Key takeaways
- SMIC said on Aug 14, 2026 that strong AI-related demand was supporting orders and that it had increased prices for some capacity.
- Its second-quarter revenue reached about $3.01 billion, while utilization rose to 93.7% and average wafer selling prices increased sequentially.
- The important signal is broader than leading-edge AI processors: demand is also tightening capacity for power-management, connectivity and other supporting chips.
- Higher revenue does not automatically mean every chip category is in shortage, and one quarter should not be treated as a permanent trend.
Semiconductor Manufacturing International Corporation, better known as SMIC, has given the global chip industry a useful signal about the second-order effects of the artificial-intelligence buildout. The company said on Aug 14, 2026 that AI-related demand continued to support customer orders and that it had raised prices for its most constrained manufacturing capacity.
The announcement followed a strong second quarter. SMIC reported revenue of approximately $3.01 billion, a 20% increase from the previous quarter. Gross margin improved to 25.3%, wafer shipments rose, and the company’s capacity-utilization rate reached 93.7%. Profit attributable to shareholders was $479.2 million, although management also identified a one-time gain that helped the result.
The headline is not simply that one Chinese foundry earned more money. The more important development is that demand related to AI systems appears to be spreading beyond the processors that receive most public attention. Servers and data centres require memory, power-management components, networking equipment, sensors, controllers and other supporting semiconductors. When demand for that wider system rises faster than available capacity, price pressure can appear across parts of the supply chain.
What SMIC actually reported
SMIC’s Hong Kong exchange filing reported second-quarter revenue of $3.0056 billion, up from $2.5055 billion in the first quarter. The company attributed the sequential growth mainly to higher wafer shipments, a higher average selling price and a change in product mix.
The company shipped roughly 2.9 million eight-inch-equivalent wafers during the quarter, about 14% more than in the previous quarter. Average selling price increased by 5.7% sequentially. Monthly production capacity, expressed on an eight-inch-equivalent basis, reached approximately 1.1 million wafers, while utilization climbed from 89.6% to 93.7%.
Management guided for third-quarter revenue to increase by 2% to 4% sequentially and expected gross margin of 26% to 28%. These are company forecasts, not guaranteed outcomes. They can change with customer demand, equipment availability, product mix, export controls and broader economic conditions.
Why AI demand affects more than AI accelerators
Public discussion often treats the AI-chip market as a contest among advanced graphics processors and custom accelerators. Those components are important, but they are only one layer of an operational AI system.
A data centre also needs chips that regulate power, move data, connect storage, control cooling equipment and coordinate servers. Consumer devices incorporating AI features need radio-frequency, display, image-processing, connectivity and power-management components. Some of these chips use mature or specialty manufacturing processes rather than the most advanced process node.
That distinction helps explain why AI investment can benefit a foundry even when it is not the primary producer of the world’s most advanced accelerators. A growing AI market can increase demand for peripheral and supporting chips across many process technologies.
The four-layer AI semiconductor demand map
The following framework separates the parts of the supply chain that are often collapsed into the phrase “AI chips.”
Layer Typical components Demand driver Main constraint
Compute GPUs, AI accelerators and CPUs Model training and inference Advanced-node capacity and packaging
Memory and storage High-bandwidth memory, DRAM and enterprise storage Larger models and data movement Memory production and integration
Connectivity Network processors, optical components and interface chips Moving data between servers Bandwidth, power and component supply
Power and control Power-management ICs, controllers and sensors Running and cooling dense systems Mature-node and specialty capacity
SMIC’s update is most useful as evidence that demand is reaching the lower two layers as well as headline compute products. It does not prove that every component in those layers is scarce.
Why prices are rising
SMIC said price negotiations took place earlier and the new pricing applied to wafers processed during the third quarter. Pricing in semiconductor manufacturing depends on process technology, wafer size, order volume, contract terms, yield requirements and how urgently customers need capacity.
When utilization approaches the practical limits of a factory, a foundry has less unused capacity available for unexpected orders. Customers may be willing to pay more to secure production, especially when delaying a component could hold up a much larger finished product.
Price increases can also reflect a richer product mix rather than a simple across-the-board increase. That is why the reported 5.7% rise in average selling price should not be interpreted as a universal price increase for every SMIC customer.
Capacity remains the central question
High utilization is positive for a manufacturer because fixed costs are spread across more output. It can also expose operational pressure. Semiconductor factories require complex equipment, long qualification cycles and consistent yields. New machinery does not become productive capacity immediately.
SMIC said it added 8,000 twelve-inch wafers of monthly capacity during the second quarter and planned to keep adjusting existing capacity while accelerating new production lines. The company reported roughly $3.4 billion of capital expenditure during the first half of 2026.
The practical question is whether new qualified capacity can arrive before customer demand changes. Too little capacity can create missed orders and higher prices; too much capacity can depress utilization and margins later.
What the result means for technology buyers
Companies buying servers, networking hardware or industrial systems should not assume that a foundry price increase will immediately appear as an identical increase in finished-product prices. The semiconductor may represent only part of the product’s total cost, and suppliers may absorb, delay or offset increases.
However, buyers can use the update as a reason to review four areas:
- Which critical products depend on a small number of chip suppliers?
- Do purchase agreements specify lead times, allocation rules and substitution options?
- Can approved alternative components be qualified before a shortage occurs?
- Are forecasts based on actual deployment plans or speculative demand?
This is supply-chain planning, not a recommendation to stockpile components. Excess inventory can be as damaging as shortages when technology changes quickly.
What the result means for the semiconductor industry
SMIC’s results support three cautious conclusions.
First, the AI infrastructure cycle is creating demand outside the most advanced compute chips. Second, mature and specialty capacity can gain pricing power when supporting components become constrained. Third, a company’s revenue mix and utilization may matter as much as the headline label attached to its technology.
There are also limits. China accounted for about 90% of SMIC’s quarterly revenue, so the company’s results are heavily influenced by domestic demand. Export restrictions can affect access to equipment and advanced technology. A one-time gain also contributed to profit, making revenue, margin, shipments and utilization more useful for assessing operating momentum than profit alone.
What to watch next
The next useful indicators are third-quarter revenue growth, gross margin, utilization, average selling price and the speed at which new production lines qualify. Investors and customers should also watch whether demand remains concentrated in a few AI-related categories or spreads more broadly across industrial, automotive, consumer and communications products.
Another question is whether customers placed orders early to protect themselves from anticipated shortages. Early ordering can strengthen one quarter while reducing demand later. Management commentary and inventory data will help distinguish durable consumption from timing effects.
The bottom line
SMIC’s price increases and record quarterly revenue show that AI’s semiconductor impact is becoming more distributed. The demand is not limited to the most visible accelerators; it is reaching the supporting components required to power, connect and control AI systems.
That is a meaningful business signal, but not proof of a permanent shortage or a guaranteed investment outcome. The most reliable interpretation will come from tracking shipments, utilization, product mix and capacity additions over several quarters.
Frequently asked questions
Why did SMIC raise chip prices?
SMIC said strong demand and tightness in its most sought-after capacity supported price adjustments. The exact change varies by product and contract.
Does SMIC manufacture AI chips?
SMIC manufactures a broad range of semiconductors, including logic and supporting chips used across computing, communications, consumer and industrial products. AI demand can affect both processors and peripheral components.
Did SMIC’s profit rise only because of operations?
No. Operating performance improved, but management also disclosed a one-time subsidiary-related gain. Readers should assess revenue, margin, shipments and utilization alongside profit.
