North American cloud giants are substantially increasing Capex, signaling an investment peak in AI infrastructure over the next two years. Shifting to large-scale, long-term strategies, they are focusing on high-end GPU Racks and accelerating in-house AI ASIC development to secure market leadership and drive rapid AI server growth.

Key Highlights

• Accelerated AI Investment: Major North American CSPs are significantly raising Capex, shifting to large-scale, long-cycle AI infrastructure strategies with strong momentum expected to continue for years.
• Focus on GPU Racks: Investment is heavily concentrated on high-end NVIDIA GPU rack solutions (GB/VR series), with NA providers taking a dominant share; AMD is also introducing competing solutions.
• Rise of Custom ASICs: To reduce reliance and improve cost efficiency, giants like Google, AWS, and Meta are accelerating in-house AI ASIC development, enhancing their technological autonomy.
• Strong Market Outlook: AI server shipment growth is projected to continue its upward revision, driven by the dual investment tracks (GPU + ASIC) and emerging sovereign AI projects.

Table of Contents

1. CSPs' Outlook on Capex Upward Revised Once Again for 2025; Investment Momentum Remains Strong for 2026
   • Average Capex Growth among Major Global CSPs in 2026
2. Five Major North American CSPs to Drive Growth of GB300 and VR200 GPU Rack Shipments in 2026
   • North America's Top Five CSPs Expected to Dominate Total GB and VR Rack Shipments by 2026, with Oracle Having Largest Share
3. AI Chip Self-Sufficiency Is a Rising Trend, and US-Based CSPs Are Expediting In-House ASIC Development and Strengthen Control over Key Technologies
   • CSPs Ramp Up AI ASIC Development to Achieve a Greater Degree of Computing Self-Sufficiency and Differentiation Competitiveness

<Total Pages: 6>