SMIC raises wafer prices by 10%: memory tightens the chip supply chain in China

China’s largest semiconductor foundry is making its move again. SMIC (Semiconductor Manufacturing International Corp.) is reportedly applying a price increase of around 10% on part of its production portfolio, mainly driven by orders tied to memory and the race to secure capacity in a market that is beginning to tighten again.

The information comes from a TrendForce analysis citing local Chinese media and industry sources. Although SMIC has not publicly confirmed this, the context aligns with a key point: the company has been operating at very high utilization levels, which reduces its ability to absorb extra demand without adjusting prices.

Why is memory behind this move?

While SMIC doesn’t compete head-to-head with the global DRAM and NAND giants, its role is important in the domestic Chinese memory ecosystem: it provides capacity at mature and specialized nodes for components related to DRAM, specific memories, and controllers. That’s where the current demand surge is focused.

TrendForce indicates that memory was one of the first areas to see price adjustments after falling to “unusually low levels,” prompting suppliers to raise rates ahead of other segments. Factors include the push from smartphones and AI applications, as well as rising raw material costs.

Simultaneously, the market is experiencing a shift that has become more evident by 2025: AI no longer only competes for GPU market share, but also for memory. TrendForce, citing sources like Commercial Times and industry experts, warns that in 2026, demand related to AI could “consume” a significant portion of global DRAM capacity, with HBM and GDDR7 growth acting as bottlenecks.

The key data: factories at capacity

If a foundry is nearing full utilization, it has only two realistic options: invest to expand (slow) or raise prices (quick). In SMIC’s case, TrendForce reports its utilization rate reached 95.8% in the third quarter of 2025 (up from 92.5% in the previous quarter). Similarly, Hua Hong — China’s second-largest foundry — reported a 109.5% utilization rate, effectively indicating it is maximizing capacity through internal metrics and ramp-ups.

In this scenario, a 10% increase can be seen as a “moderate” move to maximize revenue per wafer without risking demand collapse: when there’s a queue, price sensitivity drops and the priority shifts to securing supply.

Quick table: China’s foundry position (according to TrendForce)

Geopolitical and mature nodes: the “funnel effect” giving pricing power

TrendForce places this situation within a familiar context: export restrictions and limited access to cutting-edge equipment, but also a collateral effect that has grown over time: more Chinese companies are prioritizing domestic manufacturing of mature nodes to reduce exposure to trade disruptions.

This doesn’t automatically make mature nodes “easy”; in fact, when many projects cluster within the same technological segment (and capacity doesn’t expand at the same pace), it creates a pricing power. This power first manifests in high-volume, fast-turnover products (related memory, controllers, consumer components).

TrendForce even notes an external factor that could tighten this segment further: TSMC reportedly plans to consolidate 8-inch capacity and shut down some lines by late 2027, while also adjusting its older factory base. Additionally, sales of used equipment to VIS (Virtual Semiconductor) are expected to add a value of NT$ 2.9 billion in 2025.

Potential impacts (and why they don’t just translate directly into end-product prices)

A 10% increase “per wafer” doesn’t automatically mean the final product is 10% more expensive. Between wafer and device, factors like wafer yield, packaging, testing, assembly, logistics, margins, and market positioning all influence the final price. However, it does add upward pressure on industrial costs, especially if this trend spreads across other suppliers or utilization remains high.

Likely layer-by-layer impact:

• Chinese designers and manufacturers reliant on SMIC: higher direct costs and increased need to offset through pricing or margins.
• Consumer electronics: incremental pressure on BOM (bill of materials) costs if these costs become entrenched across multiple chips in a product.
• AI infrastructure and data centers: renewed focus on memory (not just GPUs), where any restrictions tend to amplify downstream prices due to a “domino effect,” especially during demand spikes.

In other words: it’s not an immediate “shock,” but another sign that the mature capacity segment—once considered commoditized—is reverting to a scarce resource.

Frequently Asked Questions

What exactly is a “wafer” and why does its price matter so much?
A wafer is a silicon disc on which dozens or hundreds of chips are produced in parallel. If the cost per wafer rises in a foundry, the base manufacturing cost for those chips also increases, which can partially pass on to the end products that use them.

Why would a 10% increase at SMIC relate to memory if SMIC isn’t a “DRAM manufacturer”?
Because, even if SMIC isn’t a global leader in DRAM/NAND, it supplies capacity for parts and production linked to China’s memory ecosystem—including mature nodes and ancillary chips. When memory demand tightens, it pushes prices up along the supply chain.

Could this lead to higher costs for phones, laptops, or even servers in 2026?
It could add upward pressure, especially if memory capacity and demand on mature nodes remain constrained. It’s not a direct one-to-one relationship but rather one factor among others, like AI demand, inventory levels, and manufacturing limitations.

What should companies and IT procurement teams (including sysadmins) watch for?
Signs include sustained high utilization rates, longer lead times, and “surcharge” clauses in contracts for memory-related components or mature nodes. When capacity tightens, availability problems often appear before price hikes.

Sources: tomshardware and trendforce