SMIC Raises Wafer Prices as AI Demand Tightens Capacity

SMIC has raised wafer prices after customer negotiations in Q1, despite years of intense competition among manufacturers serving mature chip markets. The Techmeme SMIC headline also says wafers processed during Q3 will carry higher charges, according to Reuters.

That change matters because foundries rarely gain broad pricing power simply by announcing it. Customers must accept new terms, keep their orders, and wait months for finished wafers. SMIC says that process has already started.

The immediate story is not that SMIC suddenly became a direct substitute for TSMC’s most advanced production. The more important conflict sits inside mature and specialty manufacturing. AI infrastructure is drawing capacity toward memory, networking, power management, and data transmission components.

That shift is tightening selected production lines while SMIC continues adding equipment and factories. It is also testing whether demand can absorb higher prices without reviving oversupply concerns.

SMIC reported second-quarter revenue of about $3.01 billion, roughly 20% above the previous quarter. Shipments increased about 14%, while the blended wafer price rose about 5%.

Those figures suggest volume and pricing improved together. Gross margin reached 25.3%, up from 20.1% in Q1, according to the company’s reported results. SMIC guided for another sequential revenue increase of 2% to 4% in Q3.

The numbers make the price decision more than a negotiating claim. However, they do not establish that every process, customer, or application now faces the same increase.

The central question is whether AI-driven scarcity has created durable pricing power across SMIC’s specialty platforms. The alternative is a shorter cycle shaped by customer stockpiling, product mix, and temporary capacity constraints.

The Techmeme SMIC Headline Marks a Real Pricing Shift

SMIC is moving from selective price discussions toward higher realized revenue per wafer.

The Techmeme SMIC item highlights two stages of that transition. Zhao Haijun says SMIC raised prices after negotiating with customers in Q1. He also says wafers processed in Q3 will cost more.

That timing reflects how foundry contracts work. A customer first reserves capacity and submits a chip design for an approved manufacturing process. SMIC then schedules wafer starts, completes fabrication, and recognizes revenue after delivery conditions are met.

A price agreement can therefore take more than one quarter to appear fully in reported results. Existing orders may retain earlier terms, while newly processed wafers carry adjusted pricing.

Q1 offered the first measurable indication. SMIC generated $2.505 billion in revenue, up 0.7% sequentially. Wafer shipments slipped 0.2%, but the blended wafer price rose 2.5%.

Management attributed that increase to stable or rising prices in selected markets where SMIC holds stronger positions. Those figures came from SMIC’s Q1 earnings discussion.

The second quarter strengthened the signal. Revenue advanced much faster than management’s original guidance, and the average selling price increased alongside shipment volume.

This combination is important. A higher average price alone can reflect a richer product mix, such as more complex wafers replacing simpler products. It does not always mean customers accepted higher rates for comparable manufacturing.

Zhao’s comments provide the missing mechanism. SMIC negotiated increases for products occupying constrained capacity, then applied higher charges as those orders moved through production.

The company took a similar approach to its fully loaded 8-inch lines earlier in 2026. Zhao said some low-layer-count analog products consumed disproportionate time on bottleneck equipment.

SMIC gave those customers a choice. They could accept higher prices, or SMIC could reduce the volume allocated to their products. Management described that process as product portfolio optimization.

This is selective pricing power, not a universal surcharge. A foundry can charge more where a product uses scarce tools, where alternatives are limited, or where demand exceeds available output.

The distinction protects the analysis from a tempting overstatement. SMIC has not disclosed a single percentage increase covering its entire customer base. Nor has it published a complete list of affected technology platforms.

The Techmeme SMIC story is still significant because the price increases are reaching actual wafer processing. The next question is why customers are accepting them while new Chinese manufacturing capacity continues to enter the market.

AI Demand Is Tightening More Than Advanced Chip Capacity

The AI buildout is pulling mature components into the same supply contest as advanced processors.

AI servers require advanced accelerators, but those processors are only one part of the system. Each server also needs power-management chips, networking components, storage controllers, sensors, and data transmission products.

Many of those supporting chips use mature or specialty manufacturing processes. Mature nodes are established production technologies that prioritize cost, reliability, and manufacturing scale over maximum transistor density.

SMIC has substantial exposure to these categories. Its platforms cover analog chips, microcontrollers, image sensors, display drivers, embedded memory, power-management products, and connectivity components.

Demand for these supporting parts can rise even when SMIC does not manufacture the leading AI accelerator itself. The foundry therefore receives a second-order benefit from data-center investment.

Zhao described this effect earlier in 2026 as an AI-related pull on capacity. In a Reuters interview, he said overseas customers were moving some orders to Chinese factories.

According to Zhao, foreign foundries had redirected resources toward AI products, memory, and high-bandwidth applications. That reduced available production for some older products previously made outside China.

The pressure also extends through the memory supply chain. AI systems consume large quantities of high-bandwidth memory, while memory manufacturers prioritize products with stronger demand and returns.

This can constrain components for smartphones, personal computers, and other electronics. Device makers may then change orders for logic, analog, or display chips because complete products require all components to arrive together.

SMIC experienced that interaction in late 2025 and early 2026. Some smartphone customers reduced orders because they worried about memory availability. The effect carried into Q1 even as other applications strengthened.

The result is not a simple shortage across every chip category. It is a reallocation problem. Capacity, equipment time, customer budgets, and supplier attention move toward products connected to AI infrastructure.

That reallocation can make an older manufacturing line more valuable. A power-management chip produced on an established process may become a bottleneck if server demand fills the relevant tools.

Foundries can respond by prioritizing higher-value orders. They can also raise prices for products that consume excessive time on constrained equipment.

SMIC’s strong utilization supports that strategy. Its Q1 utilization rate was 93.1%, even after new factory capacity entered the calculation. That level left limited flexibility on already constrained platforms.

The company had operated at 95.7% utilization during the previous quarter. It also shipped about 2.5 million 8-inch-equivalent wafers in Q1, essentially unchanged sequentially.

High utilization does not guarantee permanent scarcity. New capacity can eventually restore balance. However, it gives a foundry greater negotiating leverage while customers compete for specific processes and delivery windows.

That is why the current price movement reaches beyond an ordinary quarterly adjustment. AI spending is changing the value of production stages that sit far from the most visible accelerator chips.

Customers Need Capacity, While SMIC Needs Better Returns

The primary conflict is between customers seeking dependable mature-node supply and SMIC seeking returns on an expensive factory expansion.

SMIC has spent heavily to capture demand from Chinese chip designers and returning overseas orders. Its capital expenditure reached $8.1 billion in 2025, according to the company’s annual results.

Management expected 2026 spending to remain near that level. SMIC also warned that depreciation expenses would rise about 30% as newly installed equipment entered service.

Depreciation spreads the cost of manufacturing assets across their useful lives. It can reduce reported gross margin even when factories produce more wafers and revenue grows.

SMIC therefore faces an uncomfortable equation. It needs new capacity to serve localization demand, but every new production line adds costs before reaching efficient utilization.

The company added about 50,000 monthly 12-inch-equivalent wafers of capacity during 2025. Zhao said it planned to add roughly 40,000 more by the end of 2026.

Some equipment may not form complete production lines immediately. Export controls, missing supporting tools, installation schedules, and customer qualification can delay usable output.

Those delays increase the value of existing qualified capacity. A chip designer cannot move every order to another factory without redesign, validation, and reliability testing.

This is especially important for automotive, industrial, and power products. Customers in those markets often require longer qualification cycles and stable manufacturing records.

Once a product has passed those checks, changing foundries can create engineering expense and supply risk. SMIC can use that customer dependence when negotiating rates, provided its quality and delivery remain dependable.

Customers still have alternatives. Hua Hong Semiconductor competes across mature and specialty processes in China. United Microelectronics and GlobalFoundries also serve many established-node applications.

Reports of possible increases from UMC show that SMIC is not operating in isolation. Market expectations in early 2026 included higher prices for selected power-management and microcontroller products.

However, the competitive landscape differs sharply by process and geography. TSMC leads the most advanced nodes and reported $40.2 billion in second-quarter revenue, according to its official Q2 results.

TSMC generated 77% of wafer revenue from processes at 7 nanometers or below. Its exposure, customer mix, and economics are therefore unlike SMIC’s broader mature-node business.

Comparing their gross margins without that context would mislead readers. TSMC’s 67.7% Q2 gross margin reflects scale, technology leadership, and a richer advanced-product mix.

SMIC’s pricing decision concerns a different contest. It is trying to improve returns while funding expansion under tighter access to advanced manufacturing equipment.

Customers must decide whether a higher wafer charge costs less than delayed production or another factory qualification. SMIC must avoid pushing increases far enough to accelerate those alternatives.

That balance explains why negotiations matter. A unilateral list-price increase says little if customers leave. A negotiated increase that appears in processed wafers provides stronger evidence of leverage.

The Techmeme SMIC headline captures the moment when those negotiations begin affecting production economics. It does not settle how long customers will tolerate the new balance.

Higher Prices Do Not Remove SMIC’s Capacity Risks

SMIC’s stronger margin proves that conditions improved, but it does not prove that the improvement is permanent.

Several factors can lift average wafer revenue at the same time. Direct rate increases are one factor. Product mix, wafer size, process complexity, and customer timing can also change the reported average.

SMIC’s second-quarter increase therefore needs careful interpretation. A 5% gain in blended wafer pricing does not mean every customer paid 5% more for the same service.

The company says negotiated increases contributed to the movement. Yet it has not disclosed how much came from direct repricing and how much came from a richer mix.

Demand visibility also remains uneven. AI infrastructure orders appear strong, while smartphones and other price-sensitive products face higher memory and component costs.

A customer can reserve foundry capacity during a shortage and later reduce orders if final-device sales weaken. That pattern has repeatedly amplified semiconductor cycles.

Inventory creates another uncertainty. Customers sometimes place extra orders when they fear scarcity. Those orders can make utilization and pricing appear stronger before excess inventory reaches the supply chain.

SMIC’s Q3 guidance offers useful evidence, but not a complete answer. The company expects revenue to increase another 2% to 4% sequentially and gross margin to reach 26% to 28%.

That outlook suggests the price effect will continue beyond Q2. It also implies that utilization, scale, and product mix should offset some depreciation pressure.

Still, guidance remains a company forecast. Actual performance will depend on wafer acceptance, factory yields, equipment availability, customer schedules, and the broader economy.

Yield deserves particular attention. Yield measures the share of manufactured chips that meet specifications. Poor yield raises the cost of each usable chip, even when a factory appears busy.

SMIC previously encountered production disruption after maintenance and equipment validation work. Such problems can reduce sellable output and weaken margins independently of customer demand.

Export restrictions compound that operational risk. SMIC’s advanced manufacturing expansion faces limits on access to leading lithography systems and certain equipment support services.

Those restrictions do not prevent all capacity growth. They can still delay installations, complicate maintenance, and make process improvements more expensive.

SMIC also faces a longer-term supply concern created by its own expansion strategy. Chinese foundries are collectively adding substantial mature-node capacity.

Semicon China data cited by Reuters projected that Chinese producers would hold 37% of global 22-nanometer-to-40-nanometer capacity in 2026. The estimate rises to 41% in 2027.

More capacity supports domestic supply security, but it can weaken pricing if demand does not grow at the same pace. Hua Hong and other manufacturers will also pursue attractive specialty orders.

This creates the article’s core tradeoff. AI demand gives SMIC leverage today, while continued factory construction can reduce that leverage tomorrow.

The company needs price increases to fund capacity and absorb depreciation. Customers accept those prices because qualified output is tight. New output can eventually restore their bargaining power.

That cycle is why readers should not treat the Techmeme SMIC report as proof of permanent scarcity. It is stronger evidence that selected mature-node markets have entered a favorable pricing phase.

Three Signals Will Show Whether SMIC Can Keep Charging More

Q3 margins, customer behavior, and new capacity absorption will determine whether SMIC’s pricing power lasts.

The first signal is SMIC’s Q3 gross margin. Management guided for a range of 26% to 28%, following 25.3% in Q2 and 20.1% in Q1.

A result inside or above that range would show that higher prices are reaching recognized revenue. It would also indicate that utilization and product mix are offsetting higher depreciation.

A weaker result would not automatically disprove the pricing claim. Currency movements, startup costs, yields, and product mix can all affect gross margin.

However, a miss accompanied by flat average selling prices would weaken the argument for durable leverage. Investors should compare shipments, wafer revenue, and margin together.

The second signal is customer ordering after the Q3 increases take effect. Continued demand would show that access to qualified capacity matters more than the additional charge.

Order reductions would suggest customers had overbooked, resisted higher rates, or found alternatives. Smartphone and consumer-electronics demand deserves particular attention because those markets remain price sensitive.

Industrial and automotive orders provide a useful counterweight. Those products often have longer programs and stricter qualification requirements, which can support steadier factory demand.

The third signal is how quickly SMIC absorbs new 12-inch capacity. The company expects to add roughly 40,000 monthly 12-inch-equivalent wafers during 2026.

Utilization should remain high if AI-related components, domestic chip localization, and overseas order transfers keep pace. Falling utilization would signal that supply is catching demand.

This metric must be read carefully when a new factory begins production. Adding capacity expands the denominator before every tool reaches full output, temporarily lowering the reported rate.

Management used that explanation for Q1, when utilization declined to 93.1%. Subsequent shipment growth suggests the company filled a meaningful portion of the added capacity during Q2.

Competitor behavior will sharpen the picture. Pricing changes from UMC, GlobalFoundries, or Hua Hong would support the view that scarcity extends across mature-node manufacturing.

Aggressive discounts would point in the opposite direction. They could force SMIC to choose between protecting utilization and maintaining higher rates.

The comparison with TSMC will remain relevant but secondary. TSMC’s advanced-node growth shows the direct AI boom, while SMIC reflects demand spreading into supporting components.

That distinction is the most useful conclusion from the Techmeme SMIC story. AI infrastructure is not only increasing orders for the processors featured in product launches.

It is also changing factory allocation for ordinary but essential chips. Power control, networking, storage, and data movement all depend on manufacturing capacity that receives less attention.

SMIC now believes enough of that capacity is scarce to charge more. Its recent revenue, blended pricing, and margin support that claim.

The next quarter will test the harder proposition. Can SMIC preserve higher prices while expanding output, absorbing depreciation, and keeping customers from shifting orders?

Watch the Q3 margin first, customer order patterns second, and capacity utilization third. Together, those indicators will show whether this pricing turn has become a lasting industry shift.