Hygon Information benefits from the rollout of the Sugon 8000 domestic AI supercluster, and Morgan Stanley maintains a positive view

Overview

This report focuses on investor feedback following the launch of Hygon Information after the release of Sugon 8000. Sugon 8000 is an AI supercluster with 100,000 accelerator cards, using domestic core chips such as Hygon C86 CPU, Hygon DCU/GPGPU AI accelerators, HySW scale-up switch ASIC, PCIe 5.0 switch chip, 400/800G switch ASIC, and 400G NIC ASIC. The report believes this event validates Hygon's system-level participation capability in China's domestic AI infrastructure.

Core views

The core views are: first, Hygon is not a pure GPU company, and its "CPU+GPU" combination allows it to serve demand for general-purpose servers, AI training, and AI inference simultaneously; second, Hygon's largest shareholder, Sugon, holds 27.96%, and its ecosystem spans chips, computing, storage, networking, cooling, applications, and services, helping Hygon expand share in domestic CPU and GPU markets; third, in the era of agentic AI, the importance of CPUs in workflow control and system-level coordination increases, and Hygon's deployment base in security-sensitive sectors such as government and finance positions it well to capture demand for data sovereignty, compute isolation, and trusted infrastructure.

Analysis framework

The report argues for Hygon's investment value through industry event validation, TAM estimates, share assumptions, valuation parameters, and risk scenarios. At the industry level, it focuses on Sugon 8000's deployment in training, inference, and multi-industry applications; at the market level, it compares the growth of China's server CPU TAM and AI GPU TAM; at the company level, it evaluates Hygon's room for share gains in both CPU and GPU markets; at the valuation level, it uses the Morgan Stanley ModelWare framework and provides a target price.

Methodology notes

• Valuation methodsMorgan Stanley ModelWare

  Target price and earnings forecast framework

  The report states that unless otherwise noted, financial metrics are based on the Morgan Stanley ModelWare framework, using assumptions such as an 8.0% cost of equity, 50% payout ratio, 25.0% mid-term growth rate, and 5.0% terminal growth rate.

• Market sizingTAM analysis

  Comparison of CPU and AI GPU demand

  The report compares China's server CPU TAM and AI GPU TAM, arguing that the server CPU market including agentic AI demand can grow at a 31% CAGR, above the 23% CAGR of AI GPU TAM.

• Thematic analysisagentic AI demand framework

  The system coordination value of CPUs in multi-agent architectures

  The report believes that as AI systems evolve toward multi-agent architectures, the role of CPUs in workflow control, system-level coordination, data sovereignty, compute isolation, and trusted infrastructure will strengthen.

Asset mapping & comparison

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

• Hygon Information Technology Co., Ltd. (688041.SS)

  The core equity covered by the report, benefiting from domestic AI computing power, CPU localization, and growth in AI GPU demand.

  Strengths

  It has a CPU+GPU combination, participates in the core chip ecosystem of Sugon 8000, has a deployment base in security-sensitive sectors such as government and finance, and can expand its share of computing infrastructure with the help of the Sugon ecosystem.

  Weaknesses

  Growth and valuation of the CPU business still need market acceptance; domestic advanced process yields, capacity, and the GPU competitive landscape will affect earnings realization.

  Comparison

  Compared with most local GPU companies, Hygon covers both general-purpose CPUs and AI accelerators; the report believes its CPU business should not be assigned a valuation multiple lower than that of other GPU companies.

  Risks

  Downside risks include intensified price competition among local GPU companies, weaker-than-expected AI demand in China, and slower-than-expected improvement in domestic advanced-node foundry yields and capacity build-out.

• Sugon 8000

  An important industry catalyst and validation scenario for Hygon's chip ecosystem.

  Strengths

  It supports a scale of 100,000 accelerator cards, covers training and inference, supports more than 400 mainstream models, and has completed optimization of more than 300 supercomputing-AI integrated applications across more than 20 fields.

  Weaknesses

  The performance cited in the report is mostly based on vendor releases or conference feedback; commercialization revenue conversion and sustained orders still need to be tracked.

  Comparison

  As China's first domestic AI supercluster with 100,000 cards, it demonstrates improved completeness of local AI infrastructure in core chips, networking, storage, and system integration.

  Risks

  If actual customer adoption, model training scale, or compute utilization falls short of expectations, the revenue uplift for Hygon may be weaker than the market expects.

Key data

• Stock ratingOverweightThe most positive stock rating in Morgan Stanley's relative rating system, corresponding to better risk-adjusted returns over the next 12-18 months than the average stock under industry coverage.
• Industry viewAttractiveThe report assigns an Attractive view to the Greater China Technology Semiconductors sector.
• Target priceRmb480.00Target price history shows the target price was 480 on 2026-07-03.
• Current stock priceRmb353.00Closing price as of 2026-07-10.
• Implied upside36%Based on the Rmb480 target price and the current price of Rmb353.
• China CPU TAM forecastUS$42bn by 2030The report expects China's CPU TAM to rise to US$42 billion by 2030.
• China server CPU growth31% CAGRServer CPU growth assumption including agentic AI demand.
• China AI GPU TAM forecastUS$91bn by 2030The report expects China's AI GPU TAM to reach US$91 billion by 2030.
• AI GPU TAM growth23% CAGRUsed in the report for comparison with server CPU growth.
• China agentic AI CPU TAMUS$17bn by 2030The report estimates China's agentic AI CPU demand accounts for 21% of global demand.
• Hygon China server CPU share18%The chart indicates Hygon's market share in China's server CPU market reaches 18%.
• Hygon AI GPU share5% by 2028The chart indicates Hygon's share rises to 5% by 2028.
• Sugon shareholding27.96%The report says Sugon is Hygon's largest shareholder.
• Sugon 8000 scale100,000 acceleratorsSugon 8000 is an AI supercluster with 100,000 accelerator cards.

Impact & implications

In terms of investment implications, the launch of Sugon 8000 is not just a single-product event, but also a validation of the usability of Hygon's domestic CPUs, GPUs, and interconnect chips in large-scale AI infrastructure. If China's AI demand, localization procurement, and agentic AI applications continue to expand, Hygon could benefit simultaneously from higher server CPU share, higher AI accelerator penetration, and system-level order opportunities brought by the Sugon ecosystem. However, whether the high valuation can be realized depends on demand strength, yield improvement and advanced process capacity ramp-up, and whether price competition among local GPU vendors intensifies.

Risks

• Intensifying price competition among local GPU companies may compress margins in Hygon's AI accelerator business.
• China's AI demand may be weaker than expected, affecting TAM growth assumptions for both CPUs and GPUs.
• Slower-than-expected improvement in domestic advanced-node foundry yields and capacity build-out may constrain product ramp-up.
• U.S. executive orders, export controls, and compliance restrictions may affect related securities trading or the industry supply chain.
• The high valuation depends on sustained high growth and share gains; if delivery falls short of expectations, valuation multiples may come under pressure.

What to watch

• Follow-up customer orders, utilization, and commercialization revenue conversion for Sugon 8000.
• Progress in clock speed, performance, and security capability improvements of Hygon's next-generation CPU.
• Whether Hygon's share in China's server CPU market approaches the 18% assumed in the report.
• Whether Hygon's AI GPU share rises to 5% by 2028 as expected in the report.
• The actual pull from China's agentic AI applications on CPU demand.
• Changes in domestic advanced process capacity, yields, and supply-chain security policies.
• Price competition and product performance iteration among local GPU vendors.