
印度的可再生能源供应商正迎来数据中心需求激增的机遇,因为美国的反弹声浪日益增强以及中东冲突带来的风险,正推动运营商转向新市场。
India’s renewable energy suppliers are poised for a surge in demand from data centers, as increasing backlash in the US and risks from the Middle East conflict drive operators to new markets.
为支撑人工智能基础设施所需的绿色能源“被严重低估了”,印度领先的可再生能源供应商CleanMax Enviro Energy Solutions Ltd.的董事总经理兼创始人库尔德普·贾因表示。
The green energy needed to power the infrastructure serving artificial intelligence is “massively underappreciated,” said Kuldeep Jain, managing director and founder of CleanMax Enviro Energy Solutions Ltd., a top renewables provider in the country.
“即便仅有五吉瓦的数据中心容量进入印度——这是一个保守估计——也需要大约30至40吉瓦的可再生能源供应,”他在孟买接受采访时说。“这是需求超过供应的情况。”印度一直在为数据中心公司铺红毯,提供电价补贴、税收优惠以及更便利的土地审批。根据彭博新能源财经(BloombergNEF)的数据,这使得印度成为中国以外亚太地区数据中心开发的首选市场。
“Even if just five gigawatts of data center capacity comes to India, which is a conservative estimate, you will need about 30-40 gigawatts of renewable power,” he said in an interview in Mumbai. “It’s a case of demand exceeding supply.”India has been rolling out the red carpet for data center companies, offering power tariff subsidies, tax breaks, and easier land clearances. That’s made the country the top-ranked market for data center development in Asia-Pacific outside China, according to BloombergNEF.
贾因表示,美国公众对这些企业资源耗费巨大的设施产生反弹,以及他们面临的中东冲突扩大带来的风险,进一步提升了印度作为这些公司目的地的吸引力。7月,亚马逊的两个数据中心曾在巴林遭到伊朗袭击而受损。
The country’s appeal as a destination for these firms has been boosted by public backlash in the US against their resource-guzzling facilities and by the risks they face from the threat of widening conflict in the Middle East, Jain said. Two of Amazon’s data centers were damaged by an Iranian strike in Bahrain in July.

贾因称,支持如此快速的扩建将需要数十亿美元的可再生能源和储能投资。他补充说,CleanMax每年新增约1.5吉瓦的太阳能和风能容量,年资本支出最高可达9000亿卢比(约9.5亿美元)。
Supporting such a rapid buildout will need billions of dollars in renewables and energy storage spending, Jain said. CleanMax is adding about 1.5 gigawatts of solar and wind capacity per year, with annual capital expenditure of up to 9,000 crore rupees ($950 million), he added.
“如果要以75%的可再生能源为一吉瓦人工智能数据中心供电,需要大约三吉瓦的太阳能、三吉瓦的风能以及大约三吉瓦时的储能容量,”贾恩说。“仅在可再生能源方面,这需要约4万亿卢比(48亿美元)的资本支出。” 得到布鲁克菲尔德资产管理公司支持的这家公司已经与多家人工智能超大规模企业签署了供应协议——这些企业管理着全球数据中心网络——包括Meta Platforms公司、Alphabet公司旗下的谷歌、苹果公司和亚马逊公司。它将提供1.8吉瓦,占印度过去三年此类合同总量的大约三分之一。
“If you want to power a one gigawatt AI data cenere with 75% renewables, you need about three gigawatt of solar, three gigawatt of wind and about three gigawatt-hours of energy storage capacity,” Jain said. “This requires about Rs 40,000 crore ($4.8 billion) of capital expenditure just on the renewables side.”The Brookfield Asset Management-backed company has already signed supply deals with several AI hyperscalers — companies that manage global networks of data centers — including Meta Platforms Inc, Alphabet Inc.’s Google, Apple Inc., and Amazon.com Inc. It will provide 1.8 gigawatts, accounting for roughly a third of all such contracts in India in the last three years.
贾恩表示,人工智能超大规模企业越来越倾向于灵活的电力合同,在这些合同中,他们可以选择仅购买可再生能源项目的环境属性以抵消排放,而无需将电力实际送达其设施。
Jain said AI hyperscalers are increasingly favoring flexible power contracts, where they have the option of buying only the environmental attributes of a renewable project to offset emissions, without requiring physical delivery of electricity to their facilities.
“他们可以直接将电力送往数据中心,或者 alternatively 要求我们在交易所出售电力,并支付我们保证上网电价与交易所实际成交价之间的差价,”贾恩说。“我们看到大多数超大规模企业更倾向于保留这种灵活性。对卖家来说这无关紧要,我们获得的收益相同。” 根据彭博新能源财经(BNEF)的数据,目前印度已有1.3吉瓦的IT容量数据中心在建设中——也就是说,这些设施被设计用于向计算设备供电的功率。另有3.2吉瓦规模的项目已获得土地、电力和许可证。
“They can take power directly to data centers or alternatively ask us to sell power in the exchanges and pay us for the difference between the guaranteed tariff and the realization on the exchange,” Jain said. “We’ve seen most hyperscalers prefer to retain this flexibility. It doesn’t matter for the seller, we get the same realization.”Data centers with 1.3 gigawatts of IT capacity — that is, the power the facilities are designed to deliver to the computing equipment — are already under construction in India, according to BNEF. Another 3.2 gigawatts worth of projects have secured land, power and permits.
能源咨询公司伍德麦肯兹(Wood Mackenzie)预计,到2030年这一数字将达到12吉瓦,其中专用于人工智能的数据中心容量将激增近24倍,达到6.5吉瓦。
Energy consultancy Wood Mackenzie expects this number to reach 12 gigawatts by 2030, with AI-specific data center capacity jumping almost 24-fold to 6.5 gigawatts.