Chinese Chipmaker SMIC Reports Triple-Digit Profit Surge and Raises Prices on Domestic AI Demand

The popular imagination of the artificial intelligence boom is dominated by a single piece of hardware: the cutting-edge graphics processing unit (GPU). When investors and policymakers picture the physical reality of AI, they envision the ultra-advanced, nanometer-scale logic chips designed by Nvidia and fabricated in Taiwan. But that mental model misses the vast, unglamorous machinery required to keep those GPUs running. A server rack is just a costly box of heat and ambition without the power-management integrated circuits, logic controllers, and embedded memory that surround the headline processors.

The evidence for this hidden bottleneck arrived this week from an unexpected source. Semiconductor Manufacturing International Corporation (SMIC), China's largest contract chipmaker, reported a staggering 261.7 percent year-on-year surge in second-quarter net profit, reaching $479.2 million. Revenue topped $3 billion for the first time, defying expectations that the foundry would struggle under the weight of strict U.S. export controls.[1][3]

The windfall was not driven by breakthroughs in advanced lithography. Instead, SMIC's factories are running at near-maximum capacity—hitting 93.7 percent utilization—to churn out "mature node" components. These are the older, established semiconductor processes that handle power delivery, connectivity, and basic logic. As tech giants and startups race to build out AI data centers, the sheer volume of supporting chips required has overwhelmed existing global capacity, creating a lucrative opening for foundries willing to supply the less prestigious parts of the stack.[3][4]

The mechanism here is straightforward but often overlooked. An AI server draws exponentially more power than a traditional data center rack, requiring highly specialized power-management ICs and BCD (Bipolar-CMOS-DMOS) chips to regulate voltage and prevent catastrophic overheating. While industry leaders like TSMC and Samsung have spent the last year reallocating their 8-inch wafer capacity toward more profitable advanced nodes, they inadvertently created a vacuum at the lower end of the market.[2]

Chinese foundries have aggressively stepped into that void. SMIC shipped 2.9 million 8-inch-equivalent wafers in the second quarter, a 14 percent sequential increase. More importantly, the severe supply-demand imbalance has handed the company unexpected pricing power. Co-CEO Zhao Haijun confirmed that SMIC successfully raised prices following negotiations with customers earlier this year, driving a 5.7 percent increase in average selling prices and pushing gross margins to an impressive 25.3 percent.[3][4][6]

The price hikes are reportedly concentrated in the mature-node production lines that serve the AI sector, with industry analysts noting increases of roughly 10 percent for specific components. This dynamic is not isolated to SMIC; its smaller domestic rival, Hua Hong Grace Semiconductor, reported a 385.9 percent profit jump over the same period, citing similar momentum in AI-driven demand for logic and analog integrated circuits.[1][2]

For Beijing, the financial results offer a rare bright spot in a semiconductor landscape heavily constrained by Washington's technology embargoes. Because mature nodes do not require the extreme ultraviolet (EUV) lithography machines that are currently blocked from entering China, SMIC and Hua Hong have been able to expand capacity using readily available equipment. The domestic market, eager to secure reliable supply chains amid geopolitical uncertainty, accounted for 90 percent of SMIC's second-quarter revenue.[4][6]

Yet, significant uncertainties remain about the durability of this profit surge. SMIC is currently benefiting from a unique convergence of global capacity shifts and intense domestic stockpiling. The company has committed to massive capital expenditures—$3.4 billion in the first half of the year alone—to build new fabrication plants. As those facilities come online, the resulting depreciation charges will weigh heavily on future margins, testing whether the current pricing power is a permanent structural shift or a temporary cyclical squeeze.[4][6]

Furthermore, success in mature nodes does not solve China's broader strategic challenge. While SMIC can profitably supply the power chips and controllers that surround an AI accelerator, it remains largely cut off from the tools needed to manufacture the accelerators themselves at scale. The company's financial windfall proves that China has secured a vital, highly profitable position in the global AI supply chain, but it does not mean the country has closed the gap at the bleeding edge of semiconductor logic.[2]