TSMC Q2 Earnings Call Preview: Can AI Chip Demand Still Support High Growth?

When you look at TSMC’s Q2 earnings call, the key question is not just whether quarterly revenue grows. The more important question is whether AI chip demand can continue to support high growth, high gross margin, and elevated capital spending. Ahead of TSMC’s July 16 Q2 earnings call, the market will focus on three signals: whether Q2 guidance is met, whether HPC and advanced-node demand remain strong, and whether CoWoS and N2 capacity expansion stays on track. If these signals point in the same direction, AI demand can remain TSMC’s main growth driver; if margin pressure and capex concerns rise, the market may reassess the quality of that growth.

Key Takeaways

• TSMC’s Q2 focus is guidance execution, AI demand, and margin quality.
• HPC, 3nm, N2, and CoWoS are the main signals for growth quality.
• AI demand remains strong, but capex returns are still a valuation variable.
• High gross margin has support, but overseas fabs and N2 ramp-up add pressure.
• Investors should track earnings signals, supply-chain constraints, and trading costs together.

What Really Matters Ahead of TSMC’s Q2 Earnings Call?

The first thing to watch is not a single revenue number, but whether Q2 performance aligns with the high end of TSMC’s previous guidance, whether gross margin remains strong, and whether management continues to confirm AI and HPC demand. For global investors, TSMC’s Q2 earnings call is an important window into the health of the AI chip cycle because it connects Nvidia, AMD, Broadcom, hyperscaler custom silicon, advanced nodes, and advanced packaging.

TSMC’s Q1 earnings release provided a clear Q2 baseline: revenue was guided to US$39.0 billion to US$40.2 billion, gross margin to 65.5% to 67.5%, and operating margin to 56.5% to 58.5%. These ranges will serve as the main reference points for judging Q2 performance. Revenue near the upper end would suggest strong demand execution, while gross margin staying high would indicate that advanced-node mix, utilization, and cost control are still working in TSMC’s favor.

The monthly revenue trend also matters. TSMC’s 2026 monthly revenue data showed that January-to-May revenue increased 30.0% year over year, with April and May revenue up 17.5% and 30.1%, respectively. Monthly revenue does not replace full quarterly financial results, but it helps investors track the momentum behind Q2 revenue expectations. In the AI chip supply chain, monthly revenue can shape expectations for the quarter, full-year growth commentary, and customer pull-in behavior.

Summary: TSMC’s Q2 earnings call matters because it turns the AI story into financial evidence. You need to look at revenue, gross margin, HPC share, advanced nodes, CoWoS capacity, and capital spending together. Revenue growth alone shows that orders are strong, but it does not fully prove growth quality. When margin, utilization, and capex commentary also stay strong, the case becomes clearer that AI chip demand is translating into sustainable financial performance.

Can AI Chip Demand Still Support TSMC’s High Growth?

AI chip demand remains the main support for TSMC’s high growth, but you should not judge it simply by saying “AI is hot.” A better framework is to ask whether HPC revenue remains dominant, whether advanced-node revenue stays high, whether hyperscaler capex continues, and whether custom ASIC demand broadens the customer base. As long as these signals do not weaken materially, AI should remain the core variable in TSMC’s growth curve.

HPC Revenue Share Is the Demand Thermometer

In TSMC’s Q1 presentation, the company disclosed that HPC accounted for 61% of Q1 revenue, while smartphones represented 26%, IoT 6%, and automotive 4%. This shows that TSMC has moved further from a business once more closely tied to smartphone cycles toward a growth structure led by high-performance computing and AI infrastructure.

The node mix is equally important. In Q1, 3nm, 5nm, and 7nm contributed 25%, 36%, and 13% of wafer revenue, respectively, while 7nm and below accounted for 74% of wafer revenue. AI GPUs, AI accelerators, hyperscaler ASICs, HBM base dies, and high-speed networking chips all tend to rely on advanced nodes and high-end packaging. The higher the HPC share, the more TSMC’s revenue structure reflects the AI data center investment cycle.

Demand Is Not Just About GPUs

The market often reduces AI chip demand to Nvidia GPUs, but TSMC’s demand base is broader. AI servers also use AMD accelerators, custom chips linked to Broadcom, hyperscaler ASICs, networking chips, CPUs, I/O chips, and HBM-related base die demand. Reuters’ coverage of the AI capex chain noted that cloud service providers rely on AI chip designers such as Nvidia, AMD, and Broadcom, while continuing to expand data center investment.

This matters for TSMC because multi-customer, multi-product demand reduces dependence on one chip model. Even if a specific GPU line enters a transition period, AI ASICs, inference chips, networking chips, and HBM-related demand can still consume advanced-node and advanced-packaging capacity.

Training and Inference Both Drive Compute

TSMC management noted in its Q1 earnings call transcript that the shift from generative AI’s question-and-answer model toward agentic AI’s command-and-action model could continue to increase token consumption and compute demand. The investment implication is that AI chip demand is not only about model training; it is also about broader inference deployment.

You can divide AI chip demand into four streams:

• Training demand: Drives high-end GPUs, HBM, and CoWoS demand.
• Inference demand: Supports wider deployment of AI accelerators in data centers.
• Custom ASICs: Broaden hyperscaler demand for advanced nodes.
• Supporting chips: Networking, storage, I/O, and power management also upgrade.

Summary: AI chip demand can still support TSMC’s high growth, but the strength of that support depends on whether demand has expanded from a single GPU boom into a broader HPC ecosystem. HPC share, advanced-node revenue, hyperscaler capex, AI ASICs, and CoWoS capacity are the key signals to watch. If these indicators continue to move together, TSMC’s growth is not just a short-term order surge; it is a structural benefit from the AI infrastructure cycle.

Why 3nm, N2, and CoWoS Drive This Earnings Preview

The key issue in this earnings preview is not only whether Q2 revenue meets guidance. It is whether 3nm, N2, and CoWoS form a continuous capacity roadmap. 3nm represents current AI and HPC order absorption, N2 represents the next advanced-node transition, and CoWoS determines whether AI chips can move from wafer manufacturing into system-level packaging. Together, they shape how long TSMC’s high-growth cycle can last.

3nm Is the Main Battleground for Current Orders

3nm is not only a smartphone node. It is also becoming an important capacity resource for AI and HPC chips. In the Q1 earnings call, TSMC management said the company is increasing N3 capacity to meet strong AI application demand, including new 3nm capacity in Tainan, 3nm plans for the second Arizona fab, and 3nm plans for the second Japan fab.

This means 3nm capacity tightness is not just a short-term issue. AI GPUs, hyperscaler ASICs, flagship smartphone chips, and some HBM-related demand may all compete for advanced-node capacity. If management continues to emphasize strong N3 demand, tight capacity, and high customer visibility, the market will be more likely to view AI growth as more than a one-off pull-in cycle.

N2 Represents the Next Advanced-Node Growth Curve

N2 is central to TSMC’s next advanced-node story. Management previously said N2 entered high-volume production in the fourth quarter of 2025 and is being driven by both smartphone and HPC/AI applications. The significance of N2 is not just better performance; it is also evidence that customer roadmaps are moving toward the next generation of chips.

N2 also brings near-term cost pressure. TSMC previously expected the initial N2 ramp to dilute full-year 2026 gross margin by about 2 to 3 percentage points in the second half of the year. During the earnings call, you should watch how management discusses N2 yield, capacity, customer demand, and margin impact. If N2 demand is strong and dilution remains manageable, the long-term growth story becomes more complete.

CoWoS Is the AI Chip Delivery Bottleneck

For AI chips, wafer manufacturing is only the first step. High-end GPUs and AI accelerators must also integrate logic chips, HBM, and other components at high density. TSMC’s CoWoS® is a key packaging technology for AI, HPC, and supercomputing because it enables logic chips and HBM to be connected through an interposer and high-density interconnects.

TrendForce’s analysis of advanced packaging bottlenecks noted that rapid AI demand growth has pressured 3nm-to-2nm wafers, 2.5D/3D advanced packaging, CoWoS, substrates, equipment, and materials. These bottlenecks can support TSMC’s utilization and pricing power in the short term, but they can also limit customer delivery schedules.

Summary: 3nm, N2, and CoWoS are the most important technology themes to track in TSMC’s Q2 earnings call. 3nm determines current revenue leverage, N2 determines the next process upgrade, and CoWoS determines AI chip delivery capacity. You can view them as one continuous chain: customer demand first enters advanced nodes, then moves into advanced packaging, and finally becomes deliverable AI systems. If this chain remains tight, TSMC’s high-growth logic has stronger support.

Can TSMC Keep Gross Margin at a High Level?

TSMC’s gross margin still has near-term support because the advanced-node mix is strong, HPC share is high, and utilization remains elevated. But medium-term pressure is also clear, mainly from the initial N2 ramp, overseas fab dilution, materials and gas costs, and currency movements. You should not look only at revenue growth; you also need to judge whether growth is being delivered with healthy margin quality.

In Q1, TSMC’s gross margin reached 66.2%, and operating margin was 58.1%. Management attributed the gross margin performance to cost improvement, higher capacity utilization, and favorable foreign exchange. For Q2, gross margin was guided to 65.5% to 67.5%, with a midpoint of about 66.5%. If Q2 gross margin comes in near the high end, it would suggest that advanced nodes and utilization remain strong. If gross margin misses market expectations, it may point to cost, currency, or overseas fab dilution pressure.

The first margin support comes from product mix. The higher the share of HPC and advanced-node revenue, the easier it is for TSMC to sustain high utilization and a favorable pricing structure. Tight supply in 3nm, 5nm, and CoWoS also makes customers more willing to secure capacity earlier, improving revenue visibility.

The pressure comes from expansion itself. Early-stage advanced-node ramps usually involve higher depreciation and ramp-up costs, while overseas fabs bring higher construction and operating costs. TSMC management previously said overseas fab dilution was around 2 to 3 percentage points in the early stage and could expand to 3 to 4 percentage points in the coming years. This means global capacity expansion can improve supply security and customer confidence, but it also changes the cost structure.

Pricing pass-through is another key issue. Reuters reported that TSMC management remained positive about the continuation of the AI boom while also noting constraints across the supply chain. For investors, the question is not whether price increases are guaranteed, but whether customers are willing to pay more for advanced nodes, advanced packaging, and reliable delivery.

Summary: TSMC’s high gross margin is not supported by AI demand alone. It depends on advanced-node mix, utilization, cost control, and pricing power working together. Stronger AI demand can help offset pressure from N2, overseas fabs, and materials costs, but faster expansion also brings depreciation and operating-cost pressure. Gross margin guidance, N2 dilution, overseas fab costs, and customer price acceptance will be central to judging the quality of TSMC’s high growth.

What Risks Sit Behind AI-Led High Growth?

TSMC’s AI growth story remains strong, but the key risk is not whether AI demand exists. The real question is whether that demand can keep turning into customer capex, deliverable capacity, and stable margins. You should focus on three risk categories: hyperscaler capex returns, supply-chain bottlenecks, and export controls plus customer concentration. These variables affect order timing and the valuation premium investors are willing to assign.

The first risk is hyperscaler capex returns. AI data centers require massive investment in servers, GPUs, networking, power, and cooling. As long as generative AI and agentic AI applications continue to create real demand, cloud providers will have stronger reasons to increase AI chip orders. But if application revenue grows more slowly than infrastructure spending, the market may worry that capex payback periods are stretching.

The second risk is supply-chain bottlenecks. CoWoS, HBM, substrates, equipment, and materials can all affect AI chip delivery. Bottlenecks have a double effect on TSMC. On one hand, tight supply supports utilization and pricing power. On the other hand, if downstream customers cannot assemble and deliver AI servers smoothly, order timing may also be affected.

The third risk is export controls and customer concentration. TSMC’s [2025 annual report](https://investor.tsmc.com/sites/ir/annual-report/2025/2025 Annual Report_E.pdf) discusses the impact that regulatory restrictions, major customer changes, industry cycles, and market conditions can have on the business. AI chip export restrictions, customer program changes, and regional policy adjustments can all reshape demand. Geopolitics should not be viewed as a simple positive or negative; it is more accurately a source of disruption to order structure, customer priorities, and supply-chain layout.

Summary: TSMC’s AI growth story is powerful, but the real test is whether it can move through supply-chain and capex cycles. You should understand risk through the lens of demand quality: whether customers keep building data centers, whether CoWoS and HBM supply can match demand, whether export controls alter order paths, and whether overseas fab costs compress margin. If these risks remain manageable, AI demand provides stronger support for TSMC. If several risks worsen at the same time, the market may reassess TSMC’s valuation premium.

How Should Individual Investors Read TSMC’s Q2 Earnings Call?

You can read TSMC’s Q2 earnings call through a three-layer framework: first, whether Q2 guidance is met; second, whether full-year revenue and capex commentary remain strong; and third, whether the market has already priced in the quality of AI-driven growth. This framework is more useful than watching the share price alone, because TSMC’s valuation is shaped by revenue, margin, advanced-node cycles, customer demand, and global tech-sector sentiment.

The first layer is guidance execution. If Q2 revenue is close to the upper end of US$40.2 billion, demand execution would look strong. If gross margin is close to the upper end of 67.5%, utilization, cost control, and product mix would appear healthy. If operating margin remains strong, it would show that higher revenue is not being fully offset by higher costs.

The second layer is the full-year message. TSMC management previously expected 2026 revenue in U.S. dollar terms to grow more than 30% and indicated that capital expenditure could be close to the high end of the US$52 billion to US$56 billion range. Higher capex signals confidence in future demand, but it also raises investor attention toward future depreciation and free cash flow pressure. You need to judge whether high capex is being used to secure real demand or chase an overheated cycle.

The third layer is trading and allocation cost. Many investors track the AI chip chain through TSM ADR, U.S. semiconductor stocks, semiconductor ETFs, or related Hong Kong-listed names. Beyond company fundamentals, actual trading costs also affect the holding experience. U.S. stock trading costs usually include more than commission; they can also include platform fees, external agency fees, trading activity fees, and other charges. Taking Biya U.S. stock trading fees as an example, U.S. stock trading commission is US$0, while platform fees, external agency fees, and other charges are subject to the fee schedule and order page. This information only introduces public market information, trading rules, and fee structures, and does not constitute investment advice.

Summary: Individual investors should not read TSMC’s Q2 earnings call only as “good or bad earnings.” The better question is whether the growth has quality. Revenue near the high end, high gross margin, strong AI and HPC demand, and a clear capex return path would form a stronger setup. Revenue growth with margin pressure, rising capex, and weaker customer-demand commentary would call for more caution. The more you connect financial indicators with supply-chain signals, the less likely you are to be misled by short-term share-price moves.

If you follow TSMC, Nvidia, AMD, ASML, Broadcom, AI storage, and the AI server supply chain, it can be useful to track earnings calls, monthly revenue, valuation moves, and trading costs in one framework. Biya is a global multi-asset trading wallet that supports U.S. stocks, Hong Kong stocks, and digital asset trading. You can also use U.S. stock search to track related names and check available services in the Biya App. Availability depends on your location, identity verification result, platform rules, and applicable laws and regulations; before investing, you should assess your own risk tolerance and review fee details carefully.

FAQ

What metrics matter most in TSMC’s Q2 earnings call?

The most important metrics are revenue, gross margin, operating margin, HPC share, advanced-node revenue, CoWoS capacity, N2 ramp-up, and full-year guidance. Together, they help you judge whether AI demand is being converted into both revenue and profit.

How does AI chip demand affect TSMC revenue?

AI chip demand affects TSMC revenue through GPUs, hyperscaler ASICs, HBM base dies, high-speed networking chips, and inference chips. These products usually require advanced nodes and advanced packaging, increasing the importance of HPC revenue and 3nm/5nm capacity.

Why does CoWoS capacity matter for AI chip delivery?

CoWoS capacity matters because high-end AI accelerators need dense integration of logic chips and HBM. Even if wafer manufacturing goes smoothly, limited advanced-packaging capacity can still constrain AI server chip delivery.

What are the main risks to TSMC’s high gross margin?

The main risks are the initial N2 ramp, overseas fab dilution, materials and gas costs, currency fluctuations, and customer pricing negotiations. Strong AI demand can cushion these pressures, but it cannot fully remove cost-cycle risk.

How can individual investors track TSM ADR risks?

Individual investors should track TSMC earnings calls, U.S. dollar movements, AI customer demand, U.S. tech valuation sentiment, and trading costs. TSM ADR pricing can also be affected by U.S. market liquidity and global risk appetite toward technology stocks.

Which chip stocks can be affected by TSMC’s earnings call?

TSMC’s earnings call can affect Nvidia, AMD, Broadcom, ASML, memory stocks, advanced-packaging suppliers, and AI server supply-chain names. The direction depends on management’s commentary on AI demand, advanced nodes, CoWoS capacity, and capital expenditure.