Server Shipments Surge 28%, Computing Power Leasing Market Approaches Trillion-Yuan Threshold

A latest report from market research firm TrendForce has ignited the tech community, with predictions of over 28% annual growth in AI server shipments now redefining the competitive landscape of the computing power market.

January 20, 2026 — A "blockbuster" piece of data hit the global tech scene: the latest AI server report from market research firm TrendForce indicates that global AI server shipments in 2026 will increase by over 28% year-on-year, while the proportion of ASIC-based AI servers will surge to 27.8%.

This signals that investment in AI infrastructure has entered an unprecedented period of acceleration. A report released on the same day by Deloitte, 2026 Predictions for the Technology, Media, and Telecommunications Industry, corroborates this trend: AI inference will account for two-thirds of computing power consumption in 2026, primarily concentrated in the nearly $500 billion market for new data centers and enterprise servers.

01 The Catalyst: TrendForce Report Reveals New Market Dynamics

TrendForce's report is like a boulder thrown into a calm lake, sending ripples throughout the computing power sector. The report clearly states that North American cloud service providers are continuously expanding their investments in AI infrastructure, which will directly drive global AI server shipments to achieve over 28% annual growth in 2026.

The explosion in AI servers is not an isolated event but part of a structural transformation in the entire server market. The report predicts that global server shipments (including AI servers) will grow by 12.8% year-on-year in 2026, representing an expansion in growth momentum compared to 2025.

Behind this data lies the deep penetration of AI application scenarios. From the primary focus on large language model training in 2024-2025 to the ongoing development of AI inference services such as AI agents, LLaMA model applications, and Copilot upgrades starting in the second half of 2025, cloud service providers are actively shifting toward inference services to achieve commercialization and monetization.

02 Technological Shift: The Rise of ASICs and Surging AI Inference Demand

The explosion in demand for AI computing power is reshaping the chip market landscape. A key statistic from the TrendForce report is: The shipment share of ASIC-based AI servers will increase to 27.8% in 2026, the highest level since 2023, with their shipment growth rate even surpassing that of GPU-based AI servers.

The driver behind this shift is clear. North American tech giants like Google and Meta are actively expanding their self-developed ASIC solutions, with Google's investment in its proprietary ASICs notably stronger than most cloud service providers, positioning it as a leader in the ASIC market.

The Deloitte report further elucidates the technical basis for this trend: AI inference will account for two-thirds of computing power consumption, primarily concentrated in the nearly $500 billion market for new data centers and enterprise servers. This prediction reveals the inevitable trend of AI applications transitioning en masse from the training phase to the inference phase.

Unlike the training phase, inference is more sensitive to energy efficiency and cost, which is the technical logic behind the rapid rise of specialized chips like ASICs. Billions of dollars' worth of dedicated inference chips are being deployed in these new data centers, with some chips' power consumption rivaling or exceeding that of general-purpose chips.

03 Capital Trends: Tech Giants Bet on AI Infrastructure

The explosion in demand for AI computing power is directly reflected in the capital expenditures of tech giants. According to TrendForce statistics, the combined annual capital expenditure growth rate of North America's top five cloud service providers—Google, AWS, Meta, Microsoft, and Oracle—will reach 40% in 2026.

These massive investments are not only for building large-scale infrastructure but also for phasing out general-purpose servers acquired during the cloud investment boom between 2019 and 2021. Google and Microsoft are expected to be the most active purchasers of general-purpose servers to handle the inference traffic generated by daily AI services like Copilot and Gemini that require physical provisioning.

The primary drivers for AI server shipments in 2026 will be North American cloud service providers, various government sovereign cloud projects, and the acceleration of proprietary ASIC development and edge AI inference solutions by large cloud service providers.

Analyzing the AI chips used, GPUs still hold the largest share at 69.7%, with models equipped with NVIDIA's GB300 becoming the mainstream for shipments, while the VR200 will gradually ramp up in volume in the latter half of the year.

04 Market Opportunity: Computing Power Leasing Enters a Golden Era

Facing the explosive growth in AI computing power demand, the computing power leasing market is encountering unprecedented development opportunities. Research from the China Research and Intelligence group predicts that by 2026, the potential revenue scale of domestic computing power leasing could reach 260 billion yuan, with subsequent high-speed growth exceeding 20% annually.

This prediction is based on a simple business logic: for most enterprises, building their own computing power infrastructure involves not only high initial investment but also the risk of rapid technological obsolescence. In contrast, computing power leasing offers a flexible, scalable, and cost-controllable solution.

The prosperity of the computing power leasing market is reflected not only in scale predictions but also in actual demand. As the pace of equipment renewal in the AI industry accelerates, the latest hardware is primarily leased to enterprises for R&D and design purposes. After 2-3 years, this equipment is then leased to other businesses for general office scenarios.

This hierarchical leasing model meets both high-end AI computing power demands and enables efficient utilization of hardware resources, forming a良性 (benign) computing power recycling ecosystem.

05 Ecosystem Evolution: From Hardware Competition to Service Innovation

The explosion in AI computing power demand is driving the evolution of the entire computing power ecosystem. Operators, leveraging their triple advantages in network, capital, and customer base, are evolving from traditional communication service providers into intelligent computing power ecosystem operators.

The core of this transformation lies in the shift of computing power services from单纯的 (mere) hardware provision to comprehensive solutions. Computing power leasing is no longer just about renting servers; it has become a complete package that includes value-added services like optimized configuration, energy efficiency management, and security protection.

Market trends indicate that consumers are increasingly focusing on flexibility and practicality in consumption, with the concept of "leasing instead of buying" gradually gaining acceptance. Data analysis-based personalized computing power configuration services are becoming new growth points in the leasing market.

As AI application scenarios continue to expand, demand for AI computing power will become more diverse and specialized. From training to inference, from cloud to edge, different application scenarios require different computing power configurations, providing differentiated competitive space for computing power leasing service providers.

Google's TPUs, in addition to being used internally for Google Cloud Platform infrastructure, have begun to be actively sold externally to companies like Anthropic. This move symbolizes the dawn of a new era: Computing power is transitioning from a proprietary competitive asset for enterprises into a marketable, tradable commodity.

As the prediction of 28% growth in AI server shipments in 2026 becomes a reality and the blueprint for a 260 billion yuan computing power leasing market unfolds, a profound transformation regarding the optimal allocation of computing resources is quietly taking place on a global scale.

From the capital races of tech giants to the flexible leasing options for small and medium enterprises, the democratization of computing resources is accelerating. In the future, AI computing power will, like electricity, become a foundational resource supporting the development of the digital economy, and computing power leasing will be the key mechanism ensuring the efficient flow of this resource.