It’s Only Common Sense: Customers Need Confidence More Than Low Prices
Dan’s Biz Bookshelf: Revisiting 'The Popcorn Report'
Surging investment in AI infrastructure is expected to drive the combined CapEx of the world’s nine largest CSPs up by approximately 90% YoY in 2026, according to TrendForce’s latest AI server research.
Most recently, procurement demand for NVIDIA GB/VR rack-scale AI server platforms has strengthened noticeably among hyperscale CSPs and Tier-2 data center operators. Meanwhile, Google and AWS are set to ramp production of their next-gen in-house ASIC platforms during the second half of 2026, while Chinese CSPs are accelerating the deployment of domestic AI solutions to support LLM services.
TrendForce has therefore revised its 2026 AI server shipment forecast upward from 28% to nearly 31% YoY.
TrendForce estimates that the combined 2026 CapEx of Google, Amazon, Meta, Microsoft, Oracle, ByteDance, Tencent, Alibaba, and Baidu will exceed US$886.7 billion, with the five North American hyperscalers accounting for nearly 90%. This increased spending reflects continued investment in AI data centers, GPU clusters, liquid-cooling infrastructure, and other next-generation AI infrastructure required to support the rapidly growing demand for generative AI and large-scale models.
Breaking down AI chip strategies among North American CSPs, TrendForce observes that Google remains focused on expanding the deployment of its in-house TPUs throughout 2026 and 2027, with shipments continuing to grow rapidly in 2027. AWS is expected to deploy NVIDIA GB300 as its primary GPU AI server platform in 2026 while steadily expanding shipments of its internally developed ASICs, with further volume growth anticipated in 2027. Meta will rely primarily on NVIDIA GB/VR and AMD Helios rack-scale systems in 2026, while significantly accelerating deployment of its proprietary AI ASICs in 2027 as part of its long-term strategy to reduce AI infrastructure costs and improve inference efficiency.
CapEx of China’s four largest CSPs expected to grow by more than 80% in 2026
Chinese CSPs are also entering a new investment cycle of AI infrastructure. TrendForce forecasts that the combined CapEx of ByteDance, Tencent, Alibaba, and Baidu will increase by more than 80% YoY in 2026. Among them, ByteDance is expected to post the largest increase, with investment focused primarily on large-scale AI data centers, proprietary ASIC development, and the deployment of GPU clusters.
Looking to 2027, TrendForce estimates that the combined CapEx of the top nine CSPs will hit about $1.3 trillion, showing nearly 50% year-over-year growth. While the growth rate may slow due to a higher comparison baseline, this does not suggest a decline in AI investments.
Ongoing developments in AI inference, Agentic AI, custom ASICs, and next-generation AI models will keep fueling strong demand for computing infrastructure. As a result, CSP investments are likely to reach new record levels, expanding beyond GPU deployment to include upgrades across AI servers, liquid cooling, advanced packaging, high-speed interconnects, power infrastructure, and memory.
