美國科技巨擘投資數十億美元，數據中心建設熱潮席捲印度（2/2）

Recently the New York Times reported the following:

Data Center Surge Reaches India as American Tech Giants Invest Billions (2/2)

Megacities in southern India are attracting enormous investments to help build artificial intelligence infrastructure to serve the world’s most data-hungry country.

By Alex Travelli and Pragati K.B. - Reporting from New Delhi

Dec. 26, 2025

(continue)

India’s economic strengths and weaknesses look like a mismatch for the A.I. investments. The data centers will require cheap land, abundant electricity and water. India’s whole territory is packed with farmland, and much of it suffers from unreliable power grids and scarce water supply. It does need jobs for a huge and semi-idle labor force, but data centers require minimal staffing.

But India doesn’t want to miss out on the A.I. gold rush. It also doesn’t want to rely on data servers overseas.

Since 2018, India has been weighing laws that would require digital functions to be based on servers in India. For national security’s sake, the government already insists that banks and WhatsApp, which Meta owns, keep their data local. Mr. Mukherjee said the possibility of officials’ widening those demands was motivating foreign companies to install servers within India.

Another reason to keep the data centers closer to the people who will be using them is that data slows the farther it has to travel. Delays of even milliseconds between faraway servers can pile up during complex processing, causing whole systems to lag. Until a few years ago, most of India’s data was stored on servers in Singapore, 1,800 miles away across the Bay of Bengal.

The data centers for Google, Microsoft, Amazon and Meta will join a growing number of smaller centers spread between India’s biggest urban areas. Most are along the coasts, where they can be connected to undersea cables.

Luring giant plants to Hyderabad, an inland city of 11 million, required a feat of salesmanship on the part of the local government. That job fell to Jayesh Ranjan, a special secretary in charge of tech and investment for the state who devised India’s first data-center policy in 2016.

Hyderabad’s first data centers, which were built before A.I. started demanding more capacity, are already saturated. Now there are four.

Industrial-grade electricity is hard to come by in most of India. But Hyderabad’s data parks are connected to multiple energy sources and flooded with abundant power at wholesale prices. India’s electricity grid is balky, but in the aggregate it now generates more power than it needs and most of it is renewable. American electricity costs an average of 18 cents a kilowatt-hour, but Hyderabad’s data centers pay only 7 cents.

Mr. Ranjan said the country’s first data center to run on a gigawatt’s worth of power would be in Hyderabad. TPG, an American asset manager, is working with India’s Tata Group to build it and other A.I.-focused facilities around the country.

Another challenge for building data centers in India is access to water. In 2019, water from two rivers was redirected by a gigantic dam to irrigate most of the state around Hyderabad. And there is still plenty left for the thirsty new data centers, which consume tons of water to cool their servers.

Translation

美國科技巨擘投資數十億美元，數據中心建設熱潮席捲印度（2/2）

印度南部的大城市正吸引巨額投資，用於建立人工智能基礎設施，以滿足這個全球數據需求量最大的國家的需求。

（繼續）

印度的經濟優勢和劣勢似乎與人工智能投資並不匹配。資料數據中心需要廉價的土地、充足的電力和水資源。印度的國土遍布農田，其中許多地方電力供應不穩定，水資源匱乏。印度確實需要為龐大且半閒置的勞動力提供就業機會，但數據中心需要很少工作人員。

然而，印度不想錯過這場人工智能淘金熱。它也不想依賴海外的數據伺服器。

自2018年以來，印度一直在考慮立法，要求所有數位功能都必須基於印度境內的伺服器。出於國家安全考慮，政府已經要求銀行和Meta旗下的WhatsApp等公司將資料保留在本地。Mukherjee表示，官員可能擴大這些要求，以促使外國公司在印度境內部署伺服器。

將數據中心靠近使用者的另一個原因是，數據傳輸距離越遠，速度越慢。即使是遠距離伺服器之間幾毫秒的延遲，在複雜的處理過程中也會累積，導致整個系統運作緩慢。直至幾年前，印度的大部分資料都儲存在1800英里（約2900公里）外的新加坡的伺服器上，隔著孟加拉灣。

Google、微軟、亞馬遜和Meta的數據中心將加入印度各大城市地區日益增長的小型數據中心行列。這些數據中心大多位於沿海地區，可以連接到海底光纜。

要吸引大型中心落腳Hyderabad這座擁有1,100萬人口的內陸城市，當地政府需要展現出非凡的遊說才能。這項任務落在了Jayesh Ranjan身上，他是該邦負責科技和投資的特別秘書，也是印度首個數據中心政策的製定者，該政策於2016年出台。

Hyderabad的首批數據中心建於人工智能開始對容量需求激增之前，如今已基本飽和。現在，該市共有四個數據中心。

在印度大部分地區，工業級電力供應緊張。但Hyderabad的數據園區接入了多種能源，並以批發價格獲得充足的電力供應。印度的電網雖然存在一些問題，但整體而言，其發電量已超過自身需求，而且大部分電力來自再生能源。美國的平均電價為每千瓦時18美仙，而Hyderabad的數據中心只需支付7美仙。

Ranjan表示，印度首個耗電量達千兆瓦的數據中心將落腳Hyderabad。美國資產管理公司TPG正與印度塔塔集團合作，在印度各地建置此數據中心及其他人工智能相關設施。

在印度建造資據中心面臨的另一個主要挑戰是水資源短缺。 2019年，一座巨型水壩截流了兩條河流的水流，用於灌溉Hyderabad週邊的大部分地區。即便如此，仍有充足的水源可供這些耗水量龐大的新建數據中心使用，它們需要消耗大量水來冷卻伺服器。

So, new investment to build data centers, cloud computing and other hardware has come to India. India’s economic strengths and weaknesses look like a mismatch for the A.I. investments, yet India doesn’t want to miss out on the A.I. gold rush. It also doesn’t want to rely on data servers overseas. India now generates more electricity than it needs and most of it is renewable. In 2019, water from two rivers was redirected by a gigantic dam to irrigate most of the state around Hyderabad, and there is still plenty left to meet new demands. Apparently, Hyderabad is a good location to build huge data centers.