AI雲端工作負載：市場數據概覽（2026年第二季）

本報告調查了人工智慧雲端工作負載的趨勢，總結了訓練和推理工作負載消耗的趨勢和預測，並按類型、雲端服務供應商和地區等各種類別進行了詳細分析。

實際益處：

• 了解訓練和推理處理能力如何隨時間擴展，可以幫助您自信地制定能力策略。
• 透過清楚了解工作負載成長模式和轉折點，您可以最佳化您的 AI 投資。
• 預計區域趨勢將對基礎設施採購產生長期影響。

主要問題解答：

• 在雲端運算中，推理何時會超越訓練，成為主要的資源來源？
• 在雲端，哪種類型的推理工作負載會消耗最多的容量？
• 不同地區的推理和訓練過渡點有何不同？

研究亮點：

• 對供應商細分市場進行全面分析：一級和二級超超大規模資料中心業者、新型雲端服務供應商和主權雲端服務供應商。
• 對成長軌跡和轉折點進行詳細分析（包括推理工作負載優於訓練工作負載的時期）
• 到 2035 年雲端 AI 推理工作負載的詳細預測

目錄

本產品旨在與 AI 雲端工作負載 (MD-AICW-101) 配合閱讀。

主要發現

什麼是新的

重要預測

訓練與推理

訓練負荷

推理工作負載

調查方法

Actionable Benefits:

• Plan capacity strategies with confidence by understanding how training and inference capacity will grow over time.
• Align Artificial Intelligence (AI) investments with clear visibility into workload growth patterns and inflection points.
• Anticipate regional dynamics that will shape infrastructure procurement over time.

Critical Questions Answered:

• When will inference surpass training as the dominant capacity consumer in the cloud?
• Which types of inference workloads will be consuming the most capacity in the cloud?
• How will the inference-training inflection point differ by region?

Research Highlights:

• Comprehensive analysis of operator segments: Tier One and Tier Two hyperscalers, neocloud providers, and sovereign clouds.
• Detailed research into growth trajectories and inflection points-including when inference workloads will overtake training workloads.
• A detailed forecast of AI inference workloads in the cloud up to 2035.

Who Should Read This?

• Cloud and data center strategy leaders looking to optimize their procurement and investment roadmaps.
• Infrastructure developers and ecosystem partners aiming to anticipate regional trends.
• AI compute leaders looking to refine their roadmaps and strategies.

TABLE OF CONTENTS

This product is meant to be read in conjunction with AI Cloud Workloads MD-AICW-101

Key Findings

Whats New

Significant Forecasts

Training Versus Inference

Training Workloads

Inference Workloads

Methodology