After the discussion surrounding whether development around cutting-edge AI should slow down intensified, Wall Street began to re-evaluate the next phase of trends in the AI sector. Foreign media reports that the market is currently more concerned with one question: if the pace of model advancement slows, will technology companies also reduce their capital expenditures related to AI.
The investment in computing power may not necessarily cool down simultaneously.
Analysts believe that slowing down the development of cutting-edge technologies does not necessarily mean that related expenditures will significantly decrease. Their reasoning is that such statements are more about reducing the development speed from "extremely fast" to "relatively fast," rather than halting progress altogether.
The institution also mentioned that the computing demands for AI are increasingly driven by inference, rather than relying solely on large-scale training. The current supply of computing power is still not sufficient to meet these demands, so the likelihood of companies significantly reducing their investments is not high. Bernstein believes that stronger security measures may actually alleviate political and social concerns, thereby supporting the long-term adoption of AI.
Funds may be directed towards software and security companies.
If the total expenditure of AI does not show a significant decline, the market will be more concerned about where the funds will flow next. The ISI team under Evercore believes that in the next phase of AI transactions, there may be a greater focus on helping software companies deploy AI.
The article states that most software companies do not develop cutting-edge models themselves; instead, they provide tools and processes for businesses using existing models. Therefore, their growth depends more on whether businesses can integrate these models into their daily operations, rather than how quickly cutting-edge models are being iterated.
Under this framework, the importance of aspects such as identity management, governance, monitoring, and security may increase. The reason is that even if the update of cutting-edge models slows down, enterprises still need these fundamental capabilities when deploying AI proxies.
Market divergence focuses on pace and valuation.
RBC Capital Markets It should be noted that the initial market reaction may have overlooked another aspect. A slowdown in the development pace might indeed give laggards more time to catch up, but it could also delay the launch of stronger AI products, thereby affecting the subsequent growth prospects of some software companies.
Saxo, the Chief Investment Strategist, believes that this round of volatility is more like a shock to sentiment and valuations, rather than a collapse in AI demand itself. She pointed out that the call for increased testing should not be directly interpreted as tech companies stopping the construction of data centers or purchasing computing equipment.
However, the article also mentions that if AI company ultimately reduces the release of cutting-edge models or scales back on ultra-large-scale training projects, then the demand for high-end processors, memory chips, and advanced packaging capabilities may be lower than current market expectations. Among these, the pressure on memory chips could be even more evident, as manufacturers are expanding production in anticipation of strong demand. Once demand is delayed, the release of new production capacity could lead to oversupply and a decline in prices.
