Key Observation This Week: AI demand continues to drive capacity expansion at wafer fabs and equipment vendors, but the market is increasingly focusing on whether this expansion can be realized, if equipment supply can keep pace, and the return on AI investments.
AI continues to drive investment in advanced logic, memory, testing, and packaging, but industry observation priorities are shifting from "how strong is the demand" to "how will the demand be realized":
Can wafer fab expansion plans convert into effective capacity on schedule?
Can equipment supply keep up with construction and integration rhythms?
Can new computing power generate sustained returns?
Domestic Equipment Data: IC Equipment Growth is Rapid
In the first half of 2026, the China Electronic Specialized Equipment Industry Association (CESEIA) compiled statistics from reports of 102 member enterprises above a designated size. The sales revenue (excluding tax) of China's electronic specialized equipment industry was CNY 75.475 billion, a year-on-year increase of 1.9%; sales revenue of main products was CNY 63.596 billion, up 2.2% year-on-year.
Sales revenue for semiconductor equipment manufacturers among the association's members was CNY 55.064 billion, roughly flat year-on-year. Among these, sales revenue for integrated circuit (IC) production equipment reached CNY 47.883 billion, a year-on-year increase of 31.2%, accounting for 87% of total semiconductor equipment sales; sales revenue for solar cell production equipment was CNY 5.960 billion, down 66.5% year-on-year, attributed to overcapacity in solar cells. Driven by production equipment for compound semiconductor devices, sales revenue for other semiconductor device production equipment reached CNY 0.903 billion, an increase of 137.2%.
Chart: Compiled based on the completion status of major economic indicators for China's electronic specialized equipment industry from January to June 2026, as reported by CESEIA.
Data Source: China Electronic Specialized Equipment Industry Association.
The association expects the IC equipment market to continue growing, with specific forecasts as follows:
| Category | Association's Full-Year 2026 Forecast |
| IC Production Equipment | Approx. CNY 110 billion, growth of approx. 25% |
| Solar Cell Equipment | Approx. CNY 13 billion, decline of approx. 60% |
| Compound & Other Semiconductor Device Equipment | Approx. CNY 4 billion |
| LED Equipment | Not disclosed |
Data Source: China Electronic Specialized Equipment Industry Association.
*Note: Full-year figures are forecasts; statistical and forecast scopes are based on member enterprise report data and do not directly equate to total industry orders or realized capacity.
Meanwhile, publicly disclosed individual projects from enterprises offer insights into the stage of domestic equipment productization progress:
Hefei Kaiyue's independently developed track-type coater/developer adapted for memory processes has been shipped and delivered to memory chip manufacturing customers, but the company disclosed that subsequent work still includes client-side process verification and mass delivery. Hwatsing Technology's fully automatic panel-level CMP equipment has been shipped to advanced packaging lines for mass production application. Advanced Micro-Fabrication Equipment Inc. China (AMEC)'s cumulative shipment of advanced thin-film equipment exceeded 500 reaction chambers ("reaction chambers" refer to equipment unit metrics, not directly equivalent to complete machine counts or new customer capacity).
AI Chip Capacity Expansion Requires Capacity Planning
Wafer fab expansion is a systematic engineering project where equipment lead times determine the starting rhythm of production lines. Order-driven expansion must go through customer demand assessment, capital budgeting, factory construction, equipment procurement, and production ramp-up.
SMIC's semi-annual report for 2026 shows that the company's monthly capacity rose to 1.0965 million wafers (equivalent to 8-inch standard logic) in the second quarter, with a capacity utilization rate of 93.7%. Wafer shipments increased by 20.1% year-on-year in the quarter; total wafer shipments for the first half were 5.379 million wafers, up 14.9% year-on-year. In its Q2 earnings announcement, SMIC also cited the driving force and spillover effects of the AI industry as demand support, believing these effects would bring broad demand for IC manufacturing.
TSMC maintained its full-year capital expenditure budget of USD 60–64 billion during its Q3 2026 earnings call and stated that capex for 2026–2028 is expected to be significantly higher than the previous three years. The company linked its investment plan to customers' multi-year demand needs, specifically noting tight equipment supply and the need to collaborate early with equipment suppliers to prepare capacity.
TSMC also mentioned that it has dispatched teams to verify the progress of US data center construction and power supply conditions, incorporating these factors into capacity planning. This reflects that AI-related expansion requires not only chip demand but also supporting conditions such as data center projects and electricity. On the other hand, TSMC noted that consumer and price-sensitive markets remain affected by rising component prices and economic uncertainty, thus maintaining caution regarding business planning for related segments in 2027.
This indicates that capacity planning must simultaneously address customer demands, data center construction, and real-world conditions like electricity, while distinguishing between different end markets.
Tight Equipment Supply: Can Production Lines Run?
Equipment supply is a critical link in wafer fab expansion.
A wafer fab announcing a capital budget does not mean equipment is immediately available; nor does an equipment vendor expanding manufacturing capability guarantee that all clients can receive equipment according to plan.
In its Q3 results announced on October 14, 2026, ASML stated that customers continued to raise short- and medium-term equipment demand forecasts, with some clients wishing to place orders out to 2029 or even 2030. The company expects that, under current plans, supply in 2027 and 2028 will not fully meet customer demand and is researching ways to further boost capacity for 2028.
The same earnings release showed that ASML's Q3 net sales were EUR 11.7 billion, with 113 new lithography systems sold, up from 86 in Q2. The company also raised its full-year 2026 sales expectation to EUR 44–45 billion. The simultaneous occurrence of sales growth and insufficient supply indicates that the market contradiction is no longer just about order volume, but whether the equipment supply chain can deliver on time. TSMC also highlighted tool supply tightness and the need for early collaboration with equipment suppliers. Disclosures from both wafer fabs and equipment vendors confirm supply pressure from both ends.
According to disclosures by some domestic companies in convertible bond review materials, supply capacity is also under pressure in the back-end test equipment sector. Weice Technology stated that lead times for high-end testers used for AI are long, and supply became tight after AI business growth; the company has pre-purchased and locked in equipment needed for subsequent expansion. JHT Design, a tester handler manufacturer, indicated that rising demand for AI computing chips, HBM, and advanced packaging has driven order increases. As of the end of July 2026, the company had backlog orders of approximately CNY 253 million related to fundraising projects, covering most of its new capacity; the company also noted that existing capacity is tight, with downstream demand placing significant pressure on capacity.
Equipment expansion cannot be completed simply by adding production lines to a single factory. Key components, precision machining, software and control systems, system integration, and on-site services must all advance in sync. Wafer fabs also need to reserve facility infrastructure, cleanroom space, and equipment integration windows. Any delay in one link may postpone equipment acceptance and capacity ramp-up.
To judge whether expansion has truly progressed to the production line side, one must track equipment move-in and acceptance, availability of key process equipment, and production line uptime and yield ramp-up. Capital expenditures and equipment sales reflect investment and sales situations but do not themselves equal new wafer capacity.
Shifting from Computing Power Expansion Frenzy to Revenue and Cash Flow
Whether capacity expansion to meet AI computing power demand can sustain investment ultimately depends on whether computing power can be converted into revenue, cost savings, or customer business growth.
Alibaba's performance for the quarter ending June 2026, released in August 2026, showed AI cloud and computing service revenue of USD 7.1 billion, a year-on-year increase of 45%; adjusted EBITA for the cloud business was USD 830 million, up 133% year-on-year, with a margin of approximately 12%. In the same quarter, the company's capital expenditure approached USD 10 billion, a year-on-year increase of 75%, linking this investment to ongoing AI-related spending. Revenue and profit growth indicate that AI cloud services have formed quantifiable commercial income, while high capital expenditure shows that computing infrastructure is still being expanded.
Amazon disclosed in Q2 2026 that AWS AI business annualized run-rate revenue exceeded USD 25 billion, with year-on-year growth exceeding triple digits; meanwhile, the company's free cash flow over the past 12 months was negative USD 7.6 billion, primarily impacted by increased net purchases of property and equipment driven by AI-related investments. These two sets of data presented side-by-side show that while AI business revenue is growing rapidly, large-scale infrastructure investment is also impacting cash flow.
Microsoft disclosed in its FY2026 Q3 earnings call that its AI business annualized run-rate revenue reached USD 37 billion, a year-on-year increase of 123%; quarterly capital expenditure was USD 31.9 billion, with approximately two-thirds allocated to short-lived assets such as GPUs and CPUs. Microsoft also projected full-year 2026 capital expenditure of approximately USD 190 billion (forecast made during Q3), stating that GPU, CPU, and storage capacity remained constrained throughout 2026. Management expressed confidence in return on investment, citing strengthened demand signals, increased product usage, and improved platform efficiency.
Conclusion:
Existing disclosures present both demand growth and realization constraints: domestic IC production equipment revenue is growing, wafer fabs maintain high capital budgets, equipment integration demand forecasts are strengthening, and cloud vendors report AI business revenue growth. Simultaneously, equipment supply is tight, wafer fab construction and production ramp-up take time, and massive investments by cloud vendors have begun to impact cash flow.
AI is driving capacity expansion across the semiconductor supply chain, but expansion goals have shifted from merely adding equipment and factories to more complex coordination: equipment supply must keep pace, wafer fabs must be built on schedule, processes and yields must complete ramp-up, data centers require electricity and actual utilization, and AI services must continuously monetize.
Going forward, judgments on industry prosperity will shift more from "how much investment was announced" to "when capacity becomes available, when equipment is delivered, when customers validate, and when computing power is consistently paid for and used." Expansion plans are the starting point, equipment delivery is the middle stage, and stable output and return realization constitute the closed loop.