Cambricon's Q1 Net Profit Soars 185%: China's Domestic AI Chips Enter Commercial Monetization Phase

Explosive Performance and Capital Exodus: The Profound Transformation Behind Cambricon’s Q1 Turning Point

In Q1 2024, Cambricon delivered a financial report that stunned the market: net profit attributable to shareholders reached RMB 1.013 billion—a staggering 185% year-on-year increase. This not only marks Cambricon’s highest quarterly profit since its IPO but also represents the first time a Chinese AI chip design company has achieved scalable, sustainable positive cash flow. On the surface, this is a triumph of technological advancement; beneath the surface, however, lies a quiet but decisive divergence between industrial logic and capital logic at a critical inflection point: coinciding with the earnings release, Zhang Jianping’s investment group—long ranked among Cambricon’s top-ten shareholders and widely regarded by the market as an “industrial catalyst”—quietly exited the shareholder register. This juxtaposition of entry and exit signals that China’s domestic AI chip industry has moved beyond the “proof-of-concept” phase of “import substitution” into a new era of commercial realization—while the capital pathways underpinning its development are undergoing structural reconfiguration.

Technology Realized: A Hard-Core Leap from Lab to Data Center

Cambricon’s dramatic performance surge is neither accidental nor the result of accounting embellishment. Its core driver lies in the large-scale deployment of its MLU (Machine Learning Unit) series AI chips across intelligent computing (AI compute) data centers. According to company disclosures, orders from cloud service providers and major internet enterprises surged over 220% year-on-year in Q1 2024—with thousand-GPU inference clusters powered by the MLU370 already delivered and operating stably at three leading cloud service providers. Notably, these orders are not early-stage “policy-driven procurements,” but rather stem from genuine, mission-critical computational demands—such as large-model API calls and AIGC content generation. Third-party analysis shows that Cambricon’s chips deliver 92% of the watt-per-computation cost-effectiveness of NVIDIA’s A10 GPU for mainstream Chinese large-language model inference tasks—and even surpass it in certain structured workloads. This signifies that domestic AI chips have crossed the threshold of “functional usability” and entered the deep waters of commercial viability—where they are not just “usable,” but demonstrably “high-performing and cost-efficient.” Sustained convergence of technical specifications, coupled with China’s non-negotiable imperative for secure, sovereign computing supply chains, collectively form the foundational pillars behind this performance breakout.

Capital Divergence: Two Temporal Perspectives Behind Zhang Jianping’s Exit

Yet while the industrial sector celebrates technology’s real-world adoption, the capital markets respond with sober rationality. As an early key financial investor and strategic collaborator, Zhang Jianping—through his Pan Gu Fund and other affiliated entities—completed a full divestment ahead of the earnings release, exiting the top-ten shareholder list entirely. This move must be viewed through a longer temporal lens: Zhang’s group had been deeply involved since Cambricon’s pre-IPO round in 2018, betting on the grand narrative of “Can China build its own autonomous AI chips?” Today, with technical feasibility repeatedly validated and initial commercial pathways proven, his investment logic naturally shifts toward “value realization.” This is not a repudiation of Cambricon’s prospects, but rather the hallmark of financial capital’s “stage-based arbitrage” strategy—locking in returns at the dawn of technological maturity to redeploy funds into earlier-stage, higher-risk hard-tech ventures. In sharp contrast, industrial investors—including the National Integrated Circuit Industry Investment Fund (“Big Fund”) and CAS Holdings (affiliated with the Chinese Academy of Sciences)—have maintained or even increased their stakes. Their time horizons diverge fundamentally: the former measures returns over a 3–5-year cycle, while the latter anchors its vision to ecosystem building and standard-setting over a 10+ year horizon. This bifurcation signals that China’s domestic semiconductor investment has evolved beyond “betting on赛道 (race tracks)” to a new stage of “selecting racers + running alongside them.”

Benchmark Significance: Reshaping Semiconductor Valuation and Primary-Market Expectations

Zhang Jianping’s exit carries signal value far exceeding the scope of a single company. First, it is compelling a fundamental reconstruction of valuation logic across A-share semiconductor stocks. Historically, the market granted extremely high PS (price-to-sales) multiples to AI chip firms boasting high R&D intensity and low gross margins—implicitly assuming that “technological leadership = future monopoly.” Cambricon’s explosive profitability, however, stems precisely from optimized customer mix (government and enterprise orders now account for less than 35%) and cost reduction via product iteration (MLU370 yield improved to 82%), proving that commercial execution—not just technical prowess—is the true engine for navigating economic cycles. Secondary markets are rapidly shifting from “story-based valuation” to discounted cash-flow (DCF) models. Second, for the primary market, this event heralds a significantly accelerated exit rhythm. While domestic semiconductor IPOs declined 47% year-on-year in 2023, M&A exits rose 31%. Cambricon’s profitability blueprint offers downstream IDMs, equipment suppliers, and even materials firms a clear “acquisition-value anchor”: as import substitution evolves from “Do we have it?” to “Is it good enough?”, horizontal consolidation across the industry and vertical integration along the value chain will become the principal channels for capital realization. Finally, capital flows are quietly migrating: Wind data shows that VC/PE funding directed toward AI chip design fell 19% quarter-on-quarter in Q1 2024, while investments in next-generation technology roadmaps—including Chiplet advanced packaging and compute-in-memory architectures—rose 44%. Capital is proactively hedging against the commercialization risks of mature technologies by front-loading bets on frontier innovations.

Challenges Remain: Commercialization ≠ Risk-Free Zone

We must remain clear-eyed: this performance surge is merely a milestone—not the finish line—in the long march of domestic AI chips. Significant challenges persist. First, international giants are accelerating price-based countermeasures. Though NVIDIA’s H20 is subject to export restrictions, its successor—the L20—has already commenced small-scale testing in China, with an aggressively competitive pricing strategy. Second, ecosystem barriers remain the most critical soft constraint. Although Cambricon has adapted its platform to over 100 models, its developer community’s activity level stands at just one-seventh that of NVIDIA’s CUDA ecosystem, and toolchain usability gaps will require another two to three years to close. Moreover, geopolitical uncertainty remains unresolved: although the U.S. Department of Commerce’s latest Entity List update did not add Cambricon, it explicitly identifies “chips used for training generative AI” as a focal point for enhanced scrutiny. This means the “commercialization phase” of domestic substitution is, in essence, a high-pressure race—where success requires outpacing both internal technological iteration and external escalation of containment measures.

Cambricon’s quarterly report functions as a prism: it refracts China’s historic leap—from “technologically feasible” to “commercially viable” AI chips—and simultaneously reflects the divergent choices made by industrial and financial capital at a pivotal historical juncture. As figures like Zhang Jianping choose to step aside and realize gains, the real test has only just begun: How do we transform isolated technological breakthroughs into an impregnable, self-sustaining industrial ecosystem moat? This is no longer a question confined to laboratories—it is the ultimate examination determining the depth of China’s foundation in the intelligent age.