Analysis of the Impact of China's AI Technological Progress on US Semiconductor Valuation and Investment Competition Landscape

Now, based on the collected data and analysis, I will provide you with a complete investment research report.

---

Currently, the global artificial intelligence industry is in a period of profound transformation. China’s rapid progress in the AI field is reshaping the global semiconductor competition landscape, which has a profound impact on the valuation and investment competition of US tech stocks (especially in the semiconductor and AI chip sectors)[1]. This report aims to systematically analyze the multi-dimensional impact of China’s AI technological progress on US semiconductor companies, and provide decision-making references for investors.

Company	Market Capitalization (USD 100 million)	P/E Valuation	AI Business Exposure	China Revenue Exposure
NVIDIA	45,008	45.87	Extremely High (95/100)	15%
AMD	3,297	106.93	High (70/100)	10%
Broadcom	16,300	72.63	Extremely High (85/100)	20%
Intel	1,805	85.24	Medium (40/100)	25%
Cambricon (China)	5,832	310.35	Extremely High (100/100)	-

Data Source: Jinling API Real-Time Data[0]

---

In the field of AI training chips, China is accelerating its catch-up with US technological leadership[1][2]:

:

• Ascend 910B has approached the performance level of NVIDIA H100
• Ascend 910C has a computing power of up to 800TFLOPS, adopting 7nm process
• The Cloud Matrix 384 super node released by Huawei has surpassed the corresponding NVIDIA product in some performance indicators[1]
• Jensen Huang publicly acknowledged the competitiveness of Huawei Ascend in specific scenarios

:

• 2023
  : The gap in training capability is about 2 years
• 2025
  : The gap narrows to about 1.2 years
• 2027 Forecast
  : It is expected to control the gap within 1 year

In the inference chip field, China has achieved important breakthroughs[2][3]:

• After optimization for models such as DeepSeek-R1 and Alibaba Tongyi, the Token throughput efficiency of
  Baidu Kunlun Core P800
  and
  Alibaba PPU
  has exceeded that of NVIDIA H20 (the US export-controlled “downgraded” chip)
• Inference chips are more sensitive to costs, and Chinese manufacturers have advantages in cost performance
• It is expected that the localization rate of inference chips will exceed 50% in 2026

Chinese AI chip companies are accelerating the construction of independent software ecosystems[2][3]:

• Huawei’s CANN ecosystem continues to improve, attracting developers to migrate from CUDA
• Moore Thread’s MUSA architecture claims that developers do not need to rewrite all code
• Muxi Semiconductor’s self-developed MXMACA software stack achieves a high degree of compatibility with CUDA
• The convergence of model architectures (Transformer/Diffusion) reduces the difficulty of adaptation for domestic chips

---

Time	Policy Event	Impact Level
October 2022	BIS First Round of Export Controls	High
October 2023	H20 Chip Ban	High
December 2024	Additional Restrictive Measures	Medium-High
July 2025	H20 Export Resumption Approval	Medium
December 2025	Approval for H200 Exports to China	Medium

Data Source: Comprehensive Compilation from Online Searches[4][5]

Dong Ting, Associate Research Fellow at the Center for Strategic and Security Studies of Tsinghua University, pointed out that the recent US policy adjustments “are a recalibration of policies, more predatory”[5]:

1. Short-term Goal
   : Prevent companies such as NVIDIA from completely losing competitiveness in the Chinese market
2. Mid-term Consideration
   : Maintain the technological generation gap while retaining ecological influence
3. Long-term Strategy
   : Implement precise control within the framework of “small yard, high fence”

:

• Approximately 15-20% of its revenue comes from the Chinese market[6]
• The H20 chip was once the main product in the Chinese market
• Policy uncertainty increases the difficulty of corporate planning

:

• The proportion of revenue from China is relatively low (10-25%)
• Facing similar policy risk exposure
• Custom chip business is relatively less affected

---

:

• NVIDIA may permanently lose approximately USD 10-15 billion in annual revenue from China
• AMD is less affected, but its growth space is limited
• Broadcom’s network chip business faces substitution threats

:

• Global AI investment sentiment is disturbed by geopolitics
• Enterprise customers diversify supply chains, increasing procurement costs
• R&D investment needs to be increased to maintain technological leadership

Current valuations of US semiconductor stocks are at historical highs[6][7]:

Company	Current P/E	Historical Average P/E	Valuation Status
NVIDIA	45.87	~35	Slightly High
AMD	106.93	~40	Extremely High
Broadcom	72.63	~25	Relatively High

:

1. Intensified Chinese competition substantially affects market share
2. Narrowing of technological leadership advantages
3. Profit margin pressure (price war risk)
4. Slowdown in AI capital expenditure due to macroeconomic recession

The global AI chip market is transitioning from “NVIDIA monopoly” to “multi-polar competition”[4][6]:

:

• NVIDIA
  : 85% → 70% (gradual decline)
• AMD
  : 10% → 17% (steady increase)
• Chinese Chips
  : 2% → 10% (rapid rise)
• Others
  : 3% → 3%

---

:

• Technological Leadership: Continuous iteration of Blackwell and Rubin architectures
• Perfect Ecosystem: Deep CUDA ecosystem moat
• Order Backlog: Approximately USD 500 billion in orders, with USD 300 billion expected to be recognized in 2026[7]

:

• Rise of Chinese competitors (Huawei, Cambricon)
• Competition from self-developed chips by large-scale cloud vendors (Google TPU, AWS Trainium)
• Valuation digestion pressure

: Overweight
: Based on DCF analysis, the valuation center is approximately USD 185-210

:

• Significant performance improvement of the MI300 series
• Strong growth in data center business
• Execution advantages under the leadership of Lisa Su

:

• P/E ratio as high as 106.93, valuation relies on sustained high growth[0]
• Technological gap with NVIDIA is still large
• Competitive pressure from Intel’s recovery

: Hold
: Valuation digestion may last 6-12 months

:

• In-depth cooperation with Google TPU, stable custom chip business
• Release of VMware synergy effects
• Stable market position of network chips (Tomahawk)

:

• High dependence on a single customer (Google)
• China business risk exposure of approximately 20%
• Valuation has reflected most favorable factors

: Overweight

:

• Revenue in the first three quarters of 2025 reached RMB 4.607 billion, a year-on-year surge of 2386%[8]
• Net profit of RMB 1.605 billion, successfully turning profitable
• Plans to use RMB 2.778 billion in capital reserves to cover losses

:

• P/E ratio as high as 310.35, extremely dependent on growth expectations[9]
• P/S ratio of approximately 104, reflecting market enthusiasm for domestic substitution
• Its stock price once surpassed Moutai to become the “King of Stocks” in A-shares

:

• High valuation bubble risk
• Core technological breakthroughs still take time
• Highly dependent on domestic policy support

:

• Strongest full-stack independent and controllable capabilities
• Leading technology of Cloud Matrix 384 super node
• Dominant position in the domestic intelligent computing center market

:

• Restricted by advanced manufacturing processes (7nm vs NVIDIA’s 4nm)
• Production capacity bottlenecks restrict scale expansion
• International market expansion is limited

---

:

1. NVIDIA
   : Accumulate on dips, target price USD 190-210
   • Supporting Factors: Launch of Rubin architecture, strong demand for Blackwell
   • Risk Factors: Intensified competition from China, pressure on valuation multiples

2. AMD
   : Cautiously overweight, pay attention to valuation digestion
   • Pay attention to the release progress of the MI350 series
   • Set a stop-loss level at USD 105

3. Broadcom
   : Stable allocation, defensive choice
   • Benefits from AI infrastructure expansion
   • Volatility is lower than pure GPU companies

:

1. Gradually reduce allocation to companies with high exposure to China
2. Pay attention to semiconductor companies shifting to non-Chinese markets
3. Accumulate leading Chinese AI chip companies on dips (position controlled within 5%)

:

• Progress of China’s 7nm and below process breakthroughs
• Adoption rate of domestic AI chips in mainstream large models
• Evolution of US export control policies

:

1. Global Diversification
   : Accept the norm of “multi-polar competition”
2. Equipment and Materials
   : Pay attention to the strategic value of Japanese and European semiconductor equipment
3. Packaging and Testing
   : High-performance computing drives demand for advanced packaging

:

• Allocate a certain proportion of gold ETFs (geopolitical hedge)
• Pay attention to Japanese semiconductor revival theme funds
• Appropriately allocate Southeast Asian technology sectors

---

Risk Event	Probability	Potential Upside	Beneficiary Targets
NVIDIA’s Rubin architecture exceeds expectations	Medium	+20-30%	NVDA
China’s AI development falls short of expectations	Medium-High	+10-15%	NVDA, AMD
Global AI capital expenditure continues to expand	Medium	+15-25%	Entire Industry
Further relaxation of export controls	Low	+5-10%	NVDA

Risk Event	Probability	Potential Downside	Affected Targets
Breakthrough in China’s training chips	Medium-High	-15-25%	NVDA
AI bubble bursts	Medium	-30-40%	Entire Industry
Escalation of geopolitical conflicts	Medium	-20-30%	Entire Industry
Global economic recession	Medium	-25-35%	Entire Industry

Scenario	Probability	NVDA Stock Price by End of 2026	Key Assumptions
Optimistic	25%	$230-250	Rubin architecture is highly successful, competition from China is limited
Neutral	50%	$180-210	Competitive landscape is stable, technological gap is maintained
Pessimistic	25%	$140-165	China’s breakthrough accelerates, valuation compression occurs

---

1. Technological Gap Continues to Narrow
   : China’s AI chips have approached or even surpassed “downgraded” US products in the inference field, and the gap in the training field has narrowed to 1-2 years.

2. Limited Effectiveness of Export Controls
   : US policies have instead accelerated China’s domestic substitution process, forming a “double-edged sword” effect.

3. Valuation Faces Reassessment Pressure
   : The formation of a multi-polar competition pattern may compress the valuation multiples of US semiconductor companies, especially NVIDIA’s high premium.

4. Profound Changes in Investment Landscape
   : The global AI chip market is transitioning from “NVIDIA monopoly” to “US-China competition + US domestic competition”.

Company	Rating	Core Logic	Risk Warning
NVIDIA	Overweight	Technological leadership + AI infrastructure demand	Competition from China, valuation digestion
AMD	Hold	High growth potential + relatively high valuation	Intel’s recovery, price war
Broadcom	Overweight	Stability + custom chip advantages	Customer concentration
Cambricon	Cautious	Domestic substitution + policy support	Valuation bubble, technological gap

1. Short Term
   : Maintain core allocation to NVIDIA, add positions on dips; set reasonable stop-loss for AMD
2. Mid Term
   : Pay attention to pullback opportunities of leading Chinese players such as Huawei and Cambricon (control positions)
3. Long Term
   : Accept the multi-polar competition pattern, diversify allocation of global semiconductor assets

---

[1] Southern Metropolis Daily - “Domestic AI Chips: The Inference Track Takes Off, Who Can Break the Cambricon Myth Again?” (https://news.southcn.com/node_17a07e5926/0543a47541.shtml)

[2] Ascend Open Source Ecological Zone - “Global AI Frontier Dynamics on December 21, 2025” (https://ascendai.csdn.net/694891dfbf6b0e4b285d4422.html)

[3] 36Kr - “Domestic AI Chips, Explosive Growth” (https://m.36kr.com/p/3601821550937088)

[4] Sina Finance - “AI Chips 2025: Giant Showdown, Power Restructuring” (https://finance.sina.com.cn/cj/2025-12-31/doc-inhesxnp3665369.shtml)

[5] Wenxuecity - “Approving Exports of High-End Chips, US Adjusts Tech Competition Strategy Against China” (https://www.wenxuecity.com/news/2025/12/26/126463971.html)

[6] CMoney - “Nvidia Sprinting into the AI Era, Will Challenge USD 6 Trillion Market Capitalization in 2026!” (https://cmnews.com.tw/article/cmoneyairesearcher-c79617f9-ea68-11f0-9cec-29b3c9c6166b)

[7] CMoney - “Is Nvidia Stock Worth Investing in 2026? Experts Analyze Future Potential and Risks!” (https://www.forecastock.tw/article/cmoneyairesearcher-376c9161-ddf0-11f0-a3b9-5a0aacf7f0e5)

[8] Securities Times - “Cambricon Plans to Use Nearly RMB 2.8 Billion in Capital Reserves to Cover Losses, Net Profit Turns Positive in the First Three Quarters” (https://www.stcn.com/article/detail/3542180.html)

[9] Investing.com - “Latest Price and Quotes for Cambricon Technologies Corp Ltd (688256)” (https://cn.investing.com/equities/cambricon-technologies-corp-ltd)

[10] Wall Street CN - “How Can China’s Computing Power Grow Stronger?” (https://wallstreetcn.com/articles/3762510)

[11] Lexology - “2025 International Trade Compliance Annual Review and 2026 Outlook” (https://www.lexology.com/library/detail.aspx?g=0afcbfa0-3676-4f45-885f-ba989c1384a9)

[12] IngStart - “New Interpretation of US-China Trade Detente in 2025” (https://www.ingstart.com/blog/37797.html)

[0] Jinling API Data - Real-Time Quotes and Financial Data of US Semiconductor Companies