AI Compute Power Boom: Cambricon Hits $100B Market Cap and Reshapes China's Semiconductor Supply Chain
The AI Computing Power Industry Chain Erupts: A Triple Resonance of Hard-Tech Valuation Restructuring, Capital-Style Shift, and Global Supply Chain Repricing
Mid-2024 marks a systemic, quiet-but-unstoppable leap forward for China’s AI computing power industry chain. Cambricon’s market capitalization has historically breached RMB 1 trillion—making it the first “trillion-yuan hard-tech” benchmark on the STAR Market; the STAR 50 Index surged 26% in a single month, outperforming all other major indices; and most critically, TrendForce’s latest report confirms that demand for AI-related chips is persistently crowding out capacity on mature process nodes (28nm and above), pushing foundry pricing upward—not as a short-term negotiation tactic, but as a structural trend explicitly projected to extend through 2027. These three signals are not isolated events, but concrete manifestations of the accelerating industrialization of domestic AI computing power—signaling China’s semiconductor sector’s transition from “technology catch-up” to “value leadership.”
Behind the Market-Cap Milestone: A Paradigm Shift in Hard-Tech Valuation Logic
Cambricon’s trillion-yuan valuation appears, on the surface, to reflect concentrated market sentiment—but beneath lies a fundamental restructuring of A-share valuation frameworks for hard-tech enterprises. Historically, semiconductor firms were routinely evaluated using PE/PEG models borrowed from consumer or manufacturing sectors—emphasizing current earnings and cash flow. Yet Cambricon, as a “computing-power infrastructure” company focused on AI chip architecture design and IP licensing, derives its core value from three dimensions: the height of its technological moat, the breadth of its ecosystem compatibility, and the depth of its domestic substitution potential. In today’s valuation, roughly 40–50% already incorporates long-term penetration expectations for domestically developed large-model training chips, edge-side inference chips, and customized chips for intelligent computing centers. As CCTV.com noted, this shift reflects “sober strategic resolve”—precisely highlighting its essence: markets no longer fixate solely on Q1 financial results, but increasingly pay a premium for “technological sovereignty” and “computing infrastructure sovereignty.” This migration in valuation logic is compelling VC/PE investors in the primary market to reassess growth trajectories for chip-design firms—and prompting secondary-market investors to build multidimensional evaluation systems incorporating non-financial metrics such as technology roadmaps, customer adoption timelines, and wafer fabrication yield rates.
STAR 50’s Outperformance: Strategic Capital Reallocation Toward the “Physical Layer of Computing Power”
The STAR 50’s 26% monthly surge cannot be explained merely as thematic speculation. Semiconductor equipment makers (e.g., Advanced Micro-Fabrication Equipment Inc., NAURA), advanced packaging leaders (e.g., JCET), optical module suppliers (e.g., Innolight, New Bright Optoelectronics), and AI chip designers (e.g., Cambricon and its affiliates) collectively account for over 65% of the index’s weighting. Capital flows clearly target the “physical layer of computing power”—the foundational hardware clusters enabling large-model training and inference. Notably, during the same period, Hong Kong’s Hang Seng Tech Index rose only 1.8%, yet Huahong Semiconductor (a leader in mature-node foundry services) gained over 5%, while Zhipu AI (a large-model developer) surged more than 7%—confirming strengthened synergy across the “chip–foundry–model” value chain. This style shift exhibits strong sustainability: First, China’s “hundred-models race” has entered an intensely compute-intensive phase—training a single billion-parameter model can consume tens of thousands of NVIDIA A100/H100 GPUs, continuously releasing rigid demand for domestic alternatives. Second, tightening U.S. export controls on advanced-process equipment have paradoxically accelerated full-capacity operation and technical iteration at domestic mature-node fabs—creating a virtuous cycle of “policy pressure → capacity tightness → rising capital expenditure.”
Mature-Node Price Hikes Go Structural: The Starting Point of Global Supply Chain Repricing
TrendForce’s anchoring of the mature-node price hike cycle through 2027 rests on far more than short-term supply-demand mismatches. Data shows global utilization rates for sub-28nm mature-node foundry capacity reached 98.3% in Q2 2024—with orders for automotive-grade MCUs, CIS image sensors, power management ICs, and AI edge inference chips now accounting for 67% of total volume. Crucially, the price increase has formed a complete transmission chain: foundries raise prices → OSATs gain stronger bargaining power → equipment suppliers (e.g., ASMPT, Changchuan Technology) secure order visibility extending up to 18 months → materials vendors (e.g., Shanghai Silicon Industry Corp., Anji Technology) obtain long-term framework agreements. This means mature nodes are no longer merely “transitional technologies,” but have become core enablers of AI democratization (at the endpoint and edge) and industrial intelligence (in automotive electronics and energy control). Global giants—including GlobalFoundries and UMC—have announced over $10 billion in mature-node capacity expansions over the next three years; Chinese wafer fabs are following suit: SMIC’s Shaoxing Phase II and Yuechi Semiconductor’s Phase III both explicitly prioritize specialized 28nm/40nm processes. The long-term institutionalization of pricing mechanisms reflects a strategic reorientation of the global semiconductor value chain—from “cutting-edge performance” toward “reliable supply” and “system-level integration.”
Deepening Industrial Synergy: From Single-Point Breakthroughs to End-to-End Ecosystem Closure
Behind these three phenomena lies the rapid closure of China’s domestic AI computing-power ecosystem. Cambricon’s MLU series chips have been integrated with leading large models—including Baidu’s ERNIE and Alibaba’s Tongyi—and deployed at scale across multiple intelligent computing centers nationwide. Within the STAR 50, equipment manufacturers and foundries are deeply aligned: NAURA’s etching tools have achieved over 95% validation pass rates on SMIC production lines, accelerating domestic equipment adoption. Meanwhile, improved profitability from mature-node price hikes is feeding back into wafer fabs’ procurement of more equipment and talent investment. Equally significant is policy coordination: China’s National Medical Products Administration recently issued guidance principles for classifying brain–computer interface medical devices—a niche domain, yet one signaling accelerated regulatory compliance for vertical AI applications (e.g., AI+healthcare), thereby expanding end-use scenarios for AI chips. Ministry of Finance data reveals that profits of state-owned enterprises rose 3.5% year-on-year—even as total revenue edged down slightly—highlighting tangible quality-and-efficiency gains, with integrated circuit manufacturing and R&D contributing notably.
The eruption of the AI computing-power industry chain is, at its core, a systemic upgrade jointly driven by technological breakthroughs, market selection, and policy guidance. Cambricon’s trillion-yuan market cap is not an endpoint—but rather the starting point for domestic computing power to evolve from “functionally usable” to “highly reliable” and “trusted for mission-critical use.” The STAR 50’s leadership is not speculative froth—but rational, value-driven capital reallocation toward hard-tech’s demonstrable fundamentals. And mature-node price hikes extending through 2027 are not transient noise—but the opening chapter of a fundamental reshaping of global semiconductor division of labor. As computing power becomes the new “water, electricity, and coal” of the digital age, China’s industrial chain stands at a historic inflection point: shifting from scale advantage to value leadership. The true test lies not in peak market valuations—but in how effectively that trillion-yuan valuation solidifies into a self-reinforcing, continuously evolving technological moat; how that 26% rally translates into tangible, scalable industrial momentum; and how every pricing cycle strengthens the foundational bedrock of domestic substitution.