The appearance of the most cheapest and most efficient artificial intelligence models against the background of Deepseek in China can reshape the demand for databases, enhancing a sector that investors were already betting to continue to prosper.

For years, analysts have expected tremendous growth in databases – critical infrastructure needed to operate the world’s digital transition and train large language models (LLMS).

Deepseek Ai Model has sent investors In tension In late January, when the launch of its R1 model raised questions about the dominance of the United States in the artificial intelligence sector and whether the developer Efficiency gains The demand can prepare for the ability of the data center.

Data centers often take at least two years for construction, and orders have already been established in 2025 – this means that the launch of the R1 sabotage model is unlikely to have any immediate effect. While the release of the R1 from Deepseek initially reduced some analysts from their expectations because they wondered whether the money pumped into the sector might be “somewhat misleading”, experts told CNBC that the models that were built more licensing and with less chips are Ultimately power, can ultimately become speeding up in the market.

Deepseek highlights how databases are displayed for transformations in novels about artificial intelligence spending, according to analysts in Barclays. If the competency claims submitted from the start of the Chinese operation are confirmed, the development shows that “hundreds of billions of dollars allocated to the development of artificial intelligence, therefore, it appears that the misleading capital spending plans, said analysts led by Brendan Lynch, he said, he said, he said, he said, he said. in Note published on January 27.

They added that if artificial intelligence requires less infrastructure, it will be “the lowest quality facilities” – which is the least efficient in energy use – which may face the weakest demand and the weakest prices.

Meanwhile, UBS analysts have indicated that about a third of the current data center growth forecast depends on building and developing obstetric artificial intelligence-artificial intelligence that can create pictures of written claims. UBS said in the January 28 memo, that these expectations were not treated in a fundamental improvement in efficiency.

UBS initially expects last April last year that the global data center equipment market will grow by 10-15 % during the three-year period until 2028. Analysts said in a memorandum on Wednesday that the sector’s revenues expect the sector’s revenues to grow by 20 % in 2025 and believe that “the range is heading towards the upper side” for the growth range of at least 10-15 % at the beginning of the period 2026-2028. .

The jury “is still outside” about whether Dibsic needs a computing energy less than 20 to 30 times for each inference, Andre Koknin, UBS stock research analyst, for CNBC – referred to the data operation process through the artificial intelligence model To provide a prediction or solution of the task.

“Although it is more efficient for every symbol, it needs more symbols in every query because it is a model of thinking instead of” the flow of speech “… The bottom line is that we do not believe that it greatly reduces the demand for energy for reasoning,” explained. Kokhnin.

The Goldman SACHS Research Department expects that the “tightening” of the supply and demand in the data center in the coming years, as it reached a peak in late 2026 and then moderate from 2027 onwards.

If the gains of efficiency lead to low levels of capital spending (CEPEX) from the main investors, this can “reduce the risk of long-term supply in the market that we see in 2027 and beyond-which we believe is an important consideration that can pay more durability The least patrol in the Data Center market, “James Schneider, chief stockpile of stock research in Goldman Sachs, in a report on February 4.

Not a lot has not yet been determined about the effect of emerging technology, less than three weeks since Deepseek has published its data. R1 is not enough on its own to “change the needle” when it comes to the request, according to Andro McMelan, a partner in the RPC law office.

“The investor’s appetite will be diluted if it can be shown that this can be repeated, and therefore there will be a much lower demand for data processing in the future than it is now, or at least it will not continue the course of growth itself,” he said. McMillan, specialized in integration, purchase and data governance.

“I think in the long run, it will be really interesting to know if this structural approach is able to stick, and I think it can affect the shape of the market.”

The weak shares of transformations in the data center market decreased on January 27. Schneider ElectricityThe European company most exposed to data centers, according to UBS, has lost more than 9 %, Siemens Energy 20 % fall shares and ABB Close 6 % less a day.

Since then, some shares have regained their losses, recovering from the knee reaction from the market. MEGA-HYPERSCALEs Google and Dead He also instilled confidence, where both companies are committed Investments of billions of dollars After selling technology.

Cookien Obus said: There was no “big space for error” in the Strip. “For this reason, some shares have declined and were not bought immediately, because people already have a lot of shares and are now trying to know if this is the opportunity to add or if it is the opposite.”

He added that the low costs indicate the existence of capabilities Democracy from artificial intelligenceWhich may lead to an acceleration in the adoption of technology – which is “something that is difficult to determine.”

The Data Center market will also continue in digital transition, which occurs separately from developments in artificial intelligence. “The Truntic intelligence was almost a kind of ice on the cake, but it became a very thick layer of ice, and definitely in terms of future growth,” said Kokhnin.

Bruce Owen, President of Europe, the Middle East and Africa in Equinix, said that the company “is in good position because the artificial intelligence technology curve is determined,” adding that it is expected that the appearance of the most efficient models will be “accelerated”, for artificial intelligence.

“An additional dynamic we may see is” Jevons Paradox “, which assumes that increased efficiency in the resource can lead to an increase in the consumption of this resource,” CNBC told CNBC.

Ryan Cox, head of the Artificial Intelligence Company, at Synechron Consulting Company, also expects that Jevons Paradox Effect will see more efficient technology that ultimately leads to more demand for the data center.

“It is a truly complex equation,” CNBC told CNBC, noting that there are many opposite winds and a spirit of disagreement when it comes to determining the potential transformations of the request. Share that Synechron clients are following “safe” options to indirectly use Deepseek, such as Via Laging Face, a warehouse of artificial intelligence models.

“In general, I think efficiency will support its adoption, and I think it will continue to increase use, even as these costs are low. Sweat towards these most advanced models and broader applications, and the use of artificial intelligence, means that Cox indicated to that.