본문영역 바로가기 메인메뉴 바로가기 하단링크 바로가기

KISDI 정보통신정책연구원

KISDI 정보통신정책연구원

검색 검색 메뉴

KISDI News

  • KISDI: The AI market is shifting toward “full stack” competition

    • Pub date 2026-09-30
    • File There are no registered files.
※ URL(Korean): https://www.kisdi.re.kr/bbs/view.do?bbsSn=115131&key=m2101113055776&pageIndex=1&sc=&sw=

KISDI AI Outlook (Vol. 26, 2026): AI Full Stack Competition and Competition Policy Challenges  
 Focusing on AI Connectivity Infrastructure


KISDI: The AI market is shifting toward “full stack” competition
Priority should be given to establishing information-gathering and analytical systems to understand market changes


▲ The potential for anticompetitive effects should be examined both within individual layers and in combinations across layers
▲ Attention should be paid to the competitive, cooperative, and interdependent relationships between hyperscalers and telecommunications operators in AI connectivity infrastructure
▲ Given differences in the regulatory environments for international and domestic connectivity, priority should be placed on building systems for information collection and analysis

The Korea Information Society Development Institute (KISDI, President Sangkyu Rhee) recently published *AI Outlook* Vol. 26, titled *“AI Full Stack Competition and Competition Policy Challenges: Focusing on AI Connectivity Infrastructure.”*

The report notes that AI competition is shifting toward full stack competition, in which two or more layers of the AI ecosystem are closely linked through vertical integration or strategic alliances. Against this background, the report examines not only competition issues within a single layer, but also the potential for anticompetitive effects arising from combinations across layers, with particular attention to connectivity infrastructure.

Previous discussions of AI competition policy have focused heavily on AI computing infrastructure. However, in addition to computing capabilities, AI services also require connectivity infrastructure that links data, computing resources, services, and users in practice. In the international domain, hyperscalers have expanded international connectivity functions such as submarine cables, international backbone networks, and content delivery networks (CDNs). Within domestic markets, telecommunications operators hold fixed and mobile access networks, backbone networks, and points of contact with subscribers and business customers.

The report also emphasizes that these actors compete in some functions while depending on each other’s assets in others. Accordingly, it divides the AI ecosystem into five layers—semiconductors, computing, models and services, connectivity, and users and applications—and further subdivides the connectivity layer into international and domestic infrastructure.

The report explains that the conditions under which hyperscalers, which are strong in global computing, models, and international connectivity, combine with telecommunications operators, which possess domestic connectivity networks and customer interfaces, may affect competitive conditions in Korea’s AI market. In particular, combinations between telecommunications operators and cloud providers may, under certain conditions, strengthen market positions on both sides and raise barriers to entry. The report therefore notes the need to examine whether such combinations create the ability and incentive to restrict competition. At the same time, it adds that potential positive effects, including increased investment, risk sharing, and joint technological optimization, should also be taken into account.

The report identifies five key factors for assessing the competitive effects of full stack combinations:  
▲ the availability of access to and use of necessary functions
▲ the possibility of substitution or bypass through alternative suppliers, technologies, or routes
▲ the possibility of multi-homing and switching
▲ non-discrimination
▲ the appropriateness of using competitively sensitive information

It explains, however, that these factors should not be applied as automatic decision criteria. Rather, assessments should consider the ability and incentive to restrict access as well as actual market effects.

The report stresses the importance of securing information that makes it possible to continuously track the structure and evolution of full stack combinations. It states that, in addition to equity shares or investment amounts, it is necessary to identify contract duration, capacity commitments, technological optimization arrangements, and conditions for third-party access. It also notes the need to continuously examine how such combinations affect the emergence of new entrants and substitutes, as well as the range of available providers and technological combinations.

The report further emphasizes that connectivity infrastructure policy should also focus on maintaining both investment incentives and competitive possibilities. In the domestic market, specialized networks, private networks, or regional networks may serve as alternatives. Neutral hosts, edge infrastructure, and network APIs may also provide pathways for new entrants to participate in specific locations or functions. For international connectivity, the report explains that, from a competition perspective, it may be important to create an environment in which multiple international routes and global operators can connect to domestic data centers and telecommunications networks, and in which domestic providers and users can choose among different combinations.

With regard to the regulatory approach, the report points out that it would not be appropriate to decide in advance in favor of either ex ante regulation or ex post regulation. Ex post competition law makes it possible to confirm actual conduct and effects, but it may be difficult to restore competition once market structures have become entrenched. Ex ante measures, by contrast, may prevent entrenchment, but they also carry the risk of prematurely locking temporary bottlenecks or specific technologies into regulation. The report therefore proposes first establishing data and analytical systems capable of tracking market changes, and then using that foundation to distinguish areas requiring ex ante intervention from those that can be addressed through ex post competition law.

*KISDI AI Outlook* is available for download from the KISDI website (www.kisdi.re.kr).