Naver vs Kakao: Korea’s Two Digital Empires Compared

South Korea’s digital economy runs through two companies. Naver, the country’s dominant search engine, posted KRW 10.7 trillion in revenue in 2024 — the first time a domestic internet company crossed that threshold. Kakao, the company behind KakaoTalk’s 48.9 million monthly active users, followed with KRW 8.1 trillion in 2025 after a bruising year of legal trouble and margin compression. Together, they control more of South Korea’s online life than any comparable duopoly in Asia: search, messaging, payments, commerce, content, cloud and now artificial intelligence. Yet they’ve arrived at this point from opposite directions, and the strategic choices each is making in 2025 and 2026 — particularly around agentic AI — will determine which platform defines the next decade.

How big is each company?

Naver is the larger business by almost every financial measure. Its 2025 revenue hit KRW 12.04 trillion, up 12% year-on-year, with a market capitalisation hovering around USD 22.5 billion as of late 2025. The growth engine is diversified: search platform revenue reached KRW 3.95 trillion in 2024 (up 9.9%), commerce hit KRW 2.92 trillion (up 14.8%), and cloud revenue grew 26.1% to KRW 563.7 billion. Naver also controls Naver Webtoon, which had 180 million monthly active users globally as of late 2023, and holds a majority stake in LINE, Japan’s largest messaging app with roughly 97 million users.

Kakao’s 2025 revenue came in at KRW 8.1 trillion, a modest 3% increase, but operating profit surged 48% to KRW 732 billion — a sign that the company’s cost discipline is finally biting. Its market cap sat around USD 15-18 billion through 2025, well below Naver’s. Kakao’s real leverage isn’t in the parent company’s top line, though. It’s in the constellation of subsidiaries: Kakao Pay (46.5 million users, KRW 167.3 trillion in Q4 2024 transaction volume), Kakao Bank, Kakao Mobility, Kakao Entertainment and Kakao Games. The sum of the parts has always been Kakao’s pitch to investors — and its curse, since the holding structure makes consolidated value harder to unlock.

What’s the fundamental difference in their models?

Naver is a search-and-commerce company that built outward. Its origin is the green search bar — still the default homepage for most Korean internet users — and everything it’s added since (shopping, payments, maps, reservations, cloud) funnels back to that starting point. Naver’s 44 million monthly active users in South Korea come with high commercial intent. When someone opens Naver, they’re looking for something: a restaurant, a product, an answer. That makes Naver’s ad inventory exceptionally valuable and its commerce flywheel — Smart Store, Naver Shopping, Naver Pay — tightly integrated.

Kakao is a messaging company that built inward. KakaoTalk isn’t just dominant in South Korea; it’s practically a utility, with 97% penetration and 48.9 million MAUs in a country of 51 million. The app is the default channel for family group chats, work coordination, government notifications and gifting. Kakao’s strategy has been to layer services — payments, mobility, banking, music, webtoons — on top of that captive social graph. It’s not a search engine with a messaging app bolted on; it’s a super-app that happens to compete with a search engine.

The distinction matters because it shapes how each company monetises attention. Naver captures demand. Kakao captures time.

Where do they compete head-to-head?

Payments are the fiercest battleground. Naver Pay has over 30 million users and logged KRW 22.7 trillion in transactions in Q3 2025 alone, powered by tight integration with Naver Shopping and Smart Store. Kakao Pay’s 46.5 million users give it a larger base, and its KRW 167.3 trillion in Q4 2024 payment volume reflects the sheer breadth of its use cases — ride-hailing, parking, insurance, peer-to-peer transfers. South Korea’s mobile payment market is projected to grow from USD 44.4 billion in 2025 to USD 73.3 billion by 2031, so neither company can afford to cede ground.

Commerce tilts toward Naver, which operates Smart Store (one of South Korea’s largest e-commerce platforms) and benefits from search-driven product discovery. Kakao’s commerce efforts are more fragmented, spread across Kakao Shopping, gifting via KakaoTalk and various subsidiary integrations. Coupang remains the 800-pound gorilla both are trying to outflank.

Content is a draw, though the portfolios differ. Naver owns Webtoon, a globally scaled IP machine. Kakao Entertainment controls Kakao Webtoon (formerly Tapas and Radish) plus a music and K-drama portfolio anchored by its controversial acquisition of SM Entertainment. Both companies see content as the upstream feeder for downstream commerce and advertising.

Who’s winning the AI race in Korea?

This is where the duopoly’s divergence gets genuinely interesting. Naver is building its own models. Kakao is buying access to someone else’s.

Naver’s HyperCLOVA X is South Korea’s most advanced homegrown large language model. The company has invested over KRW 1 trillion in GPU capacity and AI infrastructure, released open-source models ranging from 0.5B to 32B parameters, and in late 2025 unveiled two omni-modal models — HyperCLOVA X SEED Think (a reasoning model) and a Native Omni Model that processes text, images and voice simultaneously. By early 2026, Naver had replaced all Chinese-developed vision encoders with proprietary ones trained natively in Korean, a move with both technical and geopolitical significance. The company’s Agent N, launched in November 2025, connects search, maps and reservations into an AI-driven workflow, with an AI shopping agent following in Q1 2026 and an AI-powered search tab in Q2.

Kakao took a different path. In February 2025, it became the first Korean company to forge a strategic partnership with OpenAI, and by October had embedded GPT-5 directly into KakaoTalk via ChatGPT for Kakao. Users can access ChatGPT within their messaging threads alongside Kakao’s affiliated mobility, payments, music and gifting services. In parallel, Kakao is developing Kanana, a proprietary on-device model that reads conversation context and offers proactive recommendations. Kanana-2, with improved tool-calling and instruction-following, is slated for full deployment in KakaoTalk in the first half of 2026. The company has also partnered with Google for device-level AI experiences, creating a deliberate two-track strategy: OpenAI for the large model, Google for the device layer, Kanana for the proprietary glue in between.

The philosophical split is stark. Naver’s betting that sovereign AI — models trained on Korean data, optimised for Korean language, hosted on Korean cloud infrastructure — will be a durable competitive advantage. Kakao’s betting that distribution trumps model ownership, and that 48 million KakaoTalk users interacting with the world’s best models will generate more value than building a model from scratch.

What are the risks for each?

Naver’s biggest vulnerability is the global AI platforms eating into search. If Korean users start going to ChatGPT or Perplexity for answers instead of Naver’s search bar, the entire commerce and advertising flywheel loses its starting point. Naver’s heavy capital expenditure on AI infrastructure — over KRW 1 trillion and counting — also compresses margins in the near term with no guarantee of proportional returns.

Kakao’s risks are more varied and more immediate. Founder Kim Beom-su was arrested in 2024 over alleged stock manipulation during the SM Entertainment acquisition. Prosecutors sought a 15-year prison term, though a Seoul court acquitted him in October 2025. The reputational damage and regulatory scrutiny haven’t fully dissipated; financial regulators could still compel Kakao to divest its stake in Kakao Bank if any senior executives are found in violation of the Capital Markets Act. Beyond governance, Kakao’s dependence on OpenAI for its core AI experience means it doesn’t control its most important new feature. If OpenAI raises prices, changes terms, or prioritises a competing partner, Kakao’s AI strategy gets rewritten overnight.

What’s the outlook?

The short answer: Naver has more strategic coherence; Kakao has more surface area. Naver’s vertically integrated approach — own the model, own the cloud, own the search bar, own the commerce — gives it clearer unit economics and a defensible AI moat. Kakao’s horizontal sprawl across messaging, fintech, mobility, entertainment and gaming gives it more touchpoints with Korean consumers but makes execution harder and governance messier.

Both companies are targeting 2026 as the year agentic AI goes mainstream in their ecosystems. Naver’s AI shopping agent and AI search tab will test whether sovereign models can match the capabilities of frontier US models on real consumer tasks. Kakao’s Kanana-in-KakaoTalk will test whether an AI assistant embedded in a messaging super-app can become the default interface for daily life in South Korea. Whichever thesis proves right, it won’t be a winner-take-all outcome. South Korea’s 51 million consumers are too valuable, and both platforms too entrenched, for either to disappear. The real question is which company captures the higher-margin layer of the stack — and right now, that looks like it’ll be decided by who builds the better AI agent, not the better app.