Original source:BofA Global ResearchFirst published on Unhidea:Original source date:
Key takeaway
BofA believes ChangXin Technology can sustain high profitability through capacity expansion and domestic Chinese demand, but that view depends on equipment availability, yields and HBM progress, and its earnings forecasts are below market consensus.
Why did BofA initiate coverage with a Buy rating after high profits?
ChangXin Technology’s gross margin and operating margin in the second quarter of 2026 were approximately 88% and 83%, respectively. The company disclosed preliminary revenue of 265 billion to 270 billion yuan for the first three quarters; on that basis, BofA forecasts full-year revenue of approximately 397 billion yuan, up from 61.8 billion yuan in 2025.
BofA believes HBM demand for AI servers can absorb new DRAM supply across the industry, while large cloud providers locking in chips through long-term agreements can also help support prices in 2027. ChangXin mainly serves domestic Chinese customers and stands to benefit from local demand for memory and data centers.
The bank assigned a Buy rating and a price target of 65 yuan, calculated by applying a price-to-earnings multiple of 16 to average earnings per share of 4.07 yuan for 2027–2028; the share price listed in the report was 43 yuan.
Why might ChangXin’s average selling price still fall amid a strong industry cycle?
BofA expects ChangXin’s average DRAM selling price to rise 9% year over year in 2027, but fall 17%, 12% and 11%, respectively, in 2028–2030. By contrast, it expects the global average DRAM price to rise 34% in 2027 and remain broadly stable in the following years.
The difference stems from the customer and product mix: in 2025, 60% of ChangXin’s DRAM revenue came from mobile applications such as phones and 27% from servers; its five largest customers accounted for 74% of company revenue. BofA believes consumer demand is more price-sensitive and customers have stronger bargaining power.
The year-over-year increase in the full-year average price in 2027 is also affected by the low base in the first half of 2026; BofA actually expects the average price to soften slightly each quarter in 2027.
How can capacity expansion offset price pressure, and what constrains it?
BofA estimates the company’s wafer capacity at around 300 thousand wafers per month at the end of 2026, rising to 470 thousand by the end of 2028 and 670 thousand by the end of 2030. Capacity expansion and process upgrades together support its forecast of annual bit shipment growth of 34% to 39% in 2027–2028.
This expansion path assumes that new facilities are completed and that the company can still obtain critical equipment and services or draw on equipment inventories. Further restrictions on advanced lithography, equipment components and technical support could slow capacity expansion and process upgrades.
The company has announced mass production using its fifth-generation DRAM process; BofA expects yields of 70% to 85% in 2027–2028, below the more than 90% yield of older processes. The new 4F2 architecture could improve chip density but may also bring higher production costs and execution risks.
How long can high margins and cash flow last?
BofA expects the operating margin to fall from about 80% in 2026 to 73% in 2028, while remaining high. Fixed-asset depreciation accounted for about 68% of cost of sales in 2025; the bank expects cost of sales per GB to rise 23% year over year in 2027, mainly due to higher depreciation and the ramp-up of new processes.
The key to maintaining high profits is for shipment growth to absorb added costs: BofA expects bit output growth to broadly catch up with depreciation growth from 2028, with unit costs stabilizing, if process upgrades and the HBM ramp proceed smoothly.
BofA forecasts capital expenditure of 100 billion and 120 billion yuan in 2027 and 2028, respectively, and operating cash flow of 442 billion and 509 billion yuan, enough to cover expansion investment. At the end of the first half of 2026, the company had approximately 143 billion yuan in cash and equivalents and approximately 140 billion yuan in debt; the bank’s projected future cash flow could also support higher dividends.
Not all profits accrue to shareholders of the listed company: through voting-rights arrangements, ChangXin consolidates two plants in Hefei and Beijing in which it holds approximately 31% to 32%. In its model for the second half of 2026 through 2028, BofA assumes non-controlling interests account for 35% of total net profit; the actual share was 28% in the first half of 2026.
What other key assumptions surround HBM and valuation?
HBM is high-bandwidth DRAM used for AI. Media reports say ChangXin has delivered small quantities of HBM3 and HBM3E samples to potential domestic customers; BofA expects HBM to account for 3% and 11% of the company’s revenue in 2027 and 2028, respectively, but production at commercial scale has yet to be demonstrated.
BofA’s earnings-per-share forecasts for 2027 and 2028 are 3.85 yuan and 4.30 yuan, respectively, approximately 20% and 27% below the market consensus listed in the report, mainly because of more conservative selling-price assumptions.
The target valuation is above the target price-to-earnings multiples of 9 to 11 that BofA assigns to global memory peers, on the grounds that ChangXin is still rapidly expanding capacity; it is also below the 30 to 40 multiples for Chinese semiconductor manufacturing peers, reflecting the cyclical volatility of the DRAM business.