CNBC projected that the emergence of cheaper AI models from China could increase the demand for data centers. Analysts reportedly tilted their forecasts after DeepSeek’s launch, anticipating immediate impacts on the industry. A report by CNBC indicated that Artificial Intelligence models built more cheaply with less powerful chips could boost the data center market. The report noted that some analysts reviewed their forecasts after DeepSeek’s launch. China’s cheaper AI models, such as DeepSeek, could reshape data center demand Analysts forecasted that data centers would experience growth in the coming years, especially after the launch of OpenAI’s ChatGPT. Newmark estimated that U.S data demand will double by 2030 to accommodate Artificial Intelligence (AI) applications. How China's DeepSeek could boost the already booming data center market https://t.co/x9kuk5cCTu — CNBC (@CNBC) February 14, 2025 Chinese startup DeepSeek launched last month, introducing a free, open-source large language model. The startup claimed it had spent under $6 million on its development. Investors reacted to the launch by questioning the large amounts of money institutions put into artificial intelligence and data centers. Some analysts reportedly altered their forecasts on data center demand, raising questions about whether the funds pumped into the AI industry were misguided. The report highlighted that data centers take approximately two years to build, and orders for 2025 have already been placed. It added that launching an RI model such as DeepSeek was unlikely to have an immediate impact. In a note published in late January, a group of analysts led by Brenda Lynch also commented on DeepSeek’s launch. They highlighted that confirmation of DeepSeek’s efficiency claims would show that the money allocated to AI development was misguided. The analysts concluded that such an event would require tech companies to reconsider their spending plans. They added that if AI needed less deployment infrastructure, it would lead to lower-quality facilities with the least energy efficiency facing weaker demands. Analysts project R1 model launch is unlikely to alter market demand Meanwhile, UBS analysts noted that one-third of the current bullish data center market projections relied on the development of generative AI. They added that the bullish forecasts did not consider improvements in efficiency such as the one highlighted by DeepSeek. In April, UBS forecasted the global data center equipment would grow by approximately 15% in three years. The analysts reviewed the forecasts this week, adding that new data and expert opinions lead to higher market projections. The firm revealed it expected the AI industry and data center sector to grow 20% this year. UBS maintained its 10% to 15% projection through 2026-2028. Andre Kukhin, equity research analyst at UBS, explained that while DeepSeek was more efficient per token, it needed more tokens per question. He noted that it was a reasoning model rather than a “flow of speech.” Kukhin expressed that he did not think the model would reduce the demand for power for inference. Goldman Sachs’ research department projected that data center supply and demand could be altered in the following years. The department revealed that demand could peak in late 2026 and then moderate in 2027. James Schneider, senior equity research analyst at Goldman Sachs, commented that if efficiency gains drove lower capital expenditures from major investors, it could mitigate the risk of long-term market oversupply projected in 2027 and beyond. He added that this was an important consideration that could increase the durability of the data center market. Andrew McMillan, partner at RPC law firm, mentioned that a lot was yet to be determined on the impact of Deepseek’s technology, given it was still in its nascent stages. He noted that R1 could not tilt the AI industry’s demand and supply balance. McMillan projected that investor appetite would be tempered if the technology could be shown to be replicable, leading to lower demand for data processing in the future than there is now. Cryptopolitan Academy: FREE Web3 Resume Cheat Sheet - Download Now
