NHN posted record-breaking revenue and operating profit in the second quarter of 2025, the best figures in the South Korean technology group's history. The driving force was explosive growth in GPU server leasing — a business within its cloud infrastructure division designed to capture surging demand for artificial intelligence computing. The results mark what many observers regard as an inflection point: a company once categorised as a diversified IT conglomerate built around gaming and digital payments is rapidly redefining itself as an AI infrastructure provider.
The numbers reflect more than incremental improvement; they signal a structural shift in where the business generates its weight. NHN Cloud, the group's cloud subsidiary, absorbed GPU resource demand from both the public and private sectors. South Korean generative AI start-ups and the AI research units of large conglomerates (known locally as chaebols) have increasingly preferred renting cloud capacity over building their own GPU infrastructure — a preference that funnelled contracts towards NHN Cloud, whose stable data-centre operations gave customers confidence. For NHN, this means the moment of reckoning has arrived: the heavy upfront capital investment it made in data-centre capacity is now converting into profit.
Market analysts describe NHN's expansion into GPU cloud as "a case where timing and positioning aligned." Globally, the three dominant hyperscalers — AWS, Microsoft Azure, and Google Cloud — command the lion's share of GPU supply. But South Korea's public sector is effectively ring-fenced from foreign providers by security certification requirements and mandates favouring domestically operated infrastructure. NHN Cloud holds the Cloud Security Assurance Programme (CSAP) certification issued by the Korean government, placing it in an enviable position to exploit that gap.
Domestic rivals are not standing still. KT Cloud and Naver Cloud are both accelerating GPU infrastructure build-outs. Naver is aggressively expanding hyperscale AI computing capacity at its Sejong data centre, while KT is targeting enterprise clients by leveraging its telecommunications backbone. NHN's counter-argument is differentiation through operational experience: its own services — spanning games, payments, and commerce — act as live proof-of-concept deployments for its cloud infrastructure, giving prospective clients a working reference rather than a sales pitch.
Similar patterns have emerged elsewhere. Japan's Sakura Internet leveraged government support and domestic AI demand to scale its GPU cloud business rapidly, achieving a recovery in both earnings and its share price. Taiwan's Chunghwa Telecom exploited its regulatory familiarity relative to global technology giants to capture AI demand from public-sector and financial clients. NHN's trajectory reads as the Korean iteration of this "local champion" playbook.
Yet the bull case is not without its caveats. The profitability of GPU cloud depends directly on the stable procurement of Nvidia chips, and global supply of high-performance processors such as the H100 and B200 continues to lag demand. Rising procurement costs remain a permanent threat to margins. There is also cyclical risk: if South Korea's AI investment boom moves past its peak and enters a consolidation phase, demand for GPU leasing could contract sharply. Some industry insiders caution that the current surge may be concentrated in the initial infrastructure build-out phase, and that growth could decelerate as corporate clients develop proprietary infrastructure or as demand stabilises.
NHN's gaming division and its fintech arm — anchored by the Payco payments platform — remain structurally stagnant. Despite the record-breaking headline figures, an excessive concentration of group-wide earnings in a single business line weakens the resilience of the overall portfolio. Analysts urging a closer reading of the quality, not just the quantity, of NHN's earnings have a point worth heeding.
The strategic implications looking ahead are reasonably clear. As AI workloads shift from the training phase towards inference — running models rather than building them — raw GPU accumulation will matter less than the ability to operate those resources efficiently. To sustain its current momentum, NHN will need to move beyond hardware supply and strengthen the software and platform layers that enable AI workload optimisation, cost-efficient operations, and customer lock-in. On the policy front, whether the South Korean government continues to direct its AI computing infrastructure support programmes towards domestic cloud providers will prove a pivotal variable for the sector's medium-term growth.
NHN's second-quarter results are not merely a reflection of a favourable market cycle. They represent a hypothesis tested and — for now — validated: that a domestic IT company can find a viable intersection of survival and growth at the infrastructure layer during the AI transition, provided it acts early, earns the right certifications, and occupies the space before the global giants can reach it.