Naver has announced that its next-generation AI, HyperCLOVA X, will compete on efficiency rather than raw power.
The battlefield for AI dominance is changing. For a long time, the race was about creating the biggest, most powerful model that could top performance charts. However, a new front has opened up, focused on practicality and affordability. Chinese AI models like DeepSeek have surged in popularity, capturing over 60% of the market on some platforms. They aren't necessarily the 'smartest,' but they are incredibly fast and cheap, making them the pragmatic choice for many everyday business tasks. This signals a market shift where token-per-performance, or getting the most bang for your buck, is becoming the most important metric.
This shift was accelerated by two key events in the high-end AI market. First, OpenAI doubled the price of its latest model, GPT-5.5, which significantly increased costs for developers using it at scale. Second, Anthropic, another leading US AI company, abruptly suspended global access to its top models, Claude Fable 5 and Mythos 5, under a US government directive. This created both a price shock and a supply crisis, pushing customers to urgently seek reliable and affordable alternatives.
Naver has been strategically preparing for this moment. The company has invested heavily in infrastructure, completing a massive 4,000-GPU cluster with NVIDIA's latest Blackwell chips. Furthermore, it has deepened its partnership with NVIDIA to build a gigawatt-scale 'AI factory' and joined the Nemotron Coalition, a group focused on developing open and efficient AI models. These moves provide Naver with the computing power and technical foundation needed to develop and serve high-quality AI at a lower cost.
Ultimately, Naver is making a calculated bet on the rise of Sovereign AI. The company believes that businesses and governments, especially in Korea and Japan, will increasingly demand AI solutions that are not only cost-effective but also secure, keeping sensitive data within national borders. By targeting this growing need with an efficient model built on a robust domestic infrastructure, Naver is positioning itself not just as a competitor, but as a crucial alternative in a volatile global AI landscape.
- Glossary
- Token-per-performance: A measure of AI cost-effectiveness. It reflects how much value (speed, quality) you get for each unit of text ('token') you process, similar to 'miles per gallon' for a car.
- Sovereign AI: An AI system developed and operated within a country's borders, ensuring data security and independence from foreign control. This is critical for government and sensitive enterprise data.
- MoE (Mixture of Experts): An efficient AI architecture that uses multiple specialized sub-models ('experts'). Instead of one giant model running for every task, it only activates the relevant experts, saving significant time and money.