Ai Boom Center Center does not break up but the “pause” is in trend

The scattered, database power centers have become a major part of the suburbs in some places as shopping centers and football fields. However, when Microsoft pulled the plug In the planned data centers in Ohio last month, she added to the questions about whether the nascent boom in the data center had already fell apart. WELLS FARGO report last Monday Saying that some data centers planned by Amazon Web Services are reviewed, added to market anxiety.

But the bust may have ended before it ever begins. And if nothing else, it enters the “pause” of some data centers Spending environment that remains strongS

“We continue to see an accelerated scaling of AI implementation in the data centers market, with signaling search signals reinforcing both our close and long -term growth,” said Giordano Albertacs, CEO of the Ohio Data Provider -based Verb When you call the profit last week. His shared ended the week to 22%.

Amazon and Nvidia both Confirmed last week that the data center market remains strong.

“There really is no significant change,” said Kevin Miller, Amazon Vice President at global data centers, at a conference organized by the Ham Institute of American Energy. “We continue to see a very strong demand and watch both in the next few years and in the long run and see that the numbers only rise up.”

This does not mean that strategic thinking about how, where and when to spend, does not change, since the AI ​​market is developing and the breakthroughs must be absorbed. For six weeks this year, in China, Deepseek broke out on stage, AI initiative was announced to power for $ 500 billion to Stargate power to $ 500 billion, and fears about rates and commercial wars poured the markets.

“All this has created a scenario where the data centers industry pauses a little,” says Pat Lynch, a executive manager of CBRE’s commercial real estate solutions. “I think this is a temporary break,” Lynch added, noting that the project of the project and its funnel remain significant and CBRE continues to execute deals. “I remain cautiously optimistic about future search, especially when you think of big AI training models,” Lynch said.

Microsoft has promised an investment of $ 1 billion in Ohio database based in the same area where Intel has planned chip factories, but the time line is slow.

“After careful consideration, we will not move forward with our plans to build data centers on LIKING County sites right now. We will continue to evaluate these sites in accordance with our investment strategy,” a Microsoft spokesman said in a CNBC statement.

Last week’s UBS report came to the conclusion that among all possible explanations for cancellation of the data center, the most likely Microsoft has overdone against the AI ​​Rush and is now reset by the projects that are currently the biggest meaning. He noted that Microsoft’s Lege Capex has grown 6.7 times over a period of two years, with a leasing liabilities of about $ 175 billion. “Microsoft redeem as high as possible the data center capacity as much as possible in 2022-2024 and now has the visibility to remove some of these” projects at an early stage, “UBS wrote. “We find the least support for the explanation” Demand, “his report adds.

Anat Ashkenazi, Alphabet CFO, described the cloud search environment as “tight” after His last revenue Thursday. “We can see variability at the growth rate of revenue in the cloud depending on the implementation of capacity every quarter,” she said. “We expect a relatively more capacity implementation by the end of 2025”

“We are not observing demand retreat, but a strategic redistribution,” says John Carafiel, a co -executed director of BGO, a global real estate investment manager with $ 83 billion under management, including a significant portfolio from the Data Center. The most significant players, according to him, do not back away, such as Microsoft, Google, Meta and Amazon, plan to spend over $ 300 billion on Capex this year-unleashed with AI infrastructure. And, he says, this does not include other basic players such as Openai and Oracle, both Participated in the Stargate projectS

“Instead of a bust, this is a deck adjusting to an environment where power, along with fiber, water and land, are scarce and strategic,” Karafiel said. Acceptance of a long -term enterprise will lead to the search for AI and data centers for the next decade. “We’re still not in the first Innning,” he said.

Power is the vital force of data centersHowever, data centers are not switching on and playing operations requiring copious amounts of electricity for computing power and fans to maintain the infrastructure cool. As the generative reception of AI passes from early experimentation to the application of a scale of the corporation, the need for low latency, highly effective data centers near end users will intensify, but will take time for the right set of conditions to arrange with the expected square footage of the data center.

“The new data centers are increasing so dramatically that the network cannot be up to date,” says Alan Shur, Chief Sales Officer of MicroGrid Developer Enchanted Rock. Three years ago, a large data center was 60 megawatts – enough power to supply 20,000 homes, but now he says that new data centers to support all uses of artificial intelligence want 500 megawatts or more.

This rapid growth of the use of electricity is at the top of the new demand for production and electrification of transport, which together weigh the supply and infrastructure. Data centers are a unique challenge for utilities, which must ensure that they can supply energy to all customers, even in times of peak demand. “That is why some utilities cite a long time to wait for interconnectedness for data centers,” Shur said. “Utilities need to invest in new substations and may also need to expand the transmission and generation, all of which take time,” he added.

CBRE has seen that data centers go from 2% of their portfolio in 2022 to 10% in 2024, and Lynch expects to continue to grow, and the proximity of electricity leads to the present market, as the data center builders are looking for areas with access to abundant power. Georgia, Texas and Ohio check many of the boxes that builders are looking for and if an area does not have a network capacity or infrastructure, it must be able to increase quickly.

“Having great power availability within 36 months is attractive to customers,” Lynch said.

Three percent of the world power has already been linked to data centers according to DataceNters.com.

Shur said Enchanted Rock data shows that there is a lot of power to meet the search – most of the time. From 8 760 hours a year, the network is stressed only for some of them. “If we can relieve the network search for those 100 to 500 hours, the long connection delays can be shortened,” he said.

An important distinction should be made between the idea of ​​a broader delay and some of the latest pauses accepted by major technology companies, according to a senior partner of McKinsey & Company Pankai Sahdeva, who examines the development of the Data Center and awaits a tide.

Based on the recent McKinsey modeling, which does not include tariffs, the data centers market is expected to grow in the range of 20% -25% over the next five to seven years, but there will be differences in growth rate from year to year. “It won’t be linear,” he said.

Changes in tariffs will introduce new cost pressures in AI and supply chains to the data center, especially with critical mineral tariffs on the horizon.

“These interruptions will increase hardware costs, affect the supply strategies, and require businesses to rethink their long -term public procurement models,” says John Archer, Senior Director of Delivery and Slalom Consulting Delivery Chain. In the short term, AI and cloud providers will need to apply cost reduction strategies, such as renegotiating supplier contracts and optimization of inventory.

“Longer-term, impetus to geographical diversification, joint production in tariff regions and the more in-depth integration of the analysis of the supply chain managed by AI can be expected to adapt to developing commercial policies,” Archer said.

One of the factors that has not changed is that Compute Power is currently expensive and much more is required for AI software and hardware, according to Suresh Venkates, CEO of Poet Technologies, a publicly traded company that is developing Power Solutions for data centers. “The AI ​​explosion causes data centers to find more efficient solutions, since AI requires the calculation of power in such a volume that it is unlike everything we have ever witnessed,” he said. “While a project center project can hit a wall, others are likely to emerge, as there is no indication of delaying demand for connectivity,” he added.