Maintains Overweight and Rmb1,408 target price; near-term earnings cut does not alter the long-term AI chip growth thesis

Overview

This report updates Cambricon Technology Corporation's earnings forecasts and bull, base, and bear risk-reward scenarios. Given weaker revenue growth in the second quarter of 2026, Morgan Stanley lowered 2026 estimated EPS by 3%; driven by pricing power, operating leverage, and margin improvement, it raised 2027 and 2028 estimated EPS by 3% and 1%, respectively. The report maintains an Overweight rating and Rmb1,408 target price, and continues to be positive on domestic generative AI infrastructure buildout, demand from major cloud service providers, and the company's strategic market position.

Core views

The core investment thesis includes: the company has deep cooperation with major cloud service providers, product-market fit has been validated, and MLU590 deployment provides strong order visibility; production is shifting toward SMIC, with yield still a near-term challenge, but gradual improvement is expected in the second half of 2026; the next-generation MLU690 is expected to launch in the fourth quarter of 2026, and a potential performance improvement of about 2.2x is expected to sustain technological competitiveness; current P/E and P/S ratios are high in absolute terms, but rapid growth, improved supply visibility, and the strategic position of domestic AI chips can provide valuation support.

Analysis framework

The report combines earnings forecast revisions, bull-base-bear scenario analysis, and 2026 estimated P/S multiples to assess risk-reward, and uses parameters such as cost of equity, long-term payout ratio, medium-term growth rate, and terminal growth rate to validate long-term valuation. Earnings data are mainly based on Morgan Stanley ModelWare, and marked items use consensus-estimate methodology.

Methodology notes

• Valuation methodsP/S multiple valuation

  Measures scenario values using P/S multiples corresponding to estimated 2026 revenue.

  The bull, base, and bear scenarios use 85x, 45x, and 23x estimated 2026 P/S, corresponding to Rmb2,660, Rmb1,408, and Rmb709.

• Scenario analysisBull-base-bear risk-reward framework

  Assesses upside and downside through different assumptions on revenue growth, market share, customer orders, and gross margin.

  The bull case assumes demand and share significantly exceed expectations; the base case assumes continued investment in domestic generative AI infrastructure; the bear case considers slowing capital expenditure, share loss, insufficient orders, and gross margin deterioration.

• Rating systemMorgan Stanley relative rating system

  Overweight means the risk-adjusted total return over the next 12 to 18 months is expected to be above the average level of the analyst's industry coverage universe.

  This rating is a recommended weight relative to the industry coverage universe and is not equivalent to a traditional absolute Buy rating.

• Financial forecastingMorgan Stanley ModelWare

  Morgan Stanley's proprietary model framework used to form financial forecasts and valuation inputs.

  Unless otherwise stated, the report metrics are based on this model; EPS data use the consensus-estimate method marked in the report.

Asset mapping & comparison

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

• Cambricon Technology Corporation (688256.SS)

  The A-share semiconductor and AI chip company directly covered by the report

  Strengths

  Deep cooperation with major cloud service providers, with MLU590 validating product-market fit; domestic AI infrastructure expansion brings demand; supply chain localization and next-generation product iteration are expected to improve supply and performance; scale effects and operating leverage support margin improvement.

  Weaknesses

  Revenue growth was somewhat soft in the second quarter of 2026, and domestic wafer manufacturing yield remains challenging; domestic wafer and HBM costs are relatively high; earnings and valuation are relatively sensitive to high-growth assumptions.

  Comparison

  P/E and P/S ratios are high in absolute terms, but the report believes its rapid growth, improving supply visibility, and strategic position as a key domestic AI chip supplier can provide partial valuation support.

  Risks

  Domestic AI capital expenditure and large-model commercialization fall short of expectations, leading customer orders are interrupted, market share declines, price competition intensifies, wafer costs rise, capacity utilization is insufficient, products are delayed, and valuation multiples contract.

Key data

• Rating and industry viewOverweight; industry view is AttractiveThe usual time horizon for the rating is 12 to 18 months.
• Target price and reference closing priceRmb1,408.00; Rmb1,199.93The reference closing price is as of August 7, 2026, and the target price implies about 17.3% upside.
• Risk-reward scenario valuesBull case Rmb2,660; base case Rmb1,408; bear case Rmb709Relative to the reference closing price, these imply about 121.7% upside, 17.3% upside, and 40.9% downside, respectively.
• Estimated EPS2025 Rmb4.9; 2026 Rmb11.3; 2027 Rmb19.0; 2028 Rmb24.6The 2026 forecast is lowered by 3%, while 2027 and 2028 forecasts are raised by 3% and 1%, respectively.
• Revenue CAGR scenariosBull case above 130%; base case 99%; bear case below 60%The forecast period is 2025 to 2028.
• Gross margin scenariosBull case above 60% in 2026 and 55% in 2027; base case 55% and 52%; bear case below 40% in both yearsKey influencing factors include scale effects, product mix, domestic wafer and HBM costs, price competition, and capacity utilization.
• Long-term valuation parametersCost of equity 8.4%; long-term payout ratio 40%; medium-term growth rate 16%; terminal growth rate 6.0%Cost of equity is based on a beta of 1.06, a risk-free rate of 2.0%, and an equity risk premium of 6.0%.
• Product roadmapMLU690 is expected to launch in the fourth quarter of 2026, potentially improving performance by about 2.2xTimely product launch and customer validation are important conditions for maintaining technological leadership and order growth.

Impact & implications

This revision reflects near-term pressure on revenue recognition and growth, but does not change the direction of medium- to long-term earnings upgrades. If major cloud service providers continue to expand generative AI infrastructure investment, SMIC production yields improve, and MLU690 enters mass production as planned, the company is expected to digest its current high valuation through high revenue growth, pricing power, and operating leverage. Conversely, slowing capital expenditure, discontinuous orders, or rising costs could pressure both valuation and earnings.

Risks

• Domestic AI capital expenditure or progress in large-model commercialization may fall below expectations, leading to a significant slowdown in revenue growth from 2025 to 2028.
• The company may fail to continuously secure large-scale orders from leading customers, or may suffer share loss in China's AI chip market.
• Yield improvement during the production transfer to SMIC may be slower than expected, affecting supply, delivery capability, and unit costs.
• Rising domestic wafer manufacturing and HBM costs, intensified price competition, or insufficient capacity utilization could cause 2026-2027 gross margins to fall below expectations.
• Delays in the launch, mass production, or customer validation of MLU690 could weaken product competitiveness and technological leadership.
• Current valuation multiples are high; if growth, margins, or supply improvements fail to materialize, valuation could contract significantly.
• Morgan Stanley disclosed that as of July 31, 2026, it beneficially owned 1% or more of a class of common equity securities of Cambricon Technology Corporation, and investors should be aware of potential conflicts of interest.

What to watch

• Progress in improving SMIC production yield and domestic supply chain stability in the second half of 2026.
• Whether MLU690 can be launched as planned in the fourth quarter of 2026 and achieve about 2.2x performance improvement.
• The deployment scale of MLU590 in relevant workloads at major cloud service providers and the continuity of subsequent large orders.
• Domestic generative AI infrastructure capital expenditure and the pace of large-model commercialization.
• Whether gross margins in 2026 and 2027 can reach the base-case levels of 55% and 52%, respectively.
• Whether 2026-2028 EPS forecasts continue to be driven by revenue growth, pricing power, and operating leverage.
• Changes in the share price relative to the Rmb1,408 target price and the Rmb709 to Rmb2,660 risk-reward range.