The emergence of cheaper and more effective AI models at the back of Chinese depth could reshape the demand for data centers, stimulating a sector that investors are already betting massively would continue to explode.
For years now, analysts have planned exponential growth in data centers – the critical infrastructure required to fuel the global digital transition and the formation of important language models (LLM).
The Chinese startup Deepseek IA Model sent investors in tremors at the end of January, while the launch of its R1 model has raised questions on American domination in the AI sector and if the efficiency gains of the efficiency of developer could at the cutting edge of data center capacity demand.
Data centers often take at least two years to build and orders have already been taken into account since 2025 – which means that the launch of the R1 disruptor model is unlikely to have an immediate impact. While the launch of the Deepseek R1 initially led some analysts to temper their forecasts when they wondered if the money pumped in the sector could have been somewhat “erroneous”, the experts told CNBC that the Models were built at a lower cost and with less powerful fleas could finally become an accelerator for the market.
Handy perspectives
Deepseek underlines how data centers are vulnerable to changes in accounts around AI spending, according to Barclays analysts. If the allegations of efficiency made by the Chinese startup are confirmed, the development shows that “the hundreds of billions of dollars dedicated to the development of AI therefore seem wrong and that the capital expenditure plans of hyperscalers could be reconsidered”, Analysts led by Brendan Lynch, said In A note published on January 27.
They added that, if AI requires less infrastructure, it will be the “lowest quality installations” – which are the least energy efficient – which could cope with lower demand and lower prices .
UBS analysts have noted that around a third of the growth projections of the current data centers are based on the construction and development of generative artificial intelligence – AI which can create images from written prompts. These forecasts did not take into account a radical improvement in efficiency, UBS said in a note on January 28.
UBS initially planned in April of last year that the global market for data centers equipment would increase from 10 to 15% over a period of three years to 2028. This week, bank analysts declared that new data and data and Expert calls ultimately lead to a more increased perspective on the market. The company now expects the sector’s income to increase by 20% in 2025 and saw “the range towards the upper end” of the growth range from 10 to 15% for at least the start of the 2026 period -2028, analysts said in a note on Wednesday.
The jury is “always outside” on the question of whether Deepseek needed 20 to 30 times less computing power per request for inference, Andre Kukhin, research analyst on actions at UBS, told CNBC – Referring to the data execution process via an AI model to make a prediction or resolve a task.
“Although it is more effective by token, it needs more tokens per request because it is a model of reasoning rather than” speech flow “… The main thing is that we do not think Not that this considerably reduces the demand for energy for inference “,” Kukhnin explained.
The Goldman Sachs Research Department provides for the balance of the supply and demand from the “tightening” data center in the coming years, reaching a peak at the end of 2026, then moderating from 2027.
If the efficiency of the gains stimulates the levels of capital expenditure (CAPEX) reduced by the main investors, this could “mitigate the risk of long -term excess offer that we see in 2027 and beyond – which, in our opinion on The data centers market, “James Schneider, main research analysts in actions by Goldman Sachs, noted in a February 4 report.
There is still a lot to determine on the impact of emerging technology, less than three weeks since Deepseek published his data. R1 is not enough for itself to “change the needle” with regard to demand, according to Andrew McMillan, partner of the RPC law firm.
“The appetite of investors will be tempered if it can be demonstrated as being reproducible, and therefore there will be a much lower demand for data processing in the future than today, or at least it will not continue on the Same way of growth, “said McMillan, specializing in mergers and acquisitions and data governance.
“I think that in the long term, it will be really interesting to see if this structural approach is able to settle, and I think that can affect the shape of the market.”
‘Fuel in fire’
Actions vulnerable to changes in the data centers market fell on January 27. Schneider ElectricThe European company most exposed to data centers according to UBS, has lost more than 9%, Siemens Energy stocks have lost 20% and Abb Closed 6% lower over the day.
Some actions have since recovered their losses, recovering from the instinctive reaction of the markets. Mega-hepatic profits such as alphabet Google And Meta Also instilled in confidence, the two companies are involved in investments of several billion dollars after technological sale.
There has not been “a lot of room for error” in the sector, said Kukhnin d’Ubs. “This is why some of the actions have dropped and are not immediately redeemed, because people already have a lot of actions and are now trying to determine if this is an opportunity to add or if it is the opposite.”
He added that lower costs indicate a potential democratization of AI, which could lead to acceleration in the adoption of technology – which is “something very difficult to quantify”.
The data center market will also continue to be fueled by the digital transition, which takes place separately from the progress of the AI. “The generative AI was in a way almost the frosting on the cake, but has become a very thick layer of frosting, certainly in terms of future growth,” said Kukhnin.
Bruce Owen, President Emea in Equinix, said that the company is “well positioned because the AI technology curve is bent”, adding that it expects the advent of more effective models to be a “Accelerator”, for AI.
“An additional dynamic that we could see is the” Jews Paradox “, which postulates that the increased efficiency of a resource can lead to greater consumption of this resource,” he told CNBC.
Ryan Cox, head of AI, in the Synechron consultation company, also expects the paradox effect to see more efficient technology finally leads to a greater demand for data centers.
“It is a really complex equation,” he told CNBC, noting that there are several opposite winds and rear winds when it comes to determining the potential changes in demand. He shared that Synechron customers are looking for “safe” options to indirectly use Deepseek, as via Hugging Face, a repository for AI models.
“Overall, I think efficiency will feed adoption, and I think it will continue to increase use, even if these costs drop. Race to these more advanced models and wider applications, the ‘Use of AI means that the overall demand for data centers will increase and decrease, “noted Cox.
