China's AI demand and localization support Cambricon's growth; Goldman Sachs raises target price to Rmb1,841 and maintains Buy

Overview

The report examines Cambricon's Chinese AI chip demand, localization trends, expansion into application industries, recovery in R&D capabilities, quarterly inventory signals, earnings forecast revisions, and valuation. Goldman Sachs believes these factors will drive the company's future growth. Despite the impact of cloud service providers' bargaining power and increased R&D investment on margins, it maintains its Buy rating and raises its target price.

Core views

The report's primary conclusion is that strong Chinese demand for AI chips, together with the country's push for semiconductor self-sufficiency and control, will provide the principal support for Cambricon's future growth. The customers and applications for the company's chips have expanded beyond foundation model providers, server and AI software vendors, and cloud service providers to include energy, education, finance, telecommunications, hospitals, and internet companies. Financial institutions mainly use the chips to optimize business operations and service efficiency, while internet companies use them for foundation models, multimodal AI, search, and recommendations. By 1H26, Cambricon's chips had been adapted to mainstream AI applications and deployed in large-scale training, while also entering scenarios such as smart grids, intelligent mines, smart cities, intelligent transportation, and smart retail, including semantic search, structured video analysis, visual inspection of goods, power equipment operations and maintenance, workplace safety monitoring, traffic incident detection, customer flow identification, store inspections, and theft and loss prevention. The report believes that expansion into additional application industries will support shipment demand, while continued product specification upgrades will increase the blended average selling price. Products and R&D constitute the second major growth driver. Cambricon is developing next-generation intelligent processor microarchitectures and instruction sets to optimize the training and inference of natural language processing, video and image generation, and vertical-domain foundation models, while improving programming flexibility, ease of use, performance, power consumption, and chip area. The company continues to enhance computing power, storage capacity, and bandwidth, and has completed adaptation to leading local foundation models including GLM, DeepSeek, Qwen, Kimi, and MiniMax. The company was added to the US Entity List in December 2022 and experienced talent attrition in 2023. After the number of R&D personnel fell to a low of 727 in 1H24, it recovered to 1,007 in 1H26. The report believes the recovery of the R&D team will support future chip iterations and growth. Quarterly inventory changes are an important basis for the report's assessment of the near-term revenue trajectory. Revenue grew 8% QoQ in 2Q26, below the 53% increase in 1Q26, with the report attributing the slowdown to customer procurement schedules and wafer supply. Its historical observations indicate that slow inventory growth generally corresponds to slower QoQ revenue growth in the following quarter: inventory declined 9% QoQ in 1Q26, followed by revenue growth of only 8% QoQ in 2Q26. In contrast, inventory surged 83% QoQ in 2Q26, mainly due to raw materials such as wafers, leading the report to expect a notable acceleration in QoQ revenue growth in 3Q26. Raw-material inventory increased 779% YoY in 1H26, which the report also views as a signal of advances in leading-edge processes at local foundries and progress in China's generative AI industry. Based on stronger Chinese AI demand and localization trends, Goldman Sachs raises its 2026—2030 EBITDA forecasts for Cambricon by 1%, 41%, 69%, 71%, and 82%, respectively, mainly due to higher revenue forecasts. The new revenue forecasts for 2026—2028 are Rmb23,980.4mn, Rmb91,261.6mn, and Rmb172,469.6mn, respectively, versus previous forecasts of Rmb23,661.8mn, Rmb39,106.5mn, and Rmb62,158.3mn. EBITDA forecasts for the same period are Rmb6,744.1mn, Rmb16,004.0mn, and Rmb31,225.8mn. The new EPS forecasts for 2025—2028 are Rmb4.90, Rmb10.63, Rmb22.58, and Rmb43.40, respectively, versus previous forecasts of Rmb3.29, Rmb9.78, Rmb15.90, and Rmb25.45. The report expects average YoY EPS growth of 111% in 2027—2028. Rapid expansion also entails margin trade-offs. The report expects product specification upgrades to increase average selling prices over the next several years. However, considering the growing share of Chinese cloud service provider customers, their strong bargaining power, and the industry practice of reducing average selling prices for large-volume purchases, Goldman Sachs lowers its gross margin assumptions. The company also needs to increase R&D investment covering AI chip design; chip-based GPU modules, backplanes, motherboards, servers, and other hardware; as well as foundational software such as AI application development platforms and toolkits. Forecasts show revenue growth of 453.2%, 269.1%, 280.6%, and 89.0% in 2025—2028, respectively, and EBITDA growth of 445.6%, 208.6%, 137.3%, and 95.1%. EBITDA margins over the same period are forecast at 33.6%, 28.1%, 17.5%, and 18.1%, reflecting substantial revenue growth alongside a temporary decline in margins. On valuation, Goldman Sachs continues to use a discounted EV/EBITDA methodology based on 2030E EBITDA. The target EV/EBITDA multiple is derived from the relationship between the trading EV/EBITDA of Chinese semiconductor peers and their next-year EBITDA growth and EBITDA margins. After updating peer data and Cambricon's forecasts, the company's relevant EBITDA growth metric increases from 21% previously to 29%, while its EBITDA margin declines from 31% previously to 19%. Consequently, the target EV/EBITDA multiple is reduced from 43x to 27x. Goldman Sachs applies 27x to 2030E EBITDA and discounts the result back to 2027 using an unchanged 12.7% cost of equity, deriving a 12-month target price of Rmb1,841, up from Rmb1,614.77 previously. The new target price corresponds to 82x 2027E P/E, broadly consistent with the company's average P/E of 84x since July 2024. Relative to the current share price of Rmb1,035 stated in the report, the target price implies 77.9% potential upside, supporting Goldman Sachs' decision to maintain its Buy rating.

Analysis framework

The report first assesses revenue drivers based on Chinese AI demand, the direction of localization policies, and customer application scenarios, and then evaluates product supply capabilities by examining processor iterations, model adaptation, and changes in R&D staffing. It subsequently uses the historical relationship between quarterly inventory and next-quarter revenue to assess the operating trajectory for 3Q26 and accordingly revises its revenue, EBITDA, EPS, gross margin, and R&D expense forecasts. Finally, it determines the target multiple based on the relationship between Chinese semiconductor peers' EV/EBITDA, growth rates, and margins, and discounts 2030 EBITDA back to 2027 to derive a 12-month target price.

Methodology notes

• Valuation MethodEV/EBITDA valuation

  Discounted EV/EBITDA Valuation

  The report applies a target EV/EBITDA multiple of 27x to 2030E EBITDA and then discounts it back to 2027 at a 12.7% cost of equity, deriving a 12-month target price of Rmb1,841.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Dual Drivers from Chip Demand and Average Selling Prices

  The report decomposes revenue growth into demand and shipment support from the expansion of application scenarios and an increase in the blended average selling price from product specification upgrades, while also considering the price pressure from large-volume purchases.

• Cycle and Business Conditions Framework

  Quarterly Inventory-Revenue Leading Relationship

  Based on the historical relationship between inventory changes and QoQ revenue performance in the following quarter, the report views the substantial increases in 2Q26 inventory and raw-material inventory as leading signals of stronger revenue growth in 3Q26.

• Valuation MethodEV/EBITDA valuation

  Peer Fundamentals and Valuation Multiple Comparison

  The target multiple references the relationship between the trading EV/EBITDA of Chinese semiconductor peers and their next-year EBITDA growth and EBITDA margins, and is recalibrated based on Cambricon's updated growth rate and margin.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Cambricon（688256.SS）

  The report believes the company directly benefits from Chinese demand for AI chips, semiconductor localization, expansion into application industries, and product specification upgrades.

  Strengths

  Its customers span multiple industries, and its chips have been adapted to several leading local foundation models. The number of R&D personnel has recovered, while product development encompasses chips, hardware, and foundational software.

  Weaknesses

  Chinese cloud service providers have strong bargaining power, and large-volume purchases may reduce average selling prices. The company needs to continue increasing R&D investment and has previously experienced talent attrition.

  Comparison

  The target EV/EBITDA is determined based on the relationship between Chinese semiconductor peers' valuations and their next-year EBITDA growth and margins. The target price implies 82x 2027 P/E, close to the company's historical average of 84x since July 2024.

  Risks

  Constraints on wafer supply, slower-than-expected development of cloud chips, and stronger-than-expected competition in cloud chips.

Key data

• 12-Month Target PriceRmb1,841.00Raised from Rmb1,614.77
• Current Price and Potential UpsideRmb1,035.00；77.9%The current price stated in the report and the upside implied by the target price
• RatingBuyBuy maintained; rating effective since April 22, 2025
• 2026—2030 EBITDA Forecast Revisions+1% / +41% / +69% / +71% / +82%Mainly driven by higher revenue forecasts
• 2026—2028 Revenue ForecastsRmb23,980.4mn / Rmb91,261.6mn / Rmb172,469.6mnPrevious forecasts were Rmb23,661.8mn / Rmb39,106.5mn / Rmb62,158.3mn, respectively
• 2026—2028 EBITDA ForecastsRmb6,744.1mn / Rmb16,004.0mn / Rmb31,225.8mnCorresponding to rapid earnings expansion over the forecast period
• New 2025—2028 EPS ForecastsRmb4.90 / Rmb10.63 / Rmb22.58 / Rmb43.40Previous forecasts were Rmb3.29 / Rmb9.78 / Rmb15.90 / Rmb25.45
• 2025—2028 EBITDA Margins33.6% / 28.1% / 17.5% / 18.1%Margins are under pressure from cloud service providers' bargaining power and R&D investment
• 2Q26 Revenue Growth+8% QoQBelow +53% QoQ in 1Q26
• Quarterly Inventory Changes1Q26 -9% QoQ；2Q26 +83% QoQThe 2Q26 increase mainly came from raw materials such as wafers, supporting the report's positive view on QoQ revenue performance in 3Q26
• 1H26 Raw-Material Inventory Growth+779% YoYThe report believes this also reflects advances in leading-edge processes at local foundries
• Number of R&D PersonnelIncreased from 727 to 1,007Recovered from the 1H24 low to 1H26
• Target EV/EBITDA27xPreviously 43x; applied to 2030E EBITDA
• Cost of Equity12.7%Used to discount the 2030 valuation back to 2027 and unchanged from previously
• 2027 P/E Implied by Target Price82xBroadly consistent with the average P/E of 84x since July 2024
• Market Capitalization and Enterprise ValueRmb601.9bn / Rmb599.1bnUS dollar values are $89.6bn / $89.1bn, respectively

Impact & implications

The report believes that expansion into additional application industries, adaptation to leading local models, and the recovery of the R&D team make Cambricon better positioned to capture opportunities arising from Chinese demand for AI computing power and localization. The increase in raw-material inventory in 2Q26 also supports its expectation of accelerating revenue in 3Q26. The resulting revenue upgrades substantially increase medium-term EBITDA and EPS forecasts, although the bargaining power of large cloud service providers, the customary price reductions for bulk purchases, and higher R&D investment mean that margins will not expand in tandem with revenue. Although the valuation multiple is reduced from 43x to 27x, the higher earnings forecasts still lift the target price to Rmb1,841.

Risks

• Wafer supply may be constrained; Cambricon was added to the US Entity List in December 2022.
• Cloud chip development may progress more slowly than expected.
• Competition in the cloud chip market may be stronger than expected.