The Hardware Sovereign: Why China is Listing Chipmakers Before Model Labs
In the global race for artificial intelligence dominance, the conventional wisdom has been to fund the model labs first, letting the compute infrastructure follow the demand. But in China, the capital markets are inverting this hierarchy. By prioritizing the public listing of AI chipmakers like Enflame, the Shanghai STAR Market is signaling that it views domestic silicon not merely as a commercial product, but as a sovereign strategic asset. This is a structural shift in how AI infrastructure is valued, moving from speculative venture bets to foundational national security requirements.
On September 11, 2026, Enflame (688801.SS) made its debut on the Shanghai STAR Market, raising 6.12 billion yuan (approximately $912 million). The market response was immediate and aggressive: the stock closed at roughly 435 yuan, a 179% surge over its 142.18 yuan offer price. This jump pushed the company’s market capitalization to approximately $28 billion, a significant leap from the $9.1 billion implied by its initial offer. Most telling, however, was the retail oversubscription rate of 6,109x. This is not a standard valuation metric; it is a raw demand signal from a market desperate for exposure to the domestic AI hardware supply chain.
Enflame is one of China’s four so-called GPU dragons, alongside Cambricon, MetaX, and Moore Threads. These firms are the primary architects of the country’s attempt to build a self-sufficient GPU ecosystem. While Enflame reported a net loss of 1.2 billion yuan in 2025, the company has guided for break-even or profitability by 2026-2027. The proceeds from this IPO are earmarked for the R&D and industrialization of its 5th and 6th generation AI training and inference chips, underscoring the capital-intensive nature of this sovereign buildout.
The STAR Market is becoming the epicenter of this strategy. It is the same exchange where DeepSeek, valued at $71 billion on secondary markets, plans to file for a $74 billion IPO in the second quarter of 2027. Crucially, Tencent serves as a lead strategic backer for both Enflame and DeepSeek. This dual-investment strategy reveals a clear intent: Tencent is hedging its bets by controlling both the model intelligence and the underlying silicon infrastructure, ensuring that the hardware layer is robust enough to support the software ambitions of the next generation of Chinese AI.
This rush to capitalize the hardware layer reflects a broader anxiety regarding compute access. As the industry grapples with the limitations of current hardware, the market is looking for alternatives to the NVIDIA-dominated status quo. This mirrors the broader trend of alternative silicon challenges, such as the $875 million bet on Positron AI, which seeks to disrupt the inference lock through commodity memory. Similarly, this hardware-first approach complements the Chinese open-weight compression thesis, where the goal is to maximize the utility of existing, potentially constrained, compute resources.
The 6,109x oversubscription for Enflame suggests that investors are not just buying into a chipmaker; they are buying into the infrastructure of a closed-loop AI economy. By listing the hardware layer before the model labs, the Chinese market is effectively de-risking the AI stack from the bottom up. If the model labs fail, the infrastructure remains a national asset. If they succeed, the infrastructure becomes the bottleneck that captures the value.
As we look toward the upcoming DeepSeek IPO, the message from the STAR Market is clear: in the era of sovereign AI, the most valuable real estate is not the model weights, but the silicon that runs them. The capital flow into Enflame is a definitive statement that for China, the compute landlord thesis is the only one that matters.