By | Gold Corner Finance, Author | Tian Yu
Over the past few years, three words have been inseparable from discussions about SMIC: lagging behind, constrained, and catching up.
Unable to procure top-tier equipment or receive orders for advanced chips, the market has persistently questioned when it can narrow the gap with TSMC.
For this company carrying the hopes of domestic chip manufacturing, merely surviving and getting its fabs built has already been considered a victory; expectations of achieving full capacity utilization, raising prices, and truly making substantial profits seemed like distant luxuries.
The moment of greatest triumph for SMIC has occurred precisely when it seemed least likely to be profitable.
The latest financial report shows that in the second quarter of 2026, SMIC's revenue surpassed $30 billion for the first time, a year-on-year increase of 36.1%; net profit attributable to shareholders reached $4.79 billion, a staggering increase of 261.7% year-on-year.
Even if the substantial profit increase includes one-time gains, the strong rebound in its core business is undeniable:
Second-quarter wafer shipments (measured in 8-inch equivalents) reached approximately 2.9 million units, a sequential increase of 14%; the average selling price per wafer rose sequentially by 5.7%; capacity utilization soared to 93.7%.
Fabs operating near full capacity, shipments continuing to grow, unit prices rising against the trend.
What's more intriguing is that all this is happening precisely as the global AI wave sweeps the world.
Global capital is frenziedly chasing top-tier GPUs, HBM, 3nm, 5nm, and advanced packaging. According to conventional logic, SMIC, lacking cutting-edge equipment and struggling to secure top chip orders, should be standing on the periphery of this AI wave, watching NVIDIA and TSMC take the largest slice of the pie.
After years of being "choked," has SMIC truly arrived at its turnaround moment? How did this seemingly anomalous comeback happen?
Government Subsidies Cannot Sustain This Financial Report
The answer begins with the most suspicious part of this financial report: government subsidies.
Looking back at the last recovery from a downturn, government subsidies had an immediate positive impact on SMIC.
In January 2025, smartphones were officially included in the national new purchase subsidy program, with subsidies of 15% (maximum 500 RMB per item) for models under 6000 RMB. Just one month later, SMIC confirmed during an earnings call that driven by consumer stimulus policies, customers had a strong desire to replenish inventories. Urgent orders emerged in consumer electronics, connectivity, and mobile phone sectors, leading to "off-season resilience" in the first quarter of 2025.
By May of the same year, management attributed the growth more directly: changes in the international landscape triggered customers to pull orders forward, trade-in subsidies stimulated end-demand, coupled with bottom-fishing restocking in industrial and automotive sectors, orders were transmitted up the supply chain to the wafer fab, and SMIC naturally became a beneficiary.
However, by 2026, the "subsidy-driven" logic began to show cracks.
The most significant signal is that the smartphone business is rapidly "retreating to the second tier." In the second quarter of 2025, smartphones contributed 25.2% of SMIC's wafer revenue; by the first quarter of 2026, this proportion dropped to 18.9%; in the second quarter, it further shrank to 16.9%. In just one year, it evaporated nearly 8 percentage points.
During this period, the company's revenue structure underwent a crucial transformation: consumer electronics remained the foundation with a 44.2% share; the combined share of industrial and automotive surged from 10.6% a year ago to 16.5%; computers and tablets also increased to 15.6%.
SMIC's revenue pillars are evolving from past heavy reliance on smartphones towards a more diversified and dispersed portfolio structure.
Why can't even government subsidies boost smartphones anymore?
The root cause lies in the soaring prices of memory chips. As costs of core components like memory and flash storage have risen sharply, smartphone manufacturers are forced to choose between "raising prices" or "reducing configurations."
In the past, subsidies could save consumers hundreds of RMB on a 3000 RMB phone, but now the savings from subsidies are instantly eroded by upstream cost increases.
IDC data shows that in the first half of 2026, Chinese smartphone shipments were approximately 134 million units, a year-on-year decline of 4.2%. During the "618" promotional period, sales declines approached 15%. Subsidy benefits remain, but consumers are no longer easily swayed.
However, broadening the view to the entire digital device sector paints a different picture.
Ministry of Commerce data shows that in the first half of 2026, national sales of new digital and smart products reached 79.098 million units, a year-on-year increase of 13.4%, with a sharp 32% increase in June alone. New form factors like smart glasses saw significant volume growth under subsidies, with June sales up 30.6% year-on-year.
This indicates that subsidies have not failed; rather, incremental benefits are accelerating their shift from smartphones towards smart glasses, wearables, and new smart terminals.
For SMIC, the diversification of end-device ecosystems is certainly beneficial, but policy stimulus still cannot explain the core paradox:
Why, while smartphone revenue share shrinks drastically, can SMIC's wafer shipments, average selling price, and capacity utilization achieve a robust rebound in "three-frequency resonance"?
If relying solely on short-term subsidy-driven pulse orders, a fab might manage to "trade price for volume" to sell more, but would absolutely not gain pricing power. Furthermore, the company's second-quarter capacity utilization soared to 93.7%, and for certain supply-constrained categories, channels for negotiating price increases with customers have already opened.
Policy subsidies at best gave SMIC a nudge at the bottom; they could never single-handedly propel a fab to the spotlight of near-full capacity and proactive price hikes.
This suggests that something more significant than a consumer recovery may be happening at SMIC: the supply-demand dynamics of mature processes are being rewritten.
Strangely enough, what's pushing mature processes from "overcapacity" to "supply shortage" is precisely the industry that seems farthest from SMIC — AI.
In the AI Era, SMIC is Quietly Making Money Behind the Scenes
This sounds counterintuitive.
At the peak of the AI wave, the biggest money-makers are top-tier GPUs, and the most sought-after technologies are 3nm, 5nm, and advanced packaging — none of which are SMIC's primary battlegrounds. According to past logic, the hotter AI gets, the more money NVIDIA and TSMC make, and the more glaring SMIC's gap with them becomes.
But people always focus on the most expensive computing core, overlooking what actually powers an AI server.
If an AI data center is likened to a burgeoning supercity, the GPU is the most dazzling skyscraper in town. But city operation also requires dense power grids, substations, roads, and traffic lights. Even if SMIC hasn't yet built the tallest skyscrapers, it can produce the massive amounts of "power grids" and "traffic lights" urgently needed by this city.
After high-voltage current enters an AI server, it needs multi-stage conversion, step-down, and distribution to transform into the low-voltage, high-current power required by GPUs. High-speed communication between GPUs, HBM, and switch chips also heavily relies on interface and controller chips. The more GPUs, the more peripheral support chips.
According to Infineon's measured data, the value of power semiconductors in an ordinary server is only $65 to $80. However, once upgraded to an AI server, this value explodes to $850 to $1800.
UBS predicts that the global market capacity for power semiconductors in AI data centers will surge from approximately $1.5 billion in 2025 to about $2.5 billion in 2026, and further expand to around $3.8 billion by 2028.
Most of these chips do not require 3nm or 5nm; 55nm, 65nm, or even 90nm and other mature processes are the mainstay. This is precisely SMIC's deepest stronghold.
Processes once considered "not advanced enough" are now starting to see price increases in the AI era. SMIC may not be able to earn from the top-tier GPUs, but it can steadily earn money from the fact that "for every additional GPU deployed, a ring of supporting chips is needed."
But this is only the first layer of spillover benefits AI brings to SMIC. The second, more concealed layer of benefits comes from the dramatic reshuffling of the global capacity landscape.
In recent years, TSMC and Samsung have fully tilted their core capital and engineering talent towards advanced processes and advanced packaging. As a result, TrendForce predicts that global 8-inch wafer capacity will shrink by 2.4% year-on-year in 2026.
However, overseas giants shrinking mature capacity does not mean demand from automotive, industrial control, consumer electronics, and AI server support disappears. Supply is shrinking, but demand is increasing. Mature processes once predicted to be "over-supplied" are instead becoming tight.
AI acts like a massive "water pump," siphoning resources from overseas giants into the ultra-fast lane, squeezing out space for mature processes. Faced with overseas capacity shortages and price hikes, global customers have to search for suppliers anew.
TrendForce observes that since the second half of 2025, some high-voltage process and CIS (image sensor) customers have begun transferring their production to Chinese mainland wafer fabs, aiming to lock in more certain capacity supply and cost expectations.
With its massive mature-process capacity and comprehensive platforms, SMIC has become the optimal recipient. Management confirms that an increasing number of overseas orders are turning to China due to overseas capacity squeeze. Moreover, switching wafer foundries involves high testing and time costs, making orders highly sticky once transferred.
External restrictions have slowed SMIC's pace of breakthrough into advanced processes, but the AI explosion has elevated the value of mature processes.
Through the reallocation of global capacity, AI has effectively "crowded" batches of orders directly to SMIC's doorstep.
Earned Money is Placed Back into the Game
Over the past few years, SMIC has faced a real dilemma: building more fabs means accruing larger depreciation.
In the second quarter of 2025, the company's depreciation and amortization was $8.79 billion; by the first quarter of 2026, it had risen to $10.88 billion; in the second quarter, it further climbed to $12.12 billion. In just one year, it increased by nearly 40%.
For a wafer fab, this is dangerous. Once equipment and facilities are operational, depreciation doesn't stop, regardless of orders. Lower capacity utilization means higher fixed costs allocated per wafer, and profits are more easily consumed by the massive fab.
But AI benefits have reversed this logic.
In the second quarter of 2026, SMIC's capacity utilization rose to 93.7%. The massive assets that once weighed down the income statement are now being filled with orders. Equipment that previously waited idly for customers now requires allocating capacity among different orders.
More important than full fabs is that SMIC has started raising prices.
Company management previously stated they have negotiated price adjustments with customers for certain product categories in short supply. The price increase effect began to show in the second quarter and will continue to be reflected in the third and fourth quarters.
This means SMIC is transitioning from "building fabs and waiting for orders" to "orders chasing after capacity."
However, a 93.7% capacity utilization rate also indicates limited room for increased shipments from existing fabs. For SMIC to continue growing, it must expand capacity again.
In the second quarter of 2026, SMIC's capital expenditures were $18.36 billion, a significant increase from $15.63 billion in the first quarter. According to management expectations, by the end of 2026, monthly capacity will increase by an equivalent of approximately 40,000 12-inch wafers compared to the end of 2025.
Clearly, SMIC is doing one thing: while orders are still flowing, prices can still rise, and capacity is near full, quickly build out the next round of capacity.
But this is also a high-stakes gamble.
From equipment ordering, fab construction to capacity ramp-up, it often takes one to two years for a wafer fab. The industry climate may have changed by the time capacity built today in response to booming orders truly comes online. Expanding too slowly risks missing the fleeting AI order window; expanding too fast risks facing demand cooling and depreciation backlash a few years later.
Therefore, what truly determines whether SMIC dares to step on the accelerator for this round of expansion is not just opportunity, but also cash.
Fortunately, SMIC currently holds a decent hand. In the second quarter of 2026, SMIC's operating cash flow reached $25.22 billion, and cash and cash equivalents at the end of the period exceeded $8.2 billion.
Simultaneously, accounts payable and contract liabilities increased to $3.466 billion and $768 million respectively. Supplier payment terms and customer prepayments also buffer the financial pressure from expansion.
For a wafer fab that spends billions on equipment annually, this is actually very comfortable. Because it means that SMIC is funding its expansion not just with its own money.
This is also why the biggest difference between this expansion round and previous ones may not just be "stronger AI demand," but that SMIC finally has the opportunity to form a true virtuous cycle:
More orders lead to higher capacity utilization; tighter capacity leads to price increases; price increases improve profits and cash flow; better cash flow enables further expansion; new capacity then captures more orders.
In the past, the market focused on SMIC primarily because it carried the industrial mission of China's semiconductor self-sufficiency in manufacturing. Now, it's beginning to prove that it's not only important but can also sustainably make money and achieve self-sustaining growth by relying on market orders.
SMIC may not have crafted the most dazzling gem on the AI crown, but it is turning the base that supports this crown into an increasingly profitable business.
The most substantial moments of domestic substitution have never been about loud slogans, but about fabs running at full capacity, products commanding higher prices, and the real money earned being sufficient to build the next fab.
References:
- Coretech News "1.12 Trillion Wafer Foundry Giant Earns Over 200 Million Daily, Zhao Haijun: No Possibility of Price Cuts This Year"
- Boss Finance "While Everyone Was Watching TSMC's 3nm, SMIC Quietly Raked in Massive Profits"
