Behind SMIC's Q3 Report: How Much Manufacturing Increment Can Domestic AI Chips Really Bring?
I'm LongbridgeAI, I can summarize articles.Fab utilization is ramping up, and profits are growing even faster—who's footing the bill for this growth? SMIC's latest Q3 report brings the relationship between domestic chip demand and manufacturing capacity into sharp focus.
On Nov 12, SMIC released its Q3 2026 results. Under Chinese GAAP, revenue for the first three quarters reached RMB 60.618bn (+22.4% YoY), with net profit attributable to shareholders at RMB 7.314bn (+91.6% YoY). In Q3 alone, revenue hit RMB 21.983bn (+28.1% YoY) and net profit attributable to shareholders was RMB 2.847bn (+87.6% YoY).
Rising shipment volumes and higher capacity utilization underpin these results. However, gauging the sustainability of this growth requires dissecting the sources of customer demand and the fab's ability to absorb new products.
How much of the AI-driven manufacturing upside is visible in the earnings?
The quarterly report did not break out revenue from AI chip foundry services separately; based on current disclosures, it remains difficult to quantify its contribution to growth. Analyzing customer structure and process requirements offers a clearer view of where the company's opportunities lie.
In Q3, wafer business accounted for 95.9% of main operating revenue. Within wafer revenue, consumer electronics made up 40.6%, industrial & automotive 19.3%, PC & tablet 18.3%, and smartphones 14.3%.
Compared to the same period last year, the share of wafer revenue from industrial & automotive rose from 11.9% to 19.3%, while PC & tablet increased from 15.2% to 18.3%. Smartphone share declined from 21.5% to 14.3%. Traditional consumer electronics remain the largest source, but the weight of industrial, automotive, and computing devices is increasing, signaling a shift in customer demand priorities.
Geographically, the China region accounted for 90.8% of main operating revenue, up from 86.2% YoY. For domestic fabless companies, local foundries provide crucial process adaptation and mass production support, which is vital for bringing products to market.
AI-related manufacturing demand extends far beyond high-performance computing chips themselves. Data centers also require supporting chips such as power management and interface connectors, some of which can leverage mature or specialty processes. SMIC has previously disclosed process platforms for power management and analog circuits, providing avenues to participate in the manufacturing of these supporting components.
For design houses, completing a chip design is just the starting point. Products must be adapted to manufacturing processes, undergo validation, and gain end-user adoption before generating stable procurement orders. The development of domestic AI creates opportunities for local foundries, but which specific product categories they can capture depends on collaboration between customers and fabs throughout these stages.
Capacity utilization hits 97.3%—can profitability keep pace?
In Q3, SMIC sold 3.0337 million wafers, up 5.7% QoQ and 21.4% YoY. Month-end monthly capacity stood at 1.1168 million wafers, converted to an 8-inch standard logic equivalent. Capacity utilization rose from 93.7% in Q2 to 97.3%, indicating that existing lines are running at high load levels.
Fabs must continuously bear equipment depreciation, facility operations, and personnel costs. Higher shipments allow these fixed costs to be spread across more units, creating conditions for margin improvement. However, the 97.3% figure is an aggregate metric; it does not imply that all processes or lines are equally constrained, nor does it mean capacity is freely interchangeable across different products.
Under IFRS, Q3 sales revenue was $3.190bn (+6.1% QoQ), with gross profit at $953mn and a gross margin (GPM) of 29.9%, expanding 460bps QoQ and 790bps YoY. The company explicitly attributed the QoQ GPM increase to higher capacity utilization and a favorable product mix, reflecting operational improvements on the manufacturing side.
Under Chinese GAAP, non-GAAP net profit attributable to shareholders for the quarter was RMB 2.639bn (+107.3% YoY), further supporting the view of operational improvement. However, gross profit and net profit do not always move in tandem: under IFRS, Q3 net profit attributable to shareholders was $428mn (-10.6% QoQ), impacted by other income turning from a gain to a loss compared to the previous quarter. Profit data under the two accounting standards and currencies should be compared separately.
SMIC's Q4 guidance suggests potential divergence between revenue and margins: revenue is expected to grow 4% to 6% QoQ, while GPM is forecast at 26% to 28%, below Q3 actuals. While revenue may continue to rise, per-unit profitability could retreat. The guidance does not break down the reasons, so one cannot conclude that demand for a specific product is weakening or that a particular line is facing issues.
After expansion, how to secure reasonable returns?
When existing lines are busy, new demand tests the speed of expansion. But once facilities and equipment are in place, process tuning, yield improvement, and customer validation are required before stable delivery can be achieved. The pace of new capacity additions must align with the demand for applicable products.
In Q3, capital expenditures were RMB 7.423bn, lower than Q2's RMB 12.574bn and last year's RMB 17.065bn. Quarterly capex is influenced by project progress and procurement schedules; this single data point does not reveal long-term expansion trends. The company reiterated that capacity expansion will remain customer-demand driven.
Prior investments will continue to weigh on costs. Under IFRS, depreciation and amortization included in cost of sales for Q3 amounted to approximately $946mn (+17.6% YoY). For manufacturing businesses, even if quarterly capex declines, the depreciation burden on already commissioned equipment does not disappear.
Product mix also affects investment returns. In Q3, 12-inch wafers accounted for 78.9% of wafer revenue, up from 77.0% YoY. Larger wafer sizes support economies of scale, but size and process node advancement are distinct concepts; changes in revenue share alone do not prove improvements in yield or unit costs. Profitability across different processes still needs to be assessed in conjunction with product characteristics and manufacturing efficiency.
When selecting a foundry, customers also consider process stability, product yield, and long-term delivery capability. The company's accumulated mass-production experience helps reduce integration costs during new product introduction. Meanwhile, supply conditions for equipment, spare parts, materials, and software services impact process iteration and line operations. These constraints require gradual resolution through technical investment and supply chain collaboration.
Cash flow supports continued investment. For the first three quarters, under Chinese GAAP, net cash flow from operating activities reached RMB 28.625bn (+133.0% YoY), primarily driven by increased cash received from sales. Meanwhile, cash paid for constructing fixed assets, intangible assets, and other long-term assets totaled RMB 29.142bn, slightly exceeding net operating cash inflow. Improved collections are encouraging, but construction investments still require funding.
Conclusion
Looking ahead, if domestic AI products can enter more practical applications, local manufacturing is likely to see more sustained iterative demand. Co-developing across multiple product generations may become a direction for deeper collaboration between design and manufacturing firms, expanding commercialization space for domestic chips.
References
1. SMIC 2026 Q3 Report, Nov 12, 2026: Chinese GAAP financial data, application structure, 8-inch conversion basis, and cash flows.
2. SMIC 2026 Q3 Unaudited Results Announcement: IFRS data, depreciation/amortization within cost of sales, and Q4 guidance.
3. SMIC 2024 ESG Report, p.48: Technical background on power management, analog, and other process platforms.
4. Texas Instruments Data Center Computing Technical Materials: Applications of supporting chips such as power management, interfaces, and connections.
Data as of Nov 13, 2026. RMB and USD figures, as well as data under different accounting standards, are presented separately. The financial data herein is unaudited, and Q4 guidance constitutes forward-looking information.
The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
