AI Compute Power Fuels Global Chip Stock Rally: Cambricon Hits All-Time High
Global Semiconductor and AI Chip Stocks Surge Collectively: Dual-Engine-Driven Reallocation of Compute Capital
On the final trading day of May, global tech capital markets witnessed a rare “counter-cyclical resonance”: amid Federal Reserve Chair Jerome Powell’s renewed hawkish signals and the 10-year U.S. Treasury yield approaching 4.5%, core AI compute assets—including Cambricon, SK Hynix, and Nasdaq-100 futures—rose sharply against the trend. Cambricon surged over 15% in a single day, briefly breaching the RMB 1,500 mark—a new all-time high since its IPO. Concurrently, SK Hynix’s share price hit a record peak, while Nasdaq-100 futures rose 1.4%, propelling China’s ChiNext Index up over 2% in a single session to a fresh historical high. This cross-market, cross-regional rally is no short-term sentiment-driven phenomenon. Rather, it reflects the powerful convergence of two structural forces: the hardening, tangible demand for AI compute capacity—and the deepening geopolitical dimension of technological competition. It signals that global tech capital is accelerating its strategic reallocation toward “high-certainty growth sectors.”
Compute Infrastructure Enters the Scale-Deployment Phase: From Model Iteration to Chip Commercialization
Over the past two years, large-model capability leaps have moved decisively from research labs into industrial applications. Today, leading global cloud providers and vertical-industry customers are rapidly deploying second-generation AI infrastructure. Microsoft Azure has added over 100,000 H100 GPUs to its cluster; Amazon Web Services (AWS) has launched Inf2 instances based on NVIDIA’s Blackwell architecture; and domestic “10,000-GPU cluster” projects have accelerated from planning into peak delivery. According to IDC’s latest forecast, global AI server shipments will grow 32% year-on-year in 2024—with servers equipped with domestically developed AI chips accounting for over 18% of the total for the first time. Cambricon’s explosive rally reflects market validation of the mass commercial rollout of its MLU370 chip across mission-critical sectors—including finance, government services, and telecom operators. The company’s Q1 report shows cloud-based intelligent computing chip orders doubled year-on-year, while edge AI chips achieved a 34% penetration rate in intelligent driving domain controllers. Surpassing RMB 1,500 is not merely a valuation re-rating—it marks a milestone pricing event signifying China’s high-end AI chips’ evolution from “functional” to “production-ready.”
Notably, compute expansion is no longer confined to the GPU-centric path. Cambricon’s MLU series employs an in-memory computing architecture, delivering 2.3× higher energy efficiency than NVIDIA’s A100 for specific inference workloads. Meanwhile, SK Hynix’s HBM3e memory has achieved a yield exceeding 85%, boosting effective bandwidth per GPU to 1.2 TB/s. Hardware co-optimization—spanning chips, memory, and interconnects—is now the new focal point of competitive differentiation. This full-stack upgrade is flattening the AI training cost curve, thereby accelerating AI-native application adoption across hundreds of industries. Capital markets’ sharp reaction is, at its core, an anticipatory pricing of the vast commercial opportunity unlocked by falling marginal compute costs.
Geopolitical Tech Competition Evolves: Rare-Earth Export Data Reveals Supply-Chain Resilience
As technological rivalry intensifies, the geopolitical dimension is shifting—from “sanction-based deterrence” toward “system-level confrontation.” In May, China’s exports of rare earths and related products reached 10,905 metric tons, surging 36.6% year-on-year; cumulative exports from January to May totaled 54,976 tons, up 14.9%. Behind these figures lies a global restructuring of supply chains for critical intermediates: rare-earth permanent magnets (used in high-end motors), polishing powders (for wafer fabrication), and sputtering targets (for chip manufacturing). While the U.S. tightens export controls on advanced semiconductor equipment destined for China, China is reinforcing upstream control through deep rare-earth processing capabilities—producing 90% of the world’s neodymium-iron-boron (NdFeB) magnets. Moreover, the high-purity cobalt-nickel sputtering targets essential to SK Hynix’s newly launched HBM3e chips rely heavily on China’s refined material production capacity.
At a deeper level, the correlated stock movements of Cambricon and SK Hynix reflect a rebalancing of East Asia’s semiconductor value chain. Cambricon provides AI chip design expertise; SK Hynix secures cutting-edge memory supply; and TSMC—though not explicitly named, implicitly central to this logic—delivers advanced-node foundry services. Together, they form the “iron triangle” of AI compute hardware. As the U.S.-Iran temporary agreement eases Middle Eastern geopolitical risks—and global shipping logistics and energy prices stabilize—the operational efficiency of this regional collaboration network improves. Although Hong Kong’s Hang Seng Tech Index fell 1.4%, Zhipu AI rose over 7% against the trend—highlighting how capital is increasingly concentrating on firms that truly command core technology nodes, rather than chasing broad tech themes.
Capital Reallocation Logic: Certainty-Driven Growth Replaces Rate-Sensitive Assets
With the Fed persistently emphasizing “higher for longer,” market expectations for rate cuts have significantly cooled. Against this backdrop, most tech stocks face pressure—yet the AI chip segment has charted an independent, upward course, driven fundamentally by markedly enhanced earnings visibility. Cambricon’s Q1 2024 revenue grew 112% year-on-year, with gross margin rising to 58.3%; SK Hynix’s HBM-related revenue accounted for 37% of total sales during the same period, contributing over 60% of gross profit. By contrast, traditional tech segments—including consumer electronics and SaaS software—continue grappling with soft demand and slowing subscription growth.
This divergence confirms a fundamental shift in capital allocation logic: amid rising macro uncertainty, investors are abandoning speculative “stories” in favor of assets anchored in “cash-flow visibility.” For AI chipmakers, order visibility has extended broadly to 12–18 months, with customer advance payments now reaching 30–40%—creating stable, REIT-like income characteristics. The ChiNext Index’s leadership and record highs reflect A-share markets’ localized response to this trend: semiconductors now account for 28% of total ChiNext trading volume—the highest level in three years—indicating systematic inflows of incremental capital into core hard-tech sectors.
Conclusion: Investment Paradigm Shift in the Era of Compute Sovereignty
Cambricon’s breakthrough past RMB 1,500, SK Hynix’s record highs, and Nasdaq futures’ strong rally may appear as mere price fluctuations—but they actually mirror a profound realignment in global technological power structures. As AI becomes a strategic pillar of national competitiveness, compute infrastructure transcends pure commercial investment to acquire dual economic and security attributes. Over the next six months, with the U.S. CHIPS Act subsidy guidelines finalized, China’s number of officially approved large models surpassing 200, and the global HBM4 standard formally ratified, competition over technical standards and supply-chain resilience will only intensify. Investors must move beyond single-company valuation frameworks and instead adopt “compute sovereignty” as their analytical coordinate system—identifying enterprises possessing irreplaceable capabilities at critical nodes: chip design, advanced packaging, specialty materials, and high-speed interconnects. This is not just the underlying logic of the current rally—it is the defining paradigm for technology investing over the next decade.