainty for future spending trends, particularly heading into 2027. In short, Big Tech may be spending far more on equipment—but not necessarily building much more. The AI boom is now colliding with a constrained memory market, turning price increases into a key driver of capex growth.