The Zhitong Finance App learned that investment bank Bernstein said that the slowdown in artificial intelligence (AI) model training called for by industry giants may have a negative impact on data center developers and new cloud service providers. Among them, CoreWeave (CRWV.US) is probably one of the most affected companies.
Anthropic CEO Dario Amoudi called for a slowing down of cutting-edge AI model development in a cautionary post published on September 12. In the article, he said bluntly that the risks posed by AI are “serious” and that time must be taken to deal with these risks. The appeal was quickly answered by two key figures. Musk retweeted Amoudi's post on social media platform X and added: “Dario is right.” Oltman, the leader of OpenAI, wrote on X: “I agree with Dario. We need to control the pace of advancement of cutting-edge AI.”
Bernstein analyst Maidson Rezaei said in an investor report on Monday that a slowdown in model training or a decrease in training volume will mainly affect data centers in rural areas. Analysts said, “From the perspective of infrastructure and new cloud service providers, declining or slowing training demand will drain demand from rural data center site selection, and many of these data centers are built specifically for this training demand that is not sensitive to delays.”
The analyst added, “We have a data center development project reserve with a total nameplate capacity (nameplate capacity) of 488 GW in the US, of which 170 GW is considered credible. Of this 488 GW of reserves, 36% are in rural areas, and another 34% are in Tier 3 markets (such as Western Texas). This means that 70% of development project reserves are built around training requirements (or inference needs that are not sensitive to delays).”
According to this analysis, CoreWeave has the most prominent risk exposure. The analyst said, “We estimate that 25% of CoreWeave's electricity capacity currently in operation in the US is in Tier 3 and Tier 4 markets. Additionally, approximately 74% of its contracted electricity capacity is also located in these markets.” “The backlog of orders mainly consists of 'take-or-pay' (pay-or-pay) contracts, so we don't expect this part to be threatened, but if training development slows down, we may see a decline in electricity capacity demand in rural areas where contracts have been signed but not yet sold.”
By contrast, data centers located in large and small metropolitan areas seem to be the best equipped to withstand the effects of a slowdown in training demand. This includes Equinix (EQIX.US), Digital Realty Trust (DLR.US), and Csquare (CSQR.US). The companies' data center capacity is 95%, 92%, and 94%, respectively, in large or small metropolitan areas.
Bernstein gave CoreWeave a “outperforming market” rating, and the target price was $74. Meanwhile, the investment agency gave Csquare, Digital Realty, and Equinix “outperforming the market” ratings, with target prices of $27, $226, and $1,270, respectively.
Bernstein notes that even if data centers move from training oriented to inference orientation — a necessary shift to enable more agentic (agentic) workflows — data centers based in urban areas “are still the safest and most valuable.”