AI Infrastructure Reassessed as Revenue-Generating Asset
Cloud Business Value of Naver, Samsung SDS, and Others Comes into Focus

Which Korean Companies Stand to Benefit from Meta's Entry into the Cloud Business? [Weekend Money] View original image

An analysis has emerged suggesting that the market’s assessment of AI infrastructure investments could shift, as Meta is reportedly considering entering the cloud business by selling its proprietary artificial intelligence (AI) infrastructure to external clients. The outlook is that large-scale AI infrastructure investments, which were previously viewed solely as expenses, could transform into direct revenue-generating assets in the future, leading to a renewed evaluation of the enterprise value of existing cloud service providers (CSPs).


Sohye Kim, a researcher at Hanwha Investment & Securities, stated, "Meta’s AI infrastructure investments have ballooned significantly, and it has become difficult to explain their recovery solely through advertising and chatbot revenues. There is now the possibility of converting AI infrastructure investments, once considered costs for advertising and AI chatbots, into AI infrastructure assets that can be sold externally," she analyzed.


It is reported that Meta is considering a plan to lease out its internally built graphics processing units (GPUs) and data centers to external customers, thereby turning its AI infrastructure investments from supporting assets for its advertising business into direct sources of revenue. Both approaches are being discussed: one similar to Amazon Web Services’ (AWS) Bedrock, which sells access to AI models, and another like CoreWeave and Nebius, which lease computing resources.


Kim explained, "With improved AI training efficiency, there is a high possibility that surplus computing capacity has emerged at Meta’s data centers. At this point, Meta needs a rationale to defend the return on invested capital (ROIC) from its AI infrastructure investments." She added, "Given that the pricing competitiveness of Meta’s proprietary model APIs is inferior to OpenAI, Anthropic, and Google, leasing infrastructure and selling computing power have become more realistic options."


She anticipated that this move by Meta would highlight the competitiveness of existing global CSP companies even more. "In the short term, it will be difficult for Meta to replace the positions of major CSPs such as AWS, Azure, and Google Cloud Platform (GCP). The CSP business is not determined by GPU capacity alone; there are still significant gaps with Meta in terms of security, compliance, data governance, and developer ecosystems," she explained.


Regarding domestic CSP companies, she assessed that although it is difficult to apply the same logic as for global firms, their investment appeal could increase.


She elaborated, "The rationale is based on having self-owned hyperscale data centers and a large number of GPUs for selling surplus computing externally. However, Naver and Samsung SDS do not possess GPU infrastructure on this scale, so the logic would inevitably have to be applied differently."


Nevertheless, she concluded, "The key question has now shifted to which company can recoup AI infrastructure investments with a high return on invested capital. There is ample possibility that the value of domestic software companies will be re-evaluated."



She continued, "For Naver, there is growing expectation that AI infrastructure investments previously regarded only as costs for advertising and services could be recouped through B2B sovereign AI and public cloud demand. For Samsung SDS, as demand for cloud services from global CSPs remains robust, strong sales of managed service provider (MSP) cloud management services can be expected. I maintain my previous view that the value of domestic CSP companies will be revised upward," she stated.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing