UBS remains positive on China AI hardware, but warns the cycle is approaching elevated levels

Overview

This is a China equity strategy report whose core question is how far China’s AI technology hardware upcycle is from its peak. The report argues that although crowded positioning, high valuations, increasing IPOs, and rising capital expenditure all suggest the cycle is approaching its later stage, strong revenue and earnings growth, order visibility extending into 2027, continued expansion of AI data center construction, and ongoing retail fund participation should still support performance this year. The report recommends remaining overweight hardware technology, but warns that 2027 may be more volatile and that positioning should become more concentrated in leading companies with stronger technological barriers, pricing power, and deeper penetration into the global supply chain.

Core views

The report’s core views include: first, China AI hardware technology still has strong near-term fundamentals, and earnings momentum is sufficient to support continued outperformance of the sector; second, signs of a cyclical peak are increasing, including high fund holdings and trading concentration, valuations near historical highs, more IPOs, and rising capital expenditure; third, Chinese AI companies are catching up with global capabilities in some segments, but global leaders still have the edge in earnings growth, cash flow, ROE, and margin outlook; fourth, domestic power equipment and hyperscale cloud providers are more attractive in valuation relative to global peers; fifth, 2027 will be a key year to test whether small- and mid-cap companies and niche segments can truly deliver on AI promises.

Analysis framework

The report combines a top-down assessment of cycle positioning with a bottom-up breakdown of the supply chain: it first uses earnings growth, revenue growth, contract liabilities, inventory, earnings revisions, valuation, fund holdings, trading concentration, and capital expenditure to judge the stage of the AI hardware cycle, then breaks the AI data center supply chain into segments such as power equipment, cooling, MLCC, power semiconductors, EDA/IP, semiconductor equipment, wafer foundries, HBM, advanced packaging, CCL, PCB/substrates, and others, while listing Chinese listed companies and global comparable companies.

Methodology notes

• Strategy cycle assessmentAI technology hardware upcycle indicator framework

  Use contract liabilities, forward revenue growth, forward earnings growth, inventory, and earnings revisions to measure fundamental momentum, and use valuation, positioning concentration, trading concentration, and capital expenditure to measure crowding and late-cycle risk.

  The report points out that AI technology share prices have the highest correlation with forward-looking indicators such as contract liabilities, forward revenue and earnings growth, and inventory, while showing lower correlation with macro variables such as interest rates, credit, and exchange rates.

• Supply chain analysisAI supply chain framework from silicon to solutions

  Break AI data center construction into segments including energy, power, thermal management, chip design, equipment, manufacturing, memory, packaging, materials, PCB, and interconnects.

  The report believes more powerful computing chips are driving stepwise increases in demand for power, cooling, data transmission, advanced packaging, HBM, and high-end materials, significantly raising industry entry barriers.

• Relative valuation comparisonComparison of China AI hardware with global peers

  Compare Chinese and global AI companies on revenue, earnings, cash flow, ROE, margins, valuation, and share price performance.

  The report finds that revenues of both Chinese and global AI companies have roughly doubled since 2021, but global peers have stronger earnings growth and cash flow; China AI technology hardware used to trade at a discount, but is no longer cheap now.

Asset mapping & comparison

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

• China AI technology hardware sector

  Core overweight direction

  Strengths

  Strong earnings momentum, high order visibility, continued AI data center construction, and high retail participation.

  Weaknesses

  Valuations are close to historical highs, and fund holdings and trading concentration are elevated.

  Comparison

  Earnings revisions are stronger relative to the overall China market, but ROE, margins, and cash flow are still weaker relative to global AI leaders.

  Risks

  Revenue growth peaks, capacity expansion loosens supply-demand balance, and 2027 earnings delivery falls short of expectations.

• Leading companies in optical modules, memory, GPU, CCL, and semiconductor equipment

  Key AI sub-industries favored by the report

  Strengths

  High technological barriers and strong pricing power, making them more likely to benefit from deeper penetration into the global AI supply chain.

  Weaknesses

  Popular segments are crowded trades, and some valuations already fully reflect optimistic expectations.

  Comparison

  The report believes leaders are more likely than smaller technology companies to prove their ability to deliver on AI in 2027.

  Risks

  A slowdown in global demand, customer capex adjustments, technological substitution, or supply chain constraints.

• NAURA Tech, Montage, CR Micro, JCET

  Top picks listed in the report

  Strengths

  They correspond respectively to key segments such as semiconductor equipment, memory/chip-related, power semiconductors, and advanced packaging.

  Weaknesses

  Their segments are highly exposed to cyclical and valuation volatility.

  Comparison

  The report prefers these leaders with industry position and supply-chain penetration capability over small-cap companies that rely purely on sector beta.

  Risks

  Order timing, capacity rollout, gross margin, and changes in global competition.

• Small- and mid-cap technology companies

  High-beta assets benefiting from sector beta and demand spillover

  Strengths

  They may benefit from risk appetite and spillover demand optionality during the upcycle.

  Weaknesses

  Technological barriers, customer quality, and earnings sustainability may be weaker than those of leading companies.

  Comparison

  They may rise with the sector in the short term, but in 2027 they will need more clearly to prove real AI business delivery capability.

  Risks

  Earnings misses, valuation pullback, and reversal in liquidity and crowded trading.

• Domestic power equipment and hyperscale cloud providers

  Related beneficiaries of AI data center construction

  Strengths

  AI data centers are boosting demand for reliable power, energy storage, transformers, UPS, power conversion, and cloud infrastructure.

  Weaknesses

  Some segments have long construction cycles and are affected by grid connection, equipment supply, and capex timing.

  Comparison

  The report believes domestic power equipment and hyperscale cloud providers have more attractive valuations relative to international peers.

  Risks

  Power buildout progresses more slowly than expected, capex cuts, and margins lower than overseas businesses.

Key data

• China AI hardware forward earnings growthAbout 80%The report says that historically, when forward earnings growth exceeds 30%, AI technology hardware usually outperforms the broader market by 10%-20% annually.
• Change in earnings forecasts over the past 3 monthsAI technology raised 15%, broader market cut 2%This shows a clear divergence in fundamentals between China’s technology sector and the overall market.
• Order visibilityExtended to end-2027The report says demand for AI technology components remains strong, while upstream product supply is still limited.
• Performance of global AI companies relative to Chinese AI companiesOutperformed by about 130% over the past few yearsBut in 2026, China-related AI companies outperformed global peers by about 17%.
• Performance of the AI sector since the Deepseek point in early 2025About 215% excess return and 16x P/E re-ratingThis is stronger than the roughly 100% excess return and more than 19 percentage points of valuation expansion seen in past 4G, 5G, and cloud computing cycles.
• Historical performance after revenue growth peaksUsually underperforms the broader market by about 2%The report believes that if earnings growth remains high, the scale of any short-term correction may still be manageable.
• Correlation between contract liabilities and share price0.75This is the highest among the indicators shown in the report, indicating that forward demand signals such as orders and advance receipts have strong explanatory power for share prices.
• Correlation between forward revenue growth and share price0.74The report believes forward-looking revenue and earnings indicators still support recent AI hardware performance.
• Correlation between China 10-year interest rate and share price-0.41Macro interest-rate indicators have lower correlation than forward-looking fundamental indicators.

Impact & implications

The investment implication is that investors should not simply exit China AI technology hardware in the short term because of rising valuations and crowding; instead, portfolios should gradually shift from broad beta trades toward leaders and companies that truly have the capability to deliver on AI. Earnings growth, order visibility, and data center capital expenditure remain supportive factors; at the same time, valuations are no longer cheap, positioning is concentrated, capacity is expanding, and slower growth in 2027 means sector volatility and differentiation are likely to increase.

Risks

• Fund holdings and trading concentration are close to historical highs; if risk appetite declines, drawdowns may increase.
• Domestic AI technology hardware valuations are no longer cheaper than global peers, limiting room for further multiple expansion.
• Once revenue growth peaks, historical experience suggests the next 6 months may see consolidation or relative weakness.
• Rising capital expenditure and more IPOs may signal the cycle is entering a later stage, and future supply expansion may pressure earnings.
• Global AI leaders still have stronger cash flow, ROE, and margins, and there is uncertainty around Chinese companies catching up.
• 2027 will test whether smaller companies and niche segments can truly deliver on AI demand; those that fail may be re-rated by the market.

What to watch

• Whether China AI hardware forward revenue and earnings growth can continue to stay above 30%.
• Whether contract liabilities, inventory, and order visibility continue to support demand strength.
• Whether fund holdings, trading concentration, and new share financing continue to rise.
• The pace of new AI data center capacity, power equipment orders, and capital expenditure.
• Changes in the valuation premium of domestic AI hardware relative to global peers and the CSI 300.
• When industry growth slows in 2027, the earnings divergence between leaders and smaller companies.
• Orders, gross margins, and global supply-chain progress of key leaders such as NAURA Tech, Montage, CR Micro, and JCET.