Physix FrontierGigaDevice Net Profit Soars 10x, But $2B Came from Stock Trading
As a solution architect who has long followed AI infrastructure, I am accustomed to evaluating semiconductor companies through two dimensions: "customer willingness to pay" and "technical feasibility." GigaDevice's main business—NOR Flash, NAND Flash, and MCU—is exactly the key components for IoT, automotive electronics, and AI terminal devices. From a commercial value perspective, the rising cycle of memory chips has indeed brought it real profits: global supply chain tensions, production cuts by giants like Micron, plus domestic substitution demand, have significantly boosted customer willingness to pay. Regarding technical feasibility, GigaDevice has established a foothold in 55nm NOR Flash and recently broke through to 38nm. These products don't require the most advanced processes but win with high yield rates, controllable costs, and short customer validation cycles, making implementation relatively easier.
However, making 2 billion RMB from stock trading adds a layer of fog to the originally clear growth logic. This 2 billion comes from fair value changes in stocks held in SMIC, Beijing Junzheng, etc. As a solution architect, I focus on whether the management of this capital is sustainable—if the company views this unrealized gain as "capital ammunition" for R&D or M&A, its value far exceeds the book number; if it's just passive returns from financial investments, it's essentially luck in betting on the right cycle, unrelated to the competitive moat of the core business.
From a technical feasibility standpoint, GigaDevice's core moat lies in building a "Memory + Control" integrated ecosystem. Solutions pairing memory chips (NOR/NAND) with MCUs (general microcontrollers) allow device manufacturers to save on an independent control chip, reducing BOM costs and design complexity. This particularly appeals to cost-sensitive customers in scenarios like smart homes, TWS earbuds, and wearables. In several AI edge computing projects I participated in, clients explicitly requested this integrated solution because it reduces PCB area and power consumption, showing quite high willingness to pay.
However, the difficulty in implementation is that this integrated solution requires extremely strong system-level design and software adaptation capabilities. GigaDevice's market share in the MCU field is still behind ST and NXP, but its advantage lies in binding with the domestic chip ecosystem, especially pre-certification with Alibaba Cloud and Tencent Cloud IoT platforms. If it continues to increase R&D investment (e.g., developing custom MCUs based on RISC-V), it can convert the money earned from stock trading into technical barriers. Otherwise, the 2 billion bonus will turn into a burden in the next earnings report.
Another hidden risk is the strong cyclical nature of memory chips. Since 2021, NOR Flash prices have remained high. Recently, Samsung and SK Hynix started lowering DRAM and NAND prices, which is usually a signal of industry reversal. Although GigaDevice's main products still face tight supply-demand, once downstream customers see price drop expectations, they will reduce inventory, putting pressure on gross margins. Stock trading gains might smooth out one or two reporting periods but cannot reverse the cycle—unless the company liquidates financial assets to invest in long-term assets.
Returning to the detail mentioned at the beginning. I tend to understand the 2 billion stock trading gain as a form of "strategic investment": As a core member of the National Big Fund for domestic chips, GigaDevice holds stocks in upstream and downstream industry players, intended for industrial synergy, not short-term speculation. But if the gains come entirely from stock price fluctuations rather than synergies from industrial integration, its sustainability is questionable. Just as when designing AI solutions for clients, if a client spends the majority of their budget on speculating server futures, I wouldn't approve of their strategic direction—no matter how good the technical feasibility, it must be realized through solid product delivery.
Trend Prediction: Within the next 12 months, China's memory chip industry will enter an integration period of "separating the true from the false." GigaDevice's stock price will regress from "cycle premium" to "technology premium," and the contribution of stock trading gains will gradually narrow (the market will redefine valuation P/E ratios). If the company can convert this 2-billion-level capital gain into R&D investment for Computing-in-Memory and RISC-V MCUs, it may dominate the AIoT edge-side chip market; conversely, it will be pulled back into competing with DaPu Tong and Winbond Electronics—by then, the money made from stock trading will truly just have been a stroke of luck.
Original Link: https://www.tmtpost.com/8060632.html
