TSMC Stock Price Forecast - TSM at $336: Is Taiwan Semiconductor’s AI Capex Super-Cycle Still a Buy?

NYSE:TSM – AI Infrastructure Core Asset Trading Around $336

NYSE:TSM – What The Current Price Around $336 Is Really Pricing In

NYSE:TSM trades around $336.39 today, with an intraday range of $331.37–$336.55, a 52-week range of $134.25–$351.33, and a market capitalization of roughly $1.48 trillion. The market is no longer valuing TSM as a cyclical foundry; at this level it is clearly priced as a core, quasi-monopolistic supplier of AI infrastructure. The stock trades on about 32x trailing earnings, low-20s forward P/E depending on whose 2026 EPS you use, and about 14.5x price-to-sales, which only makes sense if you assume that high-20s to 30% revenue CAGR and very high margins are sustainable through the AI super-cycle rather than peaking in the next few quarters.

NYSE:TSM – Earnings, Margins And Operating Leverage At Record Levels

For the December 2025 quarter, Taiwan Semiconductor reported revenue of roughly TWD 1.05 trillion, an increase of 20.45% year-on-year, while net income jumped 40.57% to TWD 505.74 billion. That pushed the net profit margin to 48.35%, up about 16.7 percentage points versus the prior year. Earnings per share reached TWD 19.50, up 34.95%, and EBITDA climbed to TWD 733.22 billion, an increase of 23.49%. Operating expenses were tightly controlled at TWD 88.76 billion, growing only 2.41% while revenues grew more than 20%, so incremental revenue is converting to profit at an accelerating rate. At this point you are looking at a foundry business delivering almost 50% net margins, which is closer to a software-like profile than a traditional semiconductor manufacturer and explains why TSM can support both a premium multiple and a massive capex load simultaneously.

NYSE:TSM – Balance Sheet, Cash Flow And Funding A USD 52–56 Billion Capex Run-Rate

On the balance sheet, cash and short-term investments total about TWD 3.07 trillion, up 26.71% year-on-year. Total assets stand at TWD 7.93 trillion (+18.55%), with total liabilities at TWD 2.47 trillion (+2.48%) and total equity at TWD 5.46 trillion. With a price-to-book around 1.57x, the market is not overpaying for the tangible and strategic asset base relative to the cash returns it is generating. Cash generation remains robust. Cash from operations reached TWD 725.51 billion, up 16.98%, while cash used in investing was TWD –365.96 billion and cash from financing was TWD –107.69 billion. Even with heavy investment, net change in cash was positive at TWD 297.10 billion, rising 23.36%, and free cash flow of TWD 223.09 billion was only 6.40% lower year-on-year despite clearly higher capex tied to AI. That profile – strong operating cash flow, positive net cash build, and still-solid free cash flow – is what allows TSM to sustain an annual capex guidance in the USD 52–56 billion range and still maintain balance sheet flexibility, while offering a modest ~1.0% dividend yield as a secondary consideration.

NYSE:TSM – Revenue Mix: Advanced Nodes And High-Performance Compute Dominate The Story

The growth and margin structure are driven by where NYSE:TSM earns its money. Advanced nodes now dominate. Recent disclosures show around 77% of revenue from N7 and below, with N3 contributing about 28% of sales, N5 about 35%, and N7 around 14%. Legacy nodes have become a cash annuity rather than a driver of valuation. On the end-market side, about 55% of revenue comes from high-performance computing, 32% from smartphones, and around 5% each from IoT and automotive, with the remainder in smaller categories. Practically, that means the stock is now levered to datacenter AI and compute intensity rather than to the smartphone replacement cycle. Nvidia’s datacenter GPUs, AMD accelerators, cloud TPUs, and leading smartphone SoCs all run through TSM’s advanced nodes. That concentration in high-value wafers is what supports both the revenue CAGR targets and the elevated gross margin guidance.

NYSE:TSM – Process Roadmap: N3 Maturity, N2 Ramp And A16 Extend The AI Super-Cycle To 2028–2029

The process roadmap is the backbone of the super-cycle thesis. On N3, TSM has already scaled capacity into the 150k wafers per month range and expects to stretch that to roughly 180k wafers per month in 2026 and potentially 250k wafers per month in 2027, fuelled by strong demand from AI GPUs, custom accelerators and high-end smartphones, plus incremental contribution from the Arizona Fab 2 once N3 is online in the U.S. On N5, capacity is expected to gradually decline from around 180k to 150k and 130k wafers per month as customers migrate to N3 and later N2, but N5 remains critical for platforms such as Nvidia’s H200 and other devices still within export-compliant envelopes for China. The real inflection is N2. Capacity is forecast to grow from about 35k wafers per month in 2025 to roughly 130–140k wafers per month in 2026, and potentially 200k wafers per month in 2027, anchored by Fab 20 in Hsinchu and Fab 22 in Kaohsiung. That N2 capacity is expected to host next-wave smartphone chips like Apple’s A20 and M6, Android flagships from Qualcomm and MediaTek, and, after a lag, AI CPUs/GPUs and ASICs from AMD, Google (TPU v9), AWS (Trainium 4) and Meta (MTIA 4). Above N2 sits A16, optimized for high-performance compute with backside power delivery, offering around 8–10% extra speed or 15–20% power reduction at similar density versus N2P. Future platforms like Nvidia’s Feynman and later OpenAI designs are expected to tap A16. When you combine N3 maturity, N2 ramp and A16 on top, the picture is a continuous AI-driven node progression that supports ~30% revenue growth in 2026, management’s ~25% USD revenue CAGR target through 2029, and high-50s CAGR for AI datacenter revenue rather than a near-term peak.

NYSE:TSM – Advanced Packaging Moat: CoWoS Today, CoWoP And CoPoS As The AI Chokepoint

The bottleneck in AI hardware today is as much packaging as lithography, and NYSE:TSM controls that choke point. CoWoS (Chip-on-Wafer-on-Substrate) is the current workhorse that lets TSM tightly couple large GPU dies with stacks of HBM in thermally manageable, high-bandwidth packages. CoWoS capacity is projected to rise from around 75k wafers per month in 2025 to about 120k in 2026 and 170k in 2027, driven by converting older fabs and ramping dedicated advanced packaging facilities. At the same time, TSM is developing CoPoS (Chip-on-Panel-on-Substrate) and CoWoP (Chip-on-Wafer-on-PCB) to reduce dependence on ABF substrates, cut costs and support larger package reticles. CoWoP, in particular, mounts a combined GPU–HBM–interposer module directly onto a high-end PCB, bypassing ABF supply bottlenecks and improving thermal and cost structure. From a competitive standpoint, Intel is pushing EMIB as a cheaper alternative and is courting U.S. customers, while Samsung is adding capacity at advanced nodes. But TSM still commands more than 90% of the high-end advanced packaging market, and is now building advanced packaging capacity in Arizona to give U.S. hyperscalers a domestic option. The combination of advanced nodes plus a dominant packaging platform is the lock-in: customers are buying a turnkey compute + packaging + thermal solution that is very hard for rivals to replicate at scale.

NYSE:TSM – Joule-Per-Token Economics And Value-Capture Pricing Power

What differentiates NYSE:TSM in this cycle is not only node leadership but how it prices that leadership. With N2 and later A16, TSM is explicitly targeting Joule-per-token economics: more tokens processed per watt of electricity, and more usable performance per rack within hard power and cooling limits. If a hyperscaler can deploy silicon that delivers 15–20% more tokens per watt, the lifetime energy and infrastructure savings run into billions. TSM is shifting from a classic cost-plus wafer pricing model to value-capture pricing, where wafer ASPs reflect a share of those downstream energy savings. That shift is visible: wafer ASPs have been increasing roughly 20% year-on-year for two consecutive years, yet demand remains on allocation at N5/N3 and pre-allocated at N2, and gross margins are expanding rather than compressing. Management is guiding gross margin in the 63–65% range for early 2026, versus 62.3% already delivered in Q4 2025, with longer-term targets “56% and higher” even after the drag from overseas fabs. That kind of margin structure on tens of billions of dollars in quarterly revenue only works if pricing power is real and durable. Right now, the hyperscalers’ constraint is power and cooling capacity more than chip budgets; that environment favors the supplier who can engineer and then monetize energy efficiency.

NYSE:TSM – Valuation: Premium Multiples, PEG Advantage And Peer Context

At the current ~$336 price with a P/E around 32 and a forward multiple in the low-20s, NYSE:TSM is not cheap on headline metrics. Investors are paying roughly $14.50 for each dollar of sales, a multiple that is about four times what legacy peers like Intel and Qualcomm commanded in earlier periods. However, the internal and external research you provided highlights a critical point: when you adjust for expected growth, TSM’s forward PEG ratio sits around 0.8x against a sector median near 1.66x, even though TSM’s long-term EPS growth rate is around 30%, with a margin structure that is materially superior to fabless peers. That combination – high-20s to 30% top-line growth, almost 50% net margins, and 30% EPS growth – can justify a premium P/E and P/S as long as growth actually materializes and margins hold. There are scenarios in which TSM screens as outright undervalued. One of the analyses you cited argues that if the market eventually prices TSM closer to peers on a PEG basis, a **50%+ valuation gap closes, implying a price that could notionally support the $475 Fibonacci extension in technical work, or even a more aggressive $685 target if the market fully reprices the stock as the “sovereign utility of computation” for AI. Those upside cases are not base-case guarantees, but they frame why the market is willing to tolerate a high P/S ratio: the earnings and cash flow justify it if the AI build-out continues as guided.

NYSE:TSM – Core Risks: Node Economics Inversion, Capex Hangover, Competition And Geopolitics

The main fundamental risk is not headlines about an “AI bubble” but the possibility that node economics invert. If the marginal cost and complexity of N2 (especially if forced to lean too long on multi-patterning EUV rather than high-NA EUV) drives cycle times up and yields down, you could see a situation where gross margins fall below the long-term 56% target while capex remains above USD 50 billion per year. That would create “profitless prosperity”: revenue rising on higher ASPs, but free cash flow squeezed by heavy depreciation and inefficient throughput. In that scenario, return on capital would decline structurally and the market could re-rate NYSE:TSM from a growth monopoly (mid-20s to low-30s P/E) back toward a capital-intensive utility (low- to mid-teens P/E). That alone could halve the share price even if revenue keeps growing. A second risk is over-capex or timing mismatch – if hyperscaler AI spending slows or node ramps slip while TSM is mid-build on fabs in Taiwan, Arizona, Japan and Germany, the system could briefly have too much advanced capacity relative to demand, pressuring pricing and utilization. A third set of risks involves competition and supply-chain diversification. Intel’s 18A plus EMIB and Samsung’s 2nm push will take some share, particularly from customers and governments that want multi-sourcing and domestic capacity. TSM is responding with N3 and advanced packaging fabs in Arizona, and still enjoys a yield and ecosystem lead, but the gap is narrowing at the margin. Finally, the geopolitical overhang around Taiwan remains a structural tail-risk: low-probability on any one-year horizon but high-impact. This risk is one reason management is willing to accept short-term gross margin dilution from overseas fabs as “geopolitical insurance”.

NYSE:TSM – Technical Structure, Liquidity And Market Behaviour Around $336

From a market-structure angle, NYSE:TSM has already cleared its prior resistance area around $300, which now acts as a key support zone. The stock recently tested the 127.2% Fibonacci extension around $342 on a weekly chart and is consolidating just below that area, with higher extensions flagged around $398 (161.8%) and $475 (211.8%) in the technical work you provided. Short- and medium-term moving averages are positively aligned, with the 6-week EMA above the 13-week EMA, and both sloping higher. MACD remains strongly positive, with the MACD line well above the signal line, and there is no clear sign yet of a bearish divergence that would usually precede a major top. RSI around the mid-70s and stochastic oscillators near 90 confirm an overbought but “locked” momentum condition – typical for a hyper-growth, high-quality stock in the middle of a strong trend. Volume studies and Price-Volume Trend confirm that the breakout has been driven by fresh capital inflows rather than just short covering. The near-term risk is that any disappointment on margins, N2 timing or AI demand could trigger a sharp pullback from overbought levels toward the $300 area and, if that fails, into a broader consolidation zone. But as long as price holds above the prior breakout band and the 13-week EMA, the technical structure remains consistent with an ongoing uptrend rather than an exhaustion top.

NYSE:TSM – Investment View: Aggressive AI Compounder, Rating Tilted To Buy With Volatility Caveats

Pulling everything together, NYSE:TSM is now a leverage point on the AI economy rather than a simple cyclical foundry. At roughly $336 per share, the stock embeds a lot of good news: ~20–25%+ revenue growth, mid-60s gross margins in 2026, ~30% EPS growth, and a multi-year AI infrastructure build-out that carries through at least 2028–2029. The earnings, margin profile, balance sheet and cash flows you provided justify a premium multiple, and the PEG and profitability data actually argue the stock is not expensive relative to its growth and moat. The core risk is that N2 economics, capex intensity or AI demand fail to live up to what is now being priced in; if that happens, a re-rating toward lower multiples can be brutal. Given the current data – record margins, rising ASPs, capacity sold forward at advanced nodes, and management explicitly raising capex against real demand signals – the balance of probabilities still favours the bull case. On that basis, and strictly on the numbers and structure you provided, I would frame NYSE:TSM as a high-conviction, high-quality AI infrastructure Buy on medium- to long-term horizons, with the clear understanding that any disappointment on margins or AI capex could easily generate 30–40% drawdowns inside the trend.

That's TradingNEWS