Summary of Key Points
Recently, South Korean prosecutors conducted surprise searches at the offices of three companies—Lanqi Technology (China), Renesas Electronics (Japan), and Rambus (USA)—in South Korea, suspecting them of colluding on prices in the Memory Interface Chip (MIC) market. This incident is part of a larger trend driven by the AI industry, which is reshaping the profits in the hardware supply chain. In the past, the benefits of AI were concentrated in NVIDIA's GPUs; now, the money is flowing towards the underlying infrastructure components such as MICs, HBMs, and advanced packaging technologies, which have become high-profit areas. At the same time, global regulators are closely monitoring these lucrative segments of the AI supply chain. As Chinese companies move to core positions in the global AI supply chain, they are also facing various regulatory challenges, including anti-monopoly investigations and intellectual property issues.
I. Why MICs? From a Secondary Role to a Critical Component in AI Servers
Memory Interface Chips (MICs) were once minor components in servers, but in the AI era, they have become indispensable. Simply put, MICs act as “data dispatchers” between GPUs and memory: training large AI models requires rapid data transfer between these two components, and if the MICs cannot keep up, the entire system will slow down.
- Market Explosion: Yole Group predicts that the global MIC market size will grow from $1.36 billion today to $14 billion by 2035, an increase of tenfold in ten years.
- Surging Profits: With the widespread adoption of DDR5 servers and the use of higher-bandwidth memory in AI servers, MICs have become one of the most profitable segments of the server supply chain.
- High Concentration: Only three companies—Lanqi, Renesas, and Rambus—dominate this market, and they are key partners of memory giants like Samsung and SK Hynix. This oligopolistic structure raises suspicions of price collusion.
II. The Shift in Wealth in the AI Era: From Software to Hardware
For the past two decades, internet companies have made money from software (e.g., WeChat, Google Search), with almost zero marginal costs—developing software once costs the same whether it is sold to one person or a million people. However, the AI era has changed:
- AI Companies as Heavy Industry: Training large models requires building data centers, which cost billions of dollars. Hardware components such as GPUs, HBMs, and MICs require substantial investment.
- Money Flowing Upstream: In 2023, there was a rush for NVIDIA GPUs; starting in 2025, hardware companies like SK Hynix (HBM), TSMC (advanced packaging), and Lanqi (MICs) will begin to share the profits. For example, SK Hynix reported record profits due to high demand for HBMs.
- Hardware Becoming Valuable: Hardware companies used to have thin margins, but now, AI demand is driving up prices, and their profit margins have surpassed those of many software companies.
III. Regulators Following the Profit Trail: Focusing on High-Profit Segments of the AI Supply Chain
Anti-monopoly investigations always target areas where profits are generated. Historically, Microsoft and Intel have been investigated for monopolies. Now that AI hardware is highly profitable, it has become a focus for regulators:
- South Korea Investigating MICs: The three companies control a large portion of the MIC market, and any price collusion could affect the purchasing costs of their customers, such as Samsung and SK Hynix, which would in turn be reflected in the prices of AI servers.
- The United States Investigating the Storage Chain: On the same day as the South Korean searches, the U.S. launched an investigation under Section 337 regarding DRAM (including HBMs), involving companies like Samsung and NVIDIA. The U.S. is also expanding export controls on AI chips, targeting the entire AI infrastructure.
- Clear Signal: Global regulators are no longer focusing on individual chips but on the most profitable segments of the supply chain—those that generate the most profit are more likely to be investigated.
IV. Opportunities and Challenges for Chinese Companies
Lanqi Technology’s investigation is not an isolated incident; it reflects the inevitable outcome of Chinese companies entering the core of the global AI supply chain:
- Opportunities: Chinese companies have already made inroads into segments such as AI servers, optical modules, PCBs, liquid cooling, and MICs. Lanqi’s position as one of the three leading MIC manufacturers globally demonstrates the recognition of their technical capabilities.
- Challenges: As Chinese companies gain a larger share in the global AI supply chain, they will face more regulatory challenges, including anti-monopoly investigations, intellectual property litigation, export controls, and national security reviews. This is similar to the path taken by the photovoltaic and new energy vehicle industries: as Chinese companies transition from followers to competitors, they encounter various “rule barriers.”
- Future Trends: Lanqi is just the beginning; more Chinese AI hardware companies will come under global regulatory scrutiny. This is both a challenge and an opportunity for growth, as it represents a hurdle that must be overcome to become core players in the global AI supply chain.
Conclusion
This investigation is not just about individual companies breaking the law but reflects a new phase in the development of the AI industry: the redistribution of profits from software to hardware, and the corresponding adjustment of global regulatory frameworks. For Chinese companies, this presents both opportunities (to share in the profits of AI hardware) and challenges (to comply with international regulations). However, it is a necessary step towards becoming core players in the global AI supply chain.