Q2 2026 Cloud Infrastructure Spending Hits $143.4 Billion: The AI-Driven Boom, Hyperscaler Capex, and the Rise of Neoclouds

Unprecedented Hyperscaler Capex: Nearly $600 Billion

Amazon, Google, and Microsoft are collectively on track to spend nearly $595 billion on capital expenditures in 2026, driven overwhelmingly by AI infrastructure buildout :

• Amazon (AWS): Raised its 2026 cash capex forecast to approximately $220 billion, up from $200 billion. CEO Andy Jassy cited higher memory costs as a key reason for the increase .
• Google (Alphabet): Updated its 2026 capex range to up to $205 billion, citing an acceleration in capacity delivery to meet demand that continues to outpace supply .
• Microsoft: Approximately $175 billion in capex .

All three companies state they cannot build infrastructure fast enough to meet AI-driven demand .

AI-Specific Cloud Services: 165% Growth

AI-specific cloud workload revenue surged roughly 165% year-over-year, far outpacing overall cloud growth . Generative AI is the primary catalyst, accelerating demand for computing capacity across training, inference, and fine-tuning workloads . Synergy Research Group's chief analyst John Dinsdale noted that "GenAI-specific cloud services are growing at 165% year over year, but equally importantly, AI technology is enabling enhanced functionality and increased growth across a much broader range of cloud services" .

The Rise of Neocloud Providers

A new class of specialized AI infrastructure providers — "neoclouds" — has emerged to fill the gap between hyperscaler supply and enterprise demand:

• The neocloud market was estimated at $35.2 billion in early 2026 and is projected to reach nearly $240 billion within five years .
• Leaders like CoreWeave and Nebius are seeing Wall Street projections of growth rates exceeding 500% .
• Neoclouds offer GPU-as-a-service and model-agnostic AI compute, attracting enterprises that cannot secure enough capacity from hyperscalers .
• By Q4 2025, neocloud revenues were scaling at an unprecedented pace, reaching $9 billion for the quarter — up 223% year-over-year — and exceeding $25 billion for full-year 2025 .
• Some analysts caution that neoclouds may be temporary, arguing they are merely filling a supply chain crunch that could ease as hardware availability normalizes and traditional cloud providers evolve .

Persistent Supply Constraints

Despite GPU supply improving from acute shortage to functional balance — NVIDIA H100 cloud rental rates fell from approximately $8/hour in early 2023 to $1.80–$3.50/hour in Q2 2026 — significant bottlenecks remain :

• High-bandwidth memory (HBM) is now the binding constraint. Micron's HBM is sold out through 2026 under multi-year hyperscaler contracts, with no meaningful relief expected before 2028. HBM accounts for 30–40% of an accelerator's cost .
• GPU scarcity continues to push enterprises toward neocloud platforms and hyperscaler commitments rather than building their own infrastructure .
• Amazon specifically cited "the higher cost of memory" as a reason for hiking its 2026 capex forecast from $200 billion to $220 billion .
• TSMC's CoWoS (Chip on Wafer on Substrate) packaging capacity remains fully allocated, contributing to extended lead times .

Key Takeaway

Generative AI is the single dominant force driving all of these trends — record market size, triple-digit AI workload growth, unprecedented hyperscaler capex, a booming neocloud segment, and persistent hardware bottlenecks that keep enterprise buyers locked into cloud provisioning rather than on-premises deployment. The cloud infrastructure market has entered a new phase of accelerated investment and structural transformation, with no signs of slowing.