Big Tech's capex growth may be far slower than it looks

Recent projections from major technology firms have astonished investors, but a deeper analysis reveals that actual capital expenditure (capex) growth may be less impressive than it appears. According to a study released by RBC Capital Markets, giants like Amazon, Google, Meta, and Microsoft are set to invest nearly $600 billion this year on data centers, chips, and networking equipment to satisfy the increasing demands of artificial intelligence. At first glance, these figures suggest a rapid expansion in spending; however, RBC's findings indicate that much of this growth is being inflated by soaring memory prices. The analysts discovered that steep price increases in memory chips, particularly DRAM, high-bandwidth memory (HBM), and NAND flash, could account for approximately 45% of the projected increase in cloud capital expenditures by 2026. This surge is not primarily due to companies purchasing significantly more hardware. Rather, it reflects the rising costs of existing components. RBC estimates that spending on data center memory among the top ten hyperscalers will rise from around $107 billion in 2025 to about $237 billion in 2026. This staggering $130 billion increase is expected to represent a substantial portion of the overall capex growth within these companies. Even more telling is that about three-quarters of this memory-related increase—approximately $98 billion—stems from price hikes rather than higher quantities of units sold. The price of DRAM is predicted to more than double by 2026, while NAND flash prices are anticipated to rise by over 85%, according to TrendForce projections cited by RBC. Memory has become a crucial bottleneck in AI infrastructure, as advanced GPUs require significant amounts of high-performing DRAM and HBM, and AI data centers demand large quantities of flash storage. When RBC removes memory costs from the analysis, the picture changes dramatically. The anticipated capex growth drops to about 40% in 2026, a notable decline from the nearly 80% growth expected in 2025. While this still indicates healthy expansion, it is far less explosive than the initial totals suggest. RBC characterized this slowdown as "notable deceleration" but reassured that it is not a cause for immediate concern. The firm emphasized that fundamental investments in AI remain strong but cautioned that memory pricing is now a significant variable influencing capex trends as we approach 2027. In essence, Big Tech may be investing substantially more in certain types of equipment without a proportional increase in their operational capacity, as the AI race converges with a heated memory market.