AI Infrastructure Boom
EXECUTIVE SUMMARY
The AI infrastructure market is experiencing explosive growth, valued at approximately $72-102 billion in 2025-2026 and projected to surpass $200 billion by 2030-2031, driven by surging demand for compute, storage, and networking to support generative AI and machine learning workloads. Global spending on AI infrastructure is on track to reach $334 billion in 2025, escalating toward $900 billion by 2029 and nearly $3 trillion cumulatively by 2028, fueled by hyperscaler capex and industrial-scale data center buildouts. This boom presents compelling opportunities in hardware and adjacent sectors, though sustainability hinges on energy efficiency and supply chain resilience.
MARKET CONTEXT
The AI infrastructure landscape in 2026 reflects a fragmented yet rapidly consolidating market, with hardware dominating 68% of spending on GPUs, high-bandwidth memory, and NVMe storage systems pushing rack densities beyond 100kW. Market size estimates vary slightly across analysts: Technavio forecasts growth of $39.5 billion from 2026-2030 at a 24.7% CAGR, starting from a 2025 base implied around $160 billion; The Business Research Company pegs 2025 at $71.9 billion, expanding to $90.9 billion in 2026 (26.5% YoY) and $227 billion by 2030 (25.7% CAGR); Mordor Intelligence values it at $101.2 billion in 2026, reaching $202.5 billion by 2031 (14.9% CAGR). IDC tracks global AI infrastructure spending at $334 billion for 2025—far exceeding traditional cloud investments—with hyperscalers capturing the majority, on pace for $900 billion by 2029. North America leads with over 44% of incremental growth, bolstered by tech ecosystems and investments, while Asia-Pacific emerges as the fastest-growing region at 16.4% CAGR through 2031. On-premise deployments hold 57% share in 2025 due to data sovereignty needs, but cloud is accelerating at 15.8% CAGR amid pay-per-inference models.
KEY DEVELOPMENTS
- IDC reports AI infrastructure spending hitting $334 billion in 2025 (up from $176 billion per broader trackers), with hyperscalers driving 80%+ of capex toward GPU clusters and edge solutions; fastest segments include AI networking (InfiniBand/Ethernet) and liquid cooling at >30% CAGR.
- Morgan Stanley projects $2.9 trillion in global data center construction costs through 2028—akin to industrial buildouts—contributing ~25% to U.S. GDP growth in 2026 via power, industrial output, and services demand.
- Enterprise AI adoption surges to 72%, with generative AI at $136 billion within a $538 billion total AI market (37.3% CAGR); venture funding hit $98 billion in 2025, targeting inference optimization and vertical apps amid talent premiums of 42% for AI engineers.
INVESTMENT IMPLICATIONS
Portfolio positioning favors overweight exposure to AI infrastructure leaders in silicon (NVIDIA, AMD), networking (Broadcom, Arista), and cooling/power enablers, as GPU demand for LLMs drives 25%+ CAGRs through 2030. Hyperscaler cloud providers (AWS, Azure) offer OpEx scalability, with Trainium2/TPU v6e instances enabling multi-petaflop economics; allocate 15-20% to diversified ETFs tracking data center REITs and semis for $3 trillion buildout upside. Edge AI and hybrid models present mid-cap growth plays, balancing on-premise (57% share) with cloud's 15.8% expansion; target 10-15% CAGR returns via active strategies emphasizing energy-efficient infra amid CHIPS subsidies boosting fabs (+2.5% CAGR impact).
RISKS TO MONITOR
- Supply constraints in GPUs and high-bandwidth memory, with hyperscaler concentration risking 80% capex dependency and potential idle time from networking bottlenecks.
- Energy demands exceeding supply, as 100kW+ racks strain grids; rising focus on efficiency may cap growth if cooling/power innovations lag >30% segment CAGRs.
OUR VIEW
While consensus fixates on NVIDIA-dominated silicon, the contrarian edge lies in overlooked networking and storage bottlenecks, where InfiniBand/RoCE and NVMe systems grow at 30%+ CAGRs to eliminate GPU idle time in distributed training—offering 2-3x upside vs. headline compute plays as edge AI fragments the stack.
This research note is for informational purposes only and does not constitute investment advice. Solomon Grey Capital and its analysts may hold positions in the securities and asset classes discussed.
