↓Skip to main content
  1. Posts/

【Passport to AI Era vol.1】Overview AI Market

·5 mins
AI Data Center 中文 Passport to AI Era

In the AI era, following my previous deep dive into the truly critical strategic moats, this series of reports will guide you through a systematic study of Data Center (DC) analytics. As the inaugural piece of this series, I will take a macro perspective to examine the global DC market landscape, analyzing it primarily through the fundamental framework of supply and demand.

In an age of information overload where AI-generated content is ubiquitous, true human value lies in synthesizing massive amounts of data to extract valuable insights. This report begins by examining the demand side, moves to the supply side, and concludes with key insights and future research directions, aiming to provide a clear and valuable reference.


在 AI 時代，繼先前文章一起深入探討真正重要的護城河之後，本系列報告將陪伴大家系統性地研究資料中心（Data Center，以下簡稱 DC）的各項分析。這篇作為系列的開頭，我們用宏觀的角度來看看全球 DC 市場的全貌，主要從供給與需求的基本框架出發進行分析。

AI 時代資訊爆炸、內容到處都是，而人的真正價值就在於把這些海量訊息彙整起來，並整理出真正有用的洞見。本報告先從需求端開始看起，接著檢視供給端，最後分享一些關鍵洞見與後續的研究方向，希望能為大家帶來清晰的參考。

Demand-Side Analysis / 需求端分析 #

As the market adage goes, “Supply always comes on the heels of demand.” Currently, for Transformer-based generative AI, the fundamental unit of computation is the token. Consequently, token usage volume serves as a highly representative proxy for current AI demand intensity. While it may not capture 100% of the real-world picture, OpenRouter’s token usage analysis offers significant reference value.

正如市場上大家常說的「Supply always comes on the heels of demand」，供給永遠是跟在需求後面來的。目前以 Transformer 為主的生成式 AI，最小的運算單位就是 token，因此 token 的使用量可以很代表性地反映出當前 AI 的需求強度。雖然無法 100% 涵蓋所有實際情況，但 OpenRouter 的 token 使用分析已經很有參考價值。

According to OpenRouter’s research [1], global token usage has exhibited exponential growth from 2024 through the end of 2025. This clearly indicates that worldwide demand for AI is currently in a phase of explosive hyper-growth.

根據 OpenRouter 的研究[1]，從 2024 年到 2025 年底，全球 token 使用量呈現指數式成長，這清楚顯示全世界對 AI 的需求正處於超額爆發的階段。

The same report highlights distinct regional variations: the Asia-Pacific (APAC) region’s share of global token usage continues to climb, demonstrating that AI demand growth in APAC is significantly outpacing the global average.

同一份報告也讓我們看到明顯的區域差異：亞太地區（APAC）在全球 token 使用佔比中持續上升，顯示亞太區的 AI 需求成長速度明顯高於全球平均。

Furthermore, Stanford University’s AI Index Report [2] points out that APAC’s AI adoption rate in 2024 remained relatively low compared to North America (72% vs. 82%). This suggests substantial headroom for future growth in the APAC region.

此外，史丹佛大學的分析報告[2]指出，2024 年亞太區的 AI 使用佔比相較北美仍然偏低（72% 對比 82%），這也意味著亞太區未來還有很大的成長空間。

It is particularly noteworthy that due to the unique situation of the Chinese market, reports from OpenRouter and Stanford often underestimate China’s actual contributions. This underlying factor implies that APAC’s overall growth momentum is even stronger than it appears on paper.

特別值得注意的是，由於中國市場的特殊性，OpenRouter 和史丹佛的報告往往會低估中國的實際貢獻，這反而讓亞太區的整體成長動能看起來更強勁。


Key Demand-Side Insights / 需求端關鍵洞察 #

  • Global AI demand is experiencing explosive, exponential growth.

  • 全球 AI 需求呈現爆炸式的指數型增長。

  • Since 2024, APAC’s share of AI demand has steadily increased, solidifying its position as the world’s second-largest region for AI demand.

  • 從 2024 年開始，亞太區的 AI 需求佔比逐步提升，已成為全球第二大 AI 需求區域。

  • The APAC region retains massive potential for future growth in AI demand.

  • 亞太區對 AI 的需求在未來仍有相當大的成長潛力。

Supply-Side Analysis / 供給端分析 #

According to CBRE’s annual trend report [3], markets worldwide have aggressively expanded their DC construction scale over the past two years, with 2025 supply volumes substantially exceeding 2023 levels. North America boasts the highest supply growth rate, which aligns perfectly with our demand-side analysis and confirms it as the region with the most robust AI demand over the last two years.

根據 CBRE 的年度分析報告[3]，全球各地在過去兩年都積極擴大 DC 的建設規模，2025 年的 DC 供給量相較 2023 年都有大幅增加。其中北美地區的供給增速最高，這也跟前面需求端的分析互相呼應，確實印證北美是近兩年 AI 需求最旺盛的區域。

Meanwhile, supply volumes and growth rates in Europe and the APAC region remain roughly on par with each other.

歐洲與亞太區的供給量和增速則大致相當。

Looking at DC vacancy rates—where lower rates indicate supply struggling to keep pace with demand—North America remains the most intensely constrained market, followed by Europe, and then APAC.

如果從 DC 空置率來看（空置率越低，就代表供給越追不上需求），北美目前依然是需求最強烈的地區，其次是歐洲，再來是亞太區。

Interestingly, within the APAC region, Sydney’s overall vacancy rate dropped significantly in 2025 compared to 2023, with a steeper decline than other major hubs like Singapore and Tokyo. This reveals exceptionally robust demand for DCs in Sydney over the past two years, with supply evidently lagging behind.

有趣的是，在亞太區裡，悉尼的整體空置率在 2025 年相較 2023 年明顯下降，而且降幅高於新加坡、東京等其他主要市場，這顯示悉尼近兩年對 DC 的需求非常旺盛，供給明顯跟不上腳步。

In terms of rental costs, highly populated, land-constrained metropolitan areas (such as Tokyo, Hong Kong, and Singapore) command significantly higher monthly pricing than other regions. Conversely, cost levels across North America, Europe, and Sydney show no substantial disparities.

租金成本方面，土地資源有限、人口密集的大型城市（如東京、香港、新加坡）的月租金相對其他地區明顯偏高；而北美、歐洲以及悉尼的租金水準則沒有太大的差距。


Key Supply-Side Insights / 供給端關鍵洞見 #

  • Despite aggressive efforts on the supply side to keep up with explosive global AI demand, the market remains fundamentally undersupplied, driving a continuous decline in DC vacancy rates.

  • 雖然全球 AI 需求呈爆炸式增長，供給端已經很努力跟進，但整體來說還是供不應求，導致 DC 的空置率持續下降。

  • North America maintains its position as the largest region for global supply, closely followed by Europe and APAC.

  • 北美依然是全球供給規模最大的地區，歐洲與亞太區緊隨其後。

  • Based on monthly rental rates (representing supply costs), densely populated cities with limited footprints (e.g., Singapore, Tokyo, Hong Kong) incur higher DC supply costs due to exorbitant land premiums.

  • 以月租金（也就是供給成本）來看，地小人稠的大型城市（如新加坡、東京、香港）因為土地成本高，DC 的供給成本也相對較高。

Note / 注意事項: #

This current analysis has yet to incorporate other critical markets such as China, Malaysia, and Thailand. These regions exhibit equally robust supply pipelines paired with highly advantageous cost structures, making them essential considerations for future analyses.
這份分析目前尚未納入中國、馬來西亞及泰國等其他重要市場。這些地區同樣供給旺盛，而且成本結構相對具有優勢，未來在分析時絕對不能忽略。

Key Insights and Future Research Directions #

Synthesizing both demand and supply-side analyses, it is evident that the APAC region will be the most powerful growth engine for future AI supply and demand. Consequently, subsequent reports in this series will focus squarely on the APAC region (with particular emphasis on Singapore and Sydney). I will conduct deep-dive explorations into critical topics including power supply, cooling systems, clean energy, policy environments, site selection strategies, and supply chain.


關鍵洞見與未來研究方向 #

綜合需求與供給端的分析，我們可以清楚看到，亞太區域將會是未來 AI 需求與供給最強勁的成長引擎。因此，本系列後續的報告會把焦點放在亞太區域（特別是新加坡與悉尼），進行更深入的專題探討，涵蓋電力供應、冷卻系統、潔淨能源、政策環境、選址策略、供應鏈等重要主題。

References / 參考資料 #

[1] OpenRouter, State of AI 2025: 100T Token LLM Usage Study https://openrouter.ai/state-of-ai

[2] Stanford University, AI Index Report 2025 https://hai.stanford.edu/assets/files/hai_ai_index_report_2025.pdf

[3] CBRE, Global Data Center Trends 2025 https://www.cbre.com/insights/reports/global-data-center-trends-2025

© Chung-Hao Lee. All Rights Reserved. All content on this webpage—including but not limited to text, images, design, code, and multimedia materials—is protected under the international copyright treaties. Unauthorized reproduction, modification, distribution, public transmission, or commercial use is strictly prohibited. Legal action will be taken against infringement.
© 李崇豪。保留所有權利。 本網頁之內容（包括但不限於文字、圖片、設計、程式碼及多媒體素材）均受國際著作權條約保護。未經書面授權，嚴禁任何形式之複製、改作、散布、公開傳輸或商業利用。侵權者將依法追訴。