人工智能时代前后：数据中心行业进入能源供应受限新时代调研报告

Surveying the Data CenterIndustry as It Enters a New Ageof Constrained Energy Supply In the beginning, Taking stock of the early days ofthe “After AI” age the North American data center industry operated in a relatively stable status quo for securing grid power. For decades,data centers would secure dozens, or hundreds, of megawatts of utility capacity at a time. They would receive power on aworkable timeline. No problem. In such a pivotal time for energy in the industry, the leaders in datacenter energy management, AlphaStruxure and Schneider Electricpartnered with Data Center Frontier to survey 149 senior industryprofessionals. This is the first major survey undertaken to measurehow they’re adapting to the unfolding energy crunch. The resultstell a stark story of how much the industry’s outlook has changedin such a short time — and the ways it’s innovating to secure theenergy supply in the After AI age. Then, in late 2022, in the halls of utility regulators and at boardroom tables, everything changed. As generative AI arrivedon the scene, the industry entered a steeplechase race to win the AI prize, with obstacles arising across the course. A lackof chips. A lack of land. In a matter of months, top markets became tapped out of available real estate, while utility waittimes stretched out to seven, nine, or even 12 years. Secondary markets are also seeing rapid growth and long waits, asdata centers soak up the last-remaining pockets of available capacity. Although the race may be entering a new phase,with some in the industry starting to scale down their growth projections, the overall trajectory remains clear: more demand,more data centers, and more energy. How the industry is transitioningfrom “Before AI” to “After AI” Theme #2: Tradeoffs and timelines Theme #1: Barriers to build-out Theme #3: Assessing the alternative 44% ranked barrier slowing down data centerprojects isgrid constraints, with 92%seeing it as an obstacle.#1 of respondents indicate theiraverage quoted utility wait timesare longer than4 years. 6IN10report they would deployon-site powergeneration systems if they ran into concernsabout power availability — the top-ranked option. #1region for “Plan B” power availability if thefirst choice couldn’t provide timely powerwas the Midwest. 48% #1ranked barrier in getting more electricalcapacity is overlylong quoted wait timesfrom utilities. of respondents report the size ofa single new data center projectaverages over100 megawatts. 6IN10are building more than10 datacentersover the next five years,with25% building at least 30. 35% #1 reported theircarbon emissionsacceleratedover past two years,since the AI race began. ranked on-site power generationsolutions weresolar andbattery storage. 05Executive Foreword: Juan Macias, Vandana Singh,and Matthew Vincent 06Survey Demographics and Methods 07Theme 1: Barriers to Build-out:On wait times and scaling challenges 15Theme 2: Tradeoffs and Timelines:On evaluating options and impacts 22Theme 3: Assessing the Alternative:On workarounds and on-site power 28Conclusion:Where the industry is heading Executive Foreword The energy crunch in the “After AI” era:How data centers are adapting Juan MaciasCEO, AlphaStruxure Here’s what this report tells us about what is going on today: ◇◇Grid constraints are the #1 barrier: 92%of respondents seeit as a challenge, with long utility wait times slowingdown expansion. We are all witnessing an unparalleled moment in data centers,one characterized by a high-stakes race for more power andcompute capacity. Although we collectively hold decades ofexperience in the data center and energy industries, we’ve neverseen anything like this. ◇◇New projects are bigger than ever:Nearlyhalf (48%)ofrespondents say their new data centers now average100+MW, signaling an unprecedented scale of development. ◇◇Longer lead times are the new normal: 44%report utilitywait times of4+ years, while demand is spilling into secondarymarkets, stretching available capacity. For decades, securing grid capacity was a manageable process.But in late 2022, as generative AI took off, everything changed.Utility wait times in key markets have stretched to seven, nine,even potentially, twelve years,1forcing operators to rethink theirapproach to energy procurement, deployment, and sustainability. ◇◇The AI boom is impacting sustainability goals: 35%reportan increase in carbon emissions over the past two years asthe industry races to keep up with new demand. Vandana SinghSenior Vice President, Secure Power,North America, Schneider Electric ◇◇Alternatives are emerging: 62%of respondents reportedexploring on-site power generation. With no slowdown insight, data center operators are rethinking power procurementstrategies. To better understand how the industry is responding,Data Center Frontier, in partnership withAlphaStruxureand Schneider Electric, surveyed149 senior industryprofessionalson the shifting energy landscape. The f