1. With Q4 earnings ahead, will AI orders drive TSMC to new highs?
Likely, but not unconditionally.
TSMC’s share price has been propelled by AI-related demand for advanced logic and accelerator chips. Analysts and trading data suggest AI infrastructure spending is a central driver behind forward earnings expectations and capacity utilisation. Foundry revenue has expanded strongly on the back of AI chip orders and products built on TSMC’s leading nodes (3nm and 2nm) have high margins and stickier customer demand.
Key considerations:
AI backlog and advanced nodes: TSMC’s 2nm and advanced CoWoS packaging are core to AI chips for Nvidia, AMD and others, sustaining both revenue and margin leadership.
Guidance matters: The next leg higher will depend on guidance around capacity utilisation, pricing power, and forward AI orders. With capacity tight through 2027 and expansion plans in play, the market will scrutinise any signs of order book growth slowing.
Valuation and risk: Shares have rallied sharply and a meaningful disappointment in demand outlook — particularly outside the AI segment — could temper gains.
In summary: AI orders can support fresh highs if TSMC confirms strong future demand and provides bullish guidance, but valuation, guidance and macro factors will shape next moves.
2. With AI demand remaining strong, how much upside for memory chips remains?
Memory chips (DRAM, HBM, NAND) are experiencing a pricing and supply cycle boom right now, largely due to AI infrastructure demand and constrained production. Multiple reports indicate that DRAM and related memory prices have surged, with some manufacturers forecasting very strong profitability driven by AI workloads.
Key points:
Strong pricing environment: Memory chip prices have rebound strongly compared with the cyclical trough, benefiting manufacturers such as Samsung and SK Hynix.
AI infrastructure pull: High-bandwidth memory (HBM) and other specialised memory types are in demand for AI accelerators — often commanding premium pricing.
Cycle considerations: Historically, memory goes through boom and bust cycles. While the current phase seems strong, the sustainability of price increases and production expansion will influence the remaining upside. Continued demand from hyperscale and AI infrastructure suggests further gains are plausible over the next 12-18 months, but demand destruction at the end-product level (e.g. PC and smartphone markets) could temper long-term growth.
Upside summary: There is meaningful remaining upside in memory chips given current supply tightness and AI demand, but the market is cyclical and increasingly dependent on longer-term infrastructure build-outs rather than short-cycle consumer demand.
3. In the current global semiconductor landscape, which do I favour: foundry vs memory chips?
Both segments benefit from the AI boom, but they differ in structural positioning and risk profiles:
Foundry (e.g. TSMC, other advanced fabs)
Strengths
Structural demand from generative AI and high-performance computing.
Persistent supply tightness at leading nodes (3nm/2nm) and proprietary packaging (CoWoS).
Broader customer diversification (AI chip designers, consumer SoCs).
Risks
High capital intensity and geopolitical risk (Taiwan–China tensions).
Expansion timelines and ROI on new fabs.
Memory (e.g. DRAM, HBM)
Strengths
Immediate pricing power due to constrained supply and AI-driven demand.
Short-cycle revenue spikes linked to rapid data centre infrastructure growth.
Risks
Classic cyclical behaviour — periods of oversupply can quickly compress margins.
End-market elasticity: memory pricing and demand depend on data centre capex, consumer electronics demand and inventory cycles.
Preference framework
Longer-term structural growth (5+ years): Foundry tends to have more predictable demand from electronics and AI architecture build-outs, with higher barriers to entry and less pronounced boom-bust cycles.
Short- to medium-term cyclical upside: Memory can outperform when pricing and capacity align with strong AI demand, but carries more volatility.
Conclusion: For a diversified exposure to semiconductor growth tied to AI, foundry has a stronger structural profile given its centrality to next-generation chips and broader customer base. Memory offers significant cyclical gains in the near term, especially in a constrained supply environment, but with higher risk of volatility.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
