Meta considers selling excess AI computing resources through cloud services
On July 1, Reuters reported, citing Bloomberg, that "Meta is considering a cloud business that would sell excess AI computing resources to external clients." This plan includes making various AI models available through Meta's infrastructure and potentially renting out processing power to external customers. The proposal is still in the development stage and may change in the future. The company has not officially commented on the report.
It is difficult to definitively conclude that this issue indicates a slowdown in semiconductor demand, as Meta is not looking to reduce its investments in AI infrastructure. Instead, it appears to be an attempt to recoup substantial data center investments through external revenue. However, investors are particularly focused on the notion of "excess computing resources." If major tech companies can sell their existing or developing processing capabilities to external clients, the perception that AI infrastructure is critically lacking may weaken.
This development would provide more options for companies looking to develop AI services. Previously, they had to secure expensive GPUs directly or rent computing capacity from cloud providers like CoreWeave. With Meta potentially entering the market as a large supplier, the burden of building their own servers would decrease. While the actual demand for GPUs and HBM will not disappear immediately, the expectations for price increases and supply shortages may diminish.
Market reactions have reflected these concerns. Meta's stock rose as the potential for recouping investments became more apparent, while companies like CoreWeave and Nevious, which rent AI infrastructure, saw declines due to heightened competition fears. Investors interpreted this development as a signal of intensified competition in the rental market rather than an expansion of AI infrastructure demand.
The HBM market could also face indirect pressure. HBM is a critical component for AI servers, so the demand itself is unlikely to drop immediately. However, as the cloud rental market grows, companies may feel less pressure to build their own GPU servers. This does not mean that HBM demand will decrease right away, but it could pose challenges for maintaining high prices and profit margins.
Samsung and SK Hynix face stock premium challenges
For domestic semiconductor companies, the impact is likely to be felt more as a stock burden than as a performance setback. SK Hynix is considered a key supplier in the HBM market. The expansion of AI server investments has previously boosted both performance and stock prices. Even if Meta's cloud initiative is perceived as a signal of easing computing resource shortages, it is unlikely to disrupt SK Hynix's HBM orders immediately. However, expectations for HBM oversupply and price increases may be adjusted.
Samsung Electronics faces a more complex situation. The longer the HBM supply shortage persists, the more opportunities Samsung has to catch up. However, if the narrative of AI infrastructure shortages weakens, Samsung's negotiating power regarding HBM prices may be limited. Conversely, if Meta's cloud business expands, demand for server memory and enterprise solid-state drives (SSDs) is likely to remain stable, as large-scale memory and storage devices are still necessary for data storage, retrieval, and inference service operations.
Domestic materials, parts, and equipment stocks may also face increased price pressures. These stocks often reflect expectations for HBM expansion and advanced packaging investments. The outlook for AI server growth has also been factored into stock prices. Even if Meta's plans materialize, semiconductor manufacturers' long-term investment strategies may not change immediately. However, if the premise that "AI processing capacity is critically lacking" weakens, the situation could shift. There may be a reassessment of whether the expectations for increased equipment orders were overly optimistic.
Ultimately, Meta's exploration of an AI cloud service is more of a variable testing the expectations of supply instability rather than a signal that semiconductor demand is collapsing. The trend of increasing AI investments is likely to continue. However, if big tech companies can supply their computing resources externally, the scarcity premium for GPUs and HBM may decrease. For domestic semiconductor companies, it is more likely that the expectations for price increases reflected in stock prices will be shaken rather than an immediate drop in orders.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.