2025 Data Center Power Report
在数据中心发展的早期阶段，全行业的焦虑点几乎都聚焦在“算力供给”上——我们讨论摩尔定律的放缓、GPU的产能分配以及芯片的制程工艺。然而，当AI浪潮真正席卷而来时，我们才猛然发现，限制算力天花板的不再是硅片，而是最基础的电力。
在这份BloomEnergy的《2025年数据中心电力报告》中讲述了市场“缺电”的现状，也预示了数据中心基础设施架构即将发生根本性巨变。
1.“电荒”只是前奏，行业正跨入“吉瓦（GW）时代”: 未来五年 35 GW 的天量需求，宣告我们告别了“兆瓦级”单体机房时代。面对这种相当于中型城市的能耗巨兽，现有电网无论在容量还是建设周期上，都已捉襟见肘。
2.“孤岛运行”正从应急手段变为战略常态: 面对漫长的电网扩容期，“等待”即意味着出局。数据中心正加速转向“能源独立”，现场发电（如燃料电池）将不再是备用的“N+1”冗余，而是确保算力按时上线的“主供电源”。
3.“热-电-算”的深度耦合将重塑基础设施: 当AI负载的剧烈波动遇上高密度的现场发电，挑战已升级为发电与散热的动态平衡。未来的数据中心即是能源工厂，谁能解决这一矛盾并掌握“快速通电”的能力，谁就能扼住算力竞赛的咽喉。
第四次工业革命正蓬勃发展，其动力源于人工智能和云计算技术的突飞猛进。在美国，快速部署新的数据中心容量已成为一项战略重点，但目前面临一个主要瓶颈：电力供应。随着电力需求持续增长，电网设施却日益老化，而新的电网项目又面临着行政审批和供应链方面的重重挑战。
The fourth industrial revolution is underway, fueled by rapid advancements in AI and cloudcomputing. In the US, the rapid deployment of new data center capacity is a strategic priority, but there is a major bottleneck: power availability. Demand for power is only growing, while the electricity grid is aging and new grid projects face permitting and supply chain challenges.
尽管这些挑战已引起广泛关注，但针对潜在解决方案的讨论却相对较少。作为电力解决方案领域的领军企业，Bloom Energy 在这份《2025年数据中心电力报告》中阐述了数据中心领导者们正如何实现转变思维模式，并通过采纳创新方案来实现其战略目标及应对经济层面的紧迫需求。
While these challenges have received significant attention, there has been comparatively less discussion about potential solutions. Bloom Energy, a leader in power solutions, explains in this 2025 Data Center Power Report how data center leaders are shifting paradigms and adopting innovative solutions to meet their strategic goals and economic imperatives.
在2024年4月和11月，我们就电力架构决策问题，调研了直接参与其中的数据中心行业领袖。受访者广泛来自于超大规模运营商和主机托管开发商，职位涵盖从基层管理者到高管层的层级。为了补充这些定量调查的发现，我们还委托专人对数据中心电力架构的决策者进行了深度的定性访谈，并参考了多方外部数据源，包括劳伦斯伯克利国家实验室、麦肯锡以及高盛的研究成果。
In April and November 2024, we surveyed datacenter leaders directly involved in making power architecture decisions. Survey respondents represented a variety ofhyperscaler and colocation developer companies and ranged from managers to C-suite executives.1 To supplement findings from these quantitative surveys, we also commissioned in-depth qualitative interviews with decision makers for data center power architecture and referenced multiple external data sources, including research by the Lawrence Berkeley National Laboratory, McKinsey & Company, and Goldman Sachs.
调查结果摘要
Summary of findings
数据中心领导者预计电力供应将进一步恶化：美国的电力需求正以史无前例的速度增长，而数据中心正是这一增长的最大推手。美国电网已无力跟上这一需求步伐，新建数据中心项目将难以获得及时的电力接入。
Data center leaders expect power availability to get worse：Demand for power in the US is growing at an unprecedented rate, with data centers being the largest driver of this growth. The US grid has not been able to keep pace with this demand, and new data center projects will struggle to get timely access to power.
Leaders are increasingly turning to onsite power as a solution：Data center leaders expect approximately 30% of all data center sites to use some onsite power as a primary energy source supplemental to the grid by 2030, 2.3 times more than just seven months prior. We find that new data center announcements corroborate this expectation.
在权衡各项优先级时，领导者对“价值”的看法发生了变化 ：新建数据中心正在平衡更多的优先事项，而“通电交付周期”在价值评估中扮演着日益重要的角色。通过对数据中心领导者的调研和访谈，我们总结出了选择现场发电技术的七个关键标准。
Leaders see value differently as they balance priorities：New data centers are balancing more priorities, and time to power is playing an increasingly important role in the value equation. Our surveys and interviews with data center leaders have surfaced seven key criteria for choosing onsite power technology.
展望未来，领导者预计2025年将出现模式转变：基于我们从数据中心领导者那里获得的反馈，我们确定了预计将在2025年持续的三大关键趋势。
Looking ahead, leaders expect a paradigm shift in 2025：Based on what we have heard from data center leaders, we have identified three key trends we expect to continue in 2025.
美国已宣布即将新增 20 GW 的数据中心 IT 容量
20 GW of upcoming data center IT capacity has been announced in the US
美国数据中心 IT 容量（单位：GW）
GW of cumulative data center IT capacity in the US
美国累计数据中心 IT 容量（单位：GW）
数据中心领导者预计电力供应将进一步恶化
‘Data center leaders expect power availability to get worse
美国电力需求正以史无前例的速度增长经历了20年的需求持平后，预计2025年美国电力需求将增加 83太瓦时（TWh）—— 这相当于为额外 770万 户家庭供电
Demand for power in the US is growing at an unprecedented rate
After 20 years of flat demand,US power needs are predicted to rise by 83terawatt-hours (TWh) in 2025 – equivalent to powering an additional7.7 million homes.
Data centers are the largest driver of this growth
预计美国将拥有除中国以外最高份额的新建数据中心。自2020年以来，受数字化、云计算和人工智能的推动，仅美国托管型数据中心市场规模就翻了一番。为了满足这一供电需求，到2030年，数据中心可能需要消耗美国总电力需求的 8-12%，而目前这一比例仅为 3-4%
The US is expected to see the highest share of new data centers outside China. Since 2020, the US colocation data center market alone has doubled, driven by digitization, cloud, and AI.To power this, data centers by 2030 could require 8-12% of the total US demand compared to 3-4% today.
The US grid has not been able to keep pace with this demand.While utilities can likely generate sufficient power to meet data center needs, they face bottlenecks with transporting that power via transmission and distribution infrastructure. As a result, grid interconnection takes longer, there is more congestion on the network, and capacity is increasingly expensive.If the US continues to build high-voltage transmission infrastructure at its current rate, it will take at least 80 years to deliver what we need over the next decade.
New data center projects will struggle to get timely access to power
在美国，预计未来五年将有 55 GW（吉瓦） 的数据中心 IT容量上线。相比之下，这是纽约市平均用电容量的10倍——而且这还不包括冷却系统所需的额外电力。
In the US, 55 GW of data center IT capacity is expected to come online in the next five years. For comparison, this is 10 times the average power capacity used by New York City– and does not include the additional power needed for cooling systems.
我们已经看到数据中心IT容量的扩建正在加速，目前已宣布的容量约为 20 GW，并且我们预计这一数字将持续增长。我们预计，未来五年内至少还将宣布 35 GW 的数据中心容量，以满足预期的需求。
We are already seeing data center IT capacity buildout ramp up with ~20 GW of capacity announced so far, and we expect it to continue growing.Weexpect that at least another 35 GW of data center capacity will be announced within the next five years to meet projected data center demand.
变革的步伐正在加快
The pace of change is accelerating
我们的研究揭示了数据中心领导者为解决“通电交付周期”挑战所持的想法和采取的行动。在听取这些领导者的意见时，我们对变革的速度以及一种日益增长的预期感到惊讶，即：现场发电将在数据中心项目供电中扮演更重要的角色。
Our research sheds light on what data center leaders are thinking and doing to solve time to power challenges. In hearing from these leaders, we were surprised by the pace of change and the growing expectation `that onsite power generation will play a greater role in powering data center projects.
More data center leaders are taking responsibility for their power needs
他们通过现场发电来实现这一目标，并与电力公司紧密合作，以减轻公共电网和输电基础设施的负担。历史上，数据中心主要将现场电源用于备用目的，但我们正看到一种向将现场发电作为主要电源转变的趋势。
They do so by generating power onsite, working in close partnership with utilities to lessen the burden on the collective power grid and transmission infrastructure. Historically, data centers have used onsite power mostly for backup purposes, but we are seeing a shift toward onsite power generation as a primary source of power.
数据中心领导者日益倾向于将现场发电作为主要能源
Data center leaders increasingly expect onsite power to be adopted as a primary energy source
领导者预计，到2030年，约有 30% 的数据中心站点将使用某种形式的现场发电，这一比例是仅七个月前的 2.3倍。我们发现，新的数据中心项目公告印证了这一预期。2024年，涉及现场发电的项目公告数量超过了过去四年的总和。目前，全球已宣布的现场发电总容量已高达 8.7 GW，其中预计在2030年前落地的有 4.8 GW。
Leaders expect approximately 30% of all data center sites to use some onsite power by 2030, 2.3 times more than just seven months prior. We find that new data center announcements corroborate this expectation. In 2024, there were more announcements featuring onsite power than the previous four years combined. Total announced capacity of onsite power is now up to 8.7 GW globally, including 4.8 GW expected before 2030.
数据中心领导者预计，到2030年拥有至少部分现场发电的数据中心平均占比
Average share of data center sites expected to have at leastsome onsite power generation in 2030, as reported by datacenter leaderslo
全球各地的数据中心站点正在采纳现场发电解决方案
Data center sites around the world are adopting onsite power solutions
近期宣布采用现场发电的数据中心
Recently announced data centers using onsite power
在平衡各项优先级时，领导者对“价值”的看法发生了变化
Leaders see value differently as they balance priorities
新建数据中心需要平衡更多的优先事项
New data centers need to balance more priorities
通过我们的调研和访谈，我们询问数据中心领导者，在考量不同的技术和解决方案时，什么才是最重要的。我们了解到，决策者目前使用七个主要标准来评估各种选项。
Through our surveys and interviews, we asked data center leaders what matters when they consider different technologies and solutions. We learned that decision makers use seven primary criteria to evaluate options.
值得注意的是，这不再仅仅是关于成本和可靠性的问题。相反，他们还将以下因素的价值和影响纳入考量：通电交付周期,可持续性,功率密度,在某些情况下，还包括支持要求更高且波动剧烈的AI工作负载的能力。
Notably, it is no longer just a cost and reliability question. Instead, they also factor in the value and impact of time to power, sustainability, power density, and in some cases, the ability to support more demanding and fluctuating AI workloads.
“通电交付周期”在价值公式中扮演关键角色
Time to power plays a key role in the value equation
我们发现，如果意味着能够为即将到来的数据中心项目更快地获得电力，领导者现在愿意比七个月前多投入 50% 的资金。
We found that leaders are now ready to invest 50% more than seven months ago if that means they can access power faster for upcoming data center projects.11
这一转变突显了一个重要的洞察：抢在竞争对手之前上线是一种战略优势，特别是对于AI数据中心而言。领导者们意识到，加速获取电力是在这个快速发展的格局中确保领先地位的关键。
This shift underscores an important insight: being online ahead of competitors is a strategic advantage, particularly for AI data centers. Leaders recognize that accelerating power access is key to securing a leading position in this rapidly evolving landscape.
选择现场发电技术的七个关键标准
Seven key criteria for choosing onsite power technology
1.可靠性:
1. Reliability
将停机时间降至最低，且提供与电网持平或更优的可靠性（至少 99.9%，有时高达99.9999%），这对于满足数据中心的业务需求和客户需求至关重要。
Minimizing outages with reliability at par or superior to the grid (at least 99.9% and sometimes up to 99.9999%) is critical to delivering on the business and customer needs of a data center
2.通电交付周期
2. Time to power
让数据中心更快上线可以提供显著的经济优势（例如：更早获得营收机会）和战略优势（例如：在 AI 领域确立领导地位）。
Bringing a data center online faster can provide significant economic advantages (e.g., earlier revenue opportunities) and strategic advantages (e.g., leadership in AI)
3. Cost
除了特定技术的单项成本外，领导者们现在更全面地考虑价值和投资回报率（ROI），将经济效益和战略效益同时纳入考量。
Beyond the cost of a particular technology, leaders consider value and return on investment more holistically, factoring in both economic benefits and strategic benefits
4.负载灵活性
4. Load flexibility
AI 数据中心涉及更大的电力需求以及短时间内的剧烈波动，并非所有的电力解决方案都能应对这些负载变化。
AI data centers involve greater power demands and fluctuations over short time frames, and not all power solutions can handle these load shifts
5.可持续性：空气质量与噪音
5. Sustainability: Air quality and noise
外界对数据中心给当地社区带来的影响（如噪音扰民、废气排放）审查日益严格，数据中心需要遵守当地的行政许可标准。
There is increasing scrutiny of data centers’ impact on their local communities and a need to comply with local permitting standards
6.可持续性：排放强度
6. Sustainability: Emissions intensity
数据中心必须遵守企业的可持续发展承诺（如净零排放目标）以及政府的碳排放法规。
Data centers must comply with companies’ sustainability commitments as well as carbon emissions regulations
7.功率密度
7. Power density
具有更高功率密度的发电技术能够更高效地利用空间，从而留出更多空间用于部署产生营收的 IT 设备。
Technologies with greater power density enable a more efficient use of space for revenue-generating IT equipment
深度解析现场发电技术
Unpacking onsite power technologies
近期发布的公告展示了一系列解决方案，反映了数据中心参与者在每个项目的具体背景下，如何平衡上述标准。例如：
Recent announcements feature a range of solutions as data center players balance the above criteria in the context of each project. Some examples include:
燃气轮机：燃气-蒸汽联合循环机组（CCGT）利用一台燃气轮机和一台蒸汽轮机来发电。它们主要见于需要 500 MW 或更高容量的大型 AI 训练数据中心，因为在这个规模下它们具有成本效益。 规模较小的推理数据中心专注于已训练模型的应用，通常依赖单循环燃气轮机，有时会与电池储能系统配对使用。 虽然这些技术是目前最常见的现场发电解决方案，但漫长的供应链交付周期和较高的二氧化碳排放量是其限制因素。
Gas turbines：Combined-cycle gas turbines (CCGT) use a gas turbine and a steam turbine to generate electricity. They are primarily seen in large AI training data centers that require 500 MW or more capacity, as they are cost effective at this scale. Smaller inferencing data centers – focused on the application of trained models – often rely on simple cycle turbines, sometimes paired with batteries. While these technologies are currently the most common onsite power solution, long supply chain lead times and higher carbon dioxide emissions are limiting factors.
燃料电池：燃料电池通过将天然气或氢气转化为电能来提供连续电力。它们在数据中心项目中正日益普及，因为相比单循环燃气轮机，它们部署速度快，且产生的排放和污染更少。虽然燃料电池可能比其他某些解决方案更昂贵，但数据中心领导者往往因其更快的部署速度、更简单的行政许可流程以及更高的功率密度而优先选择它们，这些都能带来经济和战略优势。此外，由于燃料电池直接产生直流电（DC），随着数据中心为了获得简化电源架构带来的可靠性和效率效益而越来越多地探索直流供电，燃料电池也已准备好继续满足这一电力需求。
Fuel cells：Fuel cells provide continuous electricity by converting natural gas or hydrogen into electricity. They are gaining popularity in data center projects because they are quick to deploy and produce fewer emissions and less pollution than simple cycle turbines.While fuel cells can be more expensive than some other solutions, data center leaders often prioritize them for their faster deployment, easier permitting,and greater power density, which deliver both economic and strategic advantages. Since fuel cells generate direct current (DC) electricity, they are also poised to continue meeting power demands as data centers increasingly explore DC for the reliability and efficiency benefits of its simplified power architecture.
新兴技术：数据中心领导者对地热能、小型模块化反应堆（SMRs）以及带有碳捕获与封存（CCS）技术的燃气发电等新兴技术表现出越来越乐观的态度。随着各公司致力于实现可持续发展承诺，我们看到这些技术正出现在长期路线图中。 近期与这些技术相关的公告主要集中在预计于 2030年之后上线的数据中心，因为这些技术仍处于早期开发阶段。在零碳燃料选项准备好大规模部署之前，天然气解决方案预计将起到过渡桥梁的作用。 Emerging technologies：Data center leaders are showing growing optimism about emerging technologies such as geothermal power, small modular reactors (SMRs), and gas generation with carbon capture and sequestration (CCS). We are seeing these technologies appear in long-term roadmaps as companies work toward sustainability commitments.Recent announcements related to these technologies focus on data centers expected to go online after 2030, due to the technologies’ early stage of development. Natural gas solutions are expected to bridge the gap until net zero fuel options are ready for large-scale deployment.
数据中心正在使用多种技术来满足其现场电力需求
Data centers are using several technologies to meet their onsite power needs
已宣布的数据中心现场发电容量（单位：GW）
GW of announced onsite power generation for data centers
2030年前可用的数据中心现场发电公告主要涉及燃气轮机、燃料电池和往复式发动机。其他数据中心正在采纳地热能和 SMR 等新兴技术，我们预计这些技术主要将在 2030 年后变得可用。
Announcements of onsite power generation for data centers available before 2030 primarily involve turbines.fuel cells, and reciprocating engines.Other data centers are adopting emerging technologies such asgeothermal power and SMRs, which we expect would largely become available after 2030.
DKC交流群邀请
深知社创立于2017年，是数字能源行业工程技术人员学习、交流、合作、创新的虚拟社区平台。深知社秉承互联网精神，践行工程师文化，提倡全球视野、终身学习、乐于分享的博学深知理念，感兴趣的同学可以扫描下方二维码申请进群交流学习。
翻译：
校对：
张衍飞
邮储银行数据中心 运维主管
DKV（DeepKnowledge Volunteer）精英成员
公众号声明：
本文并非官方认可的中文版本，仅供读者学习参考，不得用于任何商业用途，文章内容请以英文原版为准，本文不代表深知社观点。文中内容来自互联网，如有侵权，将在24小时内删除。中文版未经公众号DeepKnowledge文字授权，请勿转载。
推荐阅读：
