CtrlS 4.4 GW 押注印度 AI 与数据中心热潮

印度对 AI 算力、云服务和数字基础设施不断增长的需求,正推动数据中心运营商扩充产能,并重新思考建址选择。
CtrlS Datacentres 的目标是到 2030 年实现 4.4 GW 的产能,投资布局既覆盖成熟的技术中心,也延伸至较小的城市。
该公司近期宣布在比哈尔邦投资 ₹10,000 crore,在菩提伽耶(Bodh Gaya)建设一座 150 MW 的 AI 工厂,同时以 ₹400 crore 在巴特那(Patna)建设数据中心,扩大其在该邦的布局。CtrlS 的巴特那数据中心目前拥有 12 MW 的 IT 产能,并具备随需求增长而扩容的空间。
在接受 AIM 独家专访时,CtrlS Datacentres 全球数据中心运营、企业托管交付、项目管理及能源研发总裁 Rahul Dhar 表示,其战略是让基础设施更贴近客户,尤其是那些需要为小城市用户提供服务的内容型企业。
Dhar 表示,这座新设施将支撑比哈尔邦及印度东部对数字基础设施的需求。
该公司还计划未来几年在二、三线城市建设 20 多座边缘数据中心,并配套更完善的光纤连接。目前,它在 10 座城市运营着 19 座数据中心,包括勒克瑙(Lucknow)、诺伊达(Noida)、加尔各答(Kolkata)和甘地讷格尔(Gandhinagar),并在布巴内斯瓦尔(Bhubaneswar)运营一座集装箱式数据中心。
Zerodha 联合创始人 Nikhil Kamath 与企业家 Sreeram Reddy Vanga 近期向 CtrlS Datacenters 合计投资 ₹250 crore。加拿大养老金计划投资委员会(Canadian Pension Plan Investment Board)今年早些时候也向该公司投资了约 ₹7,000 crore。
Dhar 将投资者对 CtrlS 的兴趣归因于公司专注于超大规模园区、对大容量电网电力的提前规划以及模块化设计。
超大规模云服务商是 CtrlS 最大的需求来源,不过该公司也为近 40 家印度公共和私营部门银行提供服务。初创企业和 neocloud 提供商是另一个不断增长的客户群体。
CtrlS 正在海得拉巴和孟买为超大规模云服务商建设绿地数据中心园区。它也在与寻求算力的 neocloud 公司洽谈,并预计在未来 12–24 个月内完成现有承诺的交付。
但它也面临来自大型科技公司(Big Tech)竞争的可能性。今年,中央政府为使用印度境内数据中心提供全球云服务的外国公司推出了 21 年免税期。
不过,Dhar 表示这不会影响 CtrlS 现有的需求,其需求由国内客户以及全球科技公司驱动。他说:“尽管有免税期,也可能会有外国投资者涌入,但印度托管服务商手头拥有良好的订单储备。”
Dhar 表示:“印度需要缩短数据中心电力需求的预测周期,并加快输电线路和光纤网络的审批,以跟上行业扩张的步伐。”
公用事业公司和政府部门通常按 7–8 年的周期来预测电力消耗。然而,AI 工作负载不断增长的电力需求以及密度越来越高的 GPU 机架,正使长期预测越来越难以维持。
Dhar 表示:“政府或各电力、公用事业部门通常要求我们提供的预测周期,可以缩短到三至四年。”
数据中心正从容量约 5–8 兆瓦的设施,发展为 10–100 兆瓦规模的园区。随着更新的芯片进入市场,运营商必须不断重新评估自身的电力需求以及现有设施的容量。
“我们是电力的大宗消费者,因此我们需要更好的规划,也需要各政府部门更清楚地了解我们需要多少电力、在什么时间需要以及在哪个地点需要,”他说。
公用事业公司需要提前掌握需求情况,以便规划输电线路、变电站以及通往数据中心站点的电力接入。Dhar 还呼吁加快路权审批,以便铺设光纤网络和输电线路,尤其是在运营商于城市郊区购置土地时。
CtrlS 正就印度近期提出的数据中心政策,向政府机构和行业协会提供意见,包括电子与信息技术部、NASSCOM 以及印度工业联合会。包括 Telangana、Uttar Pradesh、Karnataka、Tamil Nadu 和 Maharashtra 在内的多个邦已经出台了数据中心政策。
Dhar 表示,中央政府拟议的政策应解决基础设施规划和审批时限问题。
数据中心在水资源和电力消耗方面正面临越来越多的审视。根据 Mordor Intelligence 的估算,印度在 2025 年消耗了约 1500 亿升水,用于冷却约 1.5 GW 的数据中心容量。预计到 2030 年,这一数字将增长一倍以上。
CtrlS 采用闭环冷水机组来限制水损耗,并在其孟买的一处设施中回用处理后的废水,以减少对淡水的依赖。
该公司还在那格浦尔(Nagpur)拥有并运营一座装机容量近 100 MW 的自用太阳能电站。所发电力并入电网,用于抵消其孟买数据中心的用电量,并计划在海得拉巴等其他地点推出类似举措。
CtrlS 运营其设施的 PUE 约为 1.4。PUE 衡量的是数据中心总能耗与其 IT 设备能耗之比,数值越低通常意味着能效越高。
随着 GPU 部署规模扩大,运营商必须适应更高的功率密度和新的散热需求。Dhar 表示,CtrlS 已做好准备,为部署液冷基础设施的客户提供支持,其中包括 GPU 供应商和云公司。
“我们所有的数据中心都达到 tier-4 数据中心评级,”他说。这指的是数据中心的最高等级,即不存在单点故障、正常运行时间接近 100%。他补充说,公司采用外部认证,并已在海得拉巴启用一体化指挥与控制中心(Integrated Command and Control Centre),对基础设施进行集中监控。
该中心支持预测性分析和预防性维护。CtrlS 还将 AI 应用于基础设施监控,包括 CCTV 系统,以提升对运营问题的响应能力。
随着 AI 工作负载推高基础设施需求,Dhar 表示,数据中心运营商需要持续调整其设施,以适应不断变化的芯片技术、功率密度和散热系统。
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Inside CtrlS’s Massive 4.4 GW Bet on India’s AI and Data Centre Boom)
India’s growing demand for AI computing, cloud services, and digital infrastructure is pushing data centre operators to expand capacity and rethink where they build.
CtrlS Datacentres is targeting 4.4 GW of capacity by 2030, with investments spanning established technology hubs and smaller cities
It recently announced a ₹10,000 crore investment in Bihar to set up a 150 MW AI factory in Bodh Gaya, while expanding its presence in the state with a ₹400 crore data centre in Patna. CtrlS’s Patna data centre currently has 12 MW of IT capacity, with scope to scale up as demand increases.
In an exclusive interview with AIM, Rahul Dhar, President of Global Data Centre Operations, Enterprise Colo Delivery, Programme Management and Energy R&D at CtrlS Datacentres, says the strategy is to bring infrastructure closer to customers, particularly content-based businesses that need to serve users in smaller cities.
Dhar says the new facility would support demand for digital infrastructure in Bihar and eastern India.
The company also plans to establish more than 20 edge data centres across tier-2 and tier-3 cities over the coming years with improved fibre connectivity. It currently operates 19 data centres across 10 cities, including Lucknow, Noida, Kolkata, and Gandhinagar, and operates a containerised data centre in Bhubaneswar.
Nikhil Kamath, Co-founder of Zerodha, and entrepreneur Sreeram Reddy Vanga, recently invested a combined ₹250 crore in CtrlS Datacenters. Canadian Pension Plan Investment Board also invested around ₹7,000 crore in the company earlier this year.
Dhar attributes investor interest in CtrlS to its focus on hyperscale campuses, advance planning for high-capacity grid power, and modular designs.
Hyperscalers Drive Demand
Hyperscalers constitute the largest source of demand for CtrlS, although the company also serves nearly 40 Indian public and private sector banks. Startups and neocloud providers are another growing customer segment.
CtrlS is building greenfield data centre campuses for hyperscalers in Hyderabad and Mumbai. It is also in discussions with neocloud companies seeking capacity, and expects to deliver against its existing commitments over the next 12–24 months.
But it also faces the possibility of competition from Big Tech. This year, the central government introduced a 21-year tax holiday for foreign companies providing global cloud services using India-based data centres.
However, Dhar says it won’t affect CtrlS’s existing demand, which is driven by domestic customers as well as global technology companies. “Even though there is a tax holiday, and there might be an influx of foreign investors, the Indian colocation providers have a good order book in hand,” he says.
Power Availability Remains a Challenge
“India needs to shorten the period for forecasting data centre power requirements and speed up approvals for power transmission lines and fibre networks to keep pace with the industry’s expansion,” says Dhar.
Utilities and government departments typically seek power consumption on 7–8 year horizons. However, the growing power requirements of AI workloads and increasingly dense GPU racks are making long-term forecasts harder to maintain.
“The projection that the typical governments or the various power or utility departments are asking us can be reduced to three to four years,” Dhar says.
Data centres are growing from facilities with capacities of around 5–8 megawatts to campuses ranging from 10–100 MW. As newer chips enter the market, operators must continually reassess their power requirements and the capacity of existing facilities.
“We are bulk consumers of power, so we need better planning or better awareness among various government departments on how much power we need, at what time and at which location,” he says.
Utilities need advance visibility into demand to plan transmission lines, substations, and power connections to data centre sites. Dhar also calls for faster right-of-way approvals to lay fibre networks and transmission lines, particularly when operators acquire land on city outskirts.
CtrlS is providing inputs to government bodies and industry associations, including the Ministry of Electronics and Information Technology, NASSCOM, and the Confederation of Indian Industry, on India’s recently proposed data centre policy. Several states, including Telangana, Uttar Pradesh, Karnataka, Tamil Nadu, and Maharashtra, have already introduced data centre policies.
Dhar says the Centre’s proposed policy should address infrastructure planning and approval timelines.
Sustainability
Data centres face growing scrutiny over their water and electricity consumption. According to Mordor Intelligence estimates, India consumed about 150 billion litres of water in 2025 to cool about 1.5 GW of data centre capacity. This figure is expected to more than double by 2030.
CtrlS uses closed-loop chillers to limit water loss and is reusing treated wastewater at one of its Mumbai facilities to reduce dependence on fresh water.
The company also owns and operates a nearly 100 MW captive solar farm in Nagpur. The power is fed into the grid to offset consumption at its Mumbai data centres, with similar initiatives planned for other locations, including Hyderabad.
CtrlS also operates its facilities at a power usage effectiveness (PUE) of around 1.4. PUE measures a data centre’s total energy consumption against the energy used by its IT equipment, with lower values generally indicating better energy efficiency.
Changing Data Centre Design
As GPU deployments grow, operators must adapt to higher power densities and new cooling requirements. Dhar says CtrlS is prepared to support customers deploying liquid-cooled infrastructure, including GPU providers and cloud companies.
“All our data centres are rated tier-4 data centres,” he says, referring to the highest classification of data centres with zero single points of failure and close to 100% uptime. He adds that the company uses external certification and has introduced an Integrated Command and Control Centre in Hyderabad to monitor infrastructure centrally.
The centre supports predictive analysis and preventive maintenance. CtrlS is also applying AI to infrastructure monitoring, including CCTV systems, to improve responses to operational issues.
As AI workloads increase infrastructure requirements, Dhar says data centre operators will need to keep adapting their facilities to changing chip technologies, power densities, and cooling systems.
正文保留英文原文,标题与摘要由机器翻译。
原文发布于 Analytics India Magazine(Inside CtrlS’s Massive 4.4 GW Bet on India’s AI and Data Centre Boom),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。