Semiconductor Capacity Constraints to Slow AI Spending Growth, Gartner Forecasts Show
Key Takeaways
- ▸HBM and DRAM memory production capacity is now the limiting factor for AI accelerator deployment and overall AI infrastructure growth
- ▸Gartner's revised forecasts show much more modest growth projections than previous quarters, with infrastructure bumps driven primarily by pricing rather than capacity expansion
- ▸AI spending has exploded to 31.7% of total IT budgets in 2025, up from 13.7% in 2024, effectively reshaping enterprise IT spending priorities away from traditional systems
Summary
According to new Gartner market research, constraints in DRAM and HBM (High Bandwidth Memory) production are emerging as the primary bottleneck limiting AI infrastructure expansion. As memory manufacturers run at full capacity with limited room for growth, the explosive upward revisions in AI spending forecasts that characterized 2024-2025 are expected to moderate significantly in 2026-2027, with any growth driven more by opportunistic pricing than capacity expansion.
Gartner's latest forecast reveals a notable slowdown in AI infrastructure growth projections. AI infrastructure forecasts were revised upward by only 4.8 points for 2026 and 8.1 points for 2027—a marked deceleration from the double-digit growth revisions seen in previous quarters. Software forecasts saw only tenths-of-a-point increases, while services forecasts were actually revised downward slightly, suggesting further caution in the market.
Despite the moderation, the absolute scale of AI spending remains staggering. Gartner data shows AI spending surged from 13.7% of total IT budgets in 2024 to 31.7% in 2025, effectively cannibalizing spending on traditional IT systems. This trend is expected to continue, with AI potentially accounting for the majority of IT spending within the next few years as companies deprioritize legacy infrastructure in favor of AI-ready systems.
The semiconductor constraint highlights a critical inflection point in the AI boom: while demand for AI infrastructure remains voracious, the physical supply chains that underpin it—particularly HBM memory production—are now the hard ceiling on growth. This shift from demand-driven euphoria to supply-constrained reality suggests a more measured but sustained expansion of AI infrastructure investment going forward.
- Supply-side constraints on semiconductors are expected to keep near-term AI spending forecast growth modest, signaling a shift from speculative exuberance to supply-constrained reality
Editorial Opinion
The emergence of semiconductor capacity as a hard constraint on AI infrastructure growth is a necessary reality check for an industry that has seen financing and spending forecasts revised upward almost monthly for the past two years. While HBM capacity bottlenecks will slow near-term forecast revisions, they actually validate the genuine, persistent demand for AI infrastructure rather than suggesting a bubble. Companies that secure HBM supply and maintain competitive AI infrastructure offerings will likely see their advantages widen as supply remains the limiting factor.
