SMIC eyes capacity expansion as AI chip demand outpaces forecasts
Synopsis
SMIC's co-CEO revealed that wafer starts are 'far exceeding' earlier forecasts, driven by AI infrastructure demand — and BCD power-management chip orders are already locked in through end-2027, pushing the foundry to its practical capacity limit.
Key Takeaways
SMIC is evaluating equipment additions at existing factories after AI -driven chip demand outpaced its own forecasts, management said on 14 August 2026 .
Co-CEO Zhao Haijun stated that 'future wafer starts are far exceeding our previous expectations' during the company's earnings call.
Demand is strongest for supporting chips in AI servers and data centres, including logic chips, power-management products, and optical module components.
Orders for BCD (bipolar-CMOS-DMOS) power-management chips are reportedly booked through the end of 2027 .
SMIC is operating near its practical capacity limit, making the expansion decision strategically urgent.
Further details on expansion plans are expected in upcoming company announcements or briefings.
Semiconductor Manufacturing International Corporation (SMIC), China's largest contract chipmaker, is evaluating equipment additions at its existing factories after demand for mature-node chips used in AI infrastructure surged well beyond its earlier projections, company management said on Friday, 14 August 2026.
Demand Far Exceeds Earlier Projections
SMIC disclosed during an earnings call that customer orders had risen significantly above forecasts made earlier this year. The driver: a global AI infrastructure buildout that is triggering shortages in the supporting chips that sit alongside high-end AI processors in servers and data centres. Co-CEO Zhao Haijun characterised the scale of the shift in direct terms: “Future wafer starts are far exceeding our previous expectations,” referring to the volume of new chip batches entering production lines.Expansion Plans Under Review
Zhao said SMIC is actively adjusting its capacity roadmap and may install additional equipment at existing sites where floor space permits. Further specifics are expected to be disclosed in upcoming announcements or briefings, according to company management. The move signals a meaningful shift from the cautious stance many chipmakers adopted in the face of earlier demand uncertainty.Which Chips Are in Shortest Supply
Demand is concentrated in chips that support AI server and data-centre workloads — including logic chips, power-management integrated circuits, and optical module components. Orders for BCD (bipolar-CMOS-DMOS) power-management products, in particular, are reportedly booked through the end of 2027, Zhao noted, underscoring the multi-year visibility now embedded in SMIC's order book.Why It Matters: Capacity at Its Practical Limit
SMIC is operating close to what management described as its practical capacity ceiling. That constraint, combined with surging orders, places the company at an inflection point: expand aggressively or risk losing share to rival foundries. The situation also highlights a structural dynamic in the global AI supply chain — leading-edge processors depend on a broad ecosystem of mature-node supporting chips, and bottlenecks anywhere in that chain can slow deployments.What to Watch Next
Investors and industry observers will be focused on the timeline and scale of any equipment installations SMIC announces, as well as whether order visibility beyond 2027 extends further. The company's ability to convert current demand into sustained revenue will hinge on how quickly it can bring incremental capacity online without compromising yields or delivery schedules.Point of View
The policy inadvertently concentrated AI-adjacent demand — power management, logic, optical — at mature nodes where SMIC is fully competitive. What mainstream coverage often misses is that the AI supply chain's real bottleneck is not the flagship GPU, but the dozens of supporting chips that enable it, and those chips are overwhelmingly built on older process nodes. Order visibility stretching to end-2027 suggests this is a structural shift, not a cyclical blip, which changes the calculus for global foundries competing for the same customer base. If SMIC can add capacity faster than rivals, it stands to entrench itself as the indispensable supplier for a class of chips that every AI data-centre builder needs.
NationPress
14 Aug 2026
Frequently Asked Questions
Why is SMIC considering expanding its chip manufacturing capacity?
SMIC is considering expansion because demand for mature-node chips used alongside AI processors has surged well beyond its earlier forecasts. Co-CEO Zhao Haijun said during an earnings call that 'future wafer starts are far exceeding our previous expectations,' and the company is now operating near its practical capacity limit.
What types of chips are driving the demand surge at SMIC?
The strongest demand is for chips that support AI servers and data centres, including logic chips, power-management integrated circuits, and optical module components. Orders for BCD (bipolar-CMOS-DMOS) power-management products are reportedly booked through the end of 2027 .
What did SMIC's co-CEO say about the company's expansion plans?
Co-CEO Zhao Haijun said SMIC is adjusting its expansion roadmap and may install additional equipment at existing sites where space is available. He indicated that further details would be disclosed in upcoming announcements or briefings.
How does SMIC's situation reflect broader trends in the AI chip market?
The global AI infrastructure boom is creating shortages not just in cutting-edge processors but in the mature-node supporting chips that enable them — a segment where SMIC is a key supplier. This highlights a structural dependency in the AI supply chain on older-generation foundry capacity.
Who is most affected by SMIC's capacity constraints?
Customers building AI servers and data-centre infrastructure are most directly affected, as they rely on the power-management, logic, and optical module chips that SMIC produces. Any delay in capacity expansion could slow procurement timelines for companies scaling AI deployments globally.
