NYSE:TSM – AI foundry leader at $319.61 and $1.36T
Financial strength, growth and profitability at NYSE:TSM
At the latest close, NYSE:TSM trades around $319.61, with after-hours levels near $320.18, right against its 52-week ceiling of $134.25–$321.59. At this price the company sits on roughly $1.36 trillion in market value, with a trailing P/E around 32.7x, a forward multiple near the low 30s, and a cash dividend yield in the 1.0–1.1% area. The market is clearly paying a premium multiple and the numbers you pasted show why.
In Q3-2025, revenue reached about NT$989.92B, up roughly 30.3% year over year, while net income climbed to NT$452.30B with growth of around 39.1%. That translates into a net margin of approximately 45.7%, which is software-level profitability in a manufacturing business. Earnings per share jumped roughly 50.5% year over year, and EBITDA hit about NT$660.69B, confirming heavy operating leverage as capacity fills with high-margin AI and HPC wafers.
The balance sheet backs up that income statement. Cash and short-term investments stand near NT$2.75T, up roughly 27% from a year earlier. Total assets are about NT$7.35T, with total liabilities around NT$2.32T and equity near NT$5.04T. That leaves leverage low, liquidity high and gives NYSE:TSM ample room to continue funding extreme capex. Return on assets in the high-teens and return on capital above 20% show that this is not just a big balance sheet; it is an efficient one.
Free cash flow is intentionally constrained by aggressive investment. In your data, free cash flow sits around NT$49.71B, down more than 50% year over year, while cash from operations is above NT$426B and cash used for investing is roughly NT$260B in the quarter. Management is clearly prioritising cutting-edge capacity over short-term free cash flow optics. As long as that spending remains concentrated in 3nm, 2nm and advanced packaging that Nvidia, Apple, AMD and the hyperscalers will actually load, the trade-off is rational. For intraday context and technicals, the reference point stays your own NYSE:TSM real-time chart.
AI and HPC demand structurally rewiring NYSE:TSM’s revenue mix
The most important shift is not just that NYSE:TSM is growing; it is how the mix has moved. AI and high-performance compute have shifted from being growth segments into the core of the company. In the Q3 data from your articles, the HPC segment, which houses AI accelerators for Nvidia (NVDA) and AMD (AMD), accounts for around 57% of consolidated sales. Management openly called AI demand “insane” on the earnings call and then raised its own internal growth assumptions, saying demand was even stronger than anticipated three months earlier.
One of the pieces you provided estimates that AI inference and training chips already represent roughly 54% of trailing twelve-month revenue. At the same time, TSMC has lifted its 2025 revenue guidance to roughly mid-30s percent growth in USD, implying annual sales in the $112–$113B range with growth around 35–37%. These are not early-stage numbers; this is a mega-cap compounding like a much smaller growth stock.
The real flywheel sits in hyperscaler capital expenditure. The material you shared points to 2026 hyperscaler capex exceeding $600B with growth of around 36%, with AI datacenter infrastructure taking an outsized share of that budget. Because most of the relevant accelerators and advanced CPUs at the top of that stack are taped out on TSM’s 5nm, 3nm and soon 2nm nodes, NYSE:TSM becomes a direct derivative of hyperscaler AI capex, not just industry-average semiconductor demand.
A bottom-up segment model in your sources uses realistic growth rates by business line and arrives at a 4-year revenue CAGR around 28.8% for NYSE:TSM, versus a consensus revenue CAGR closer to 15.8%. That gap between what the fundamentals can deliver and what the Street is pricing is exactly where the “mispricing” thesis sits. At $319.61, the stock looks expensive on static multiples, but still underestimates the combination of volume and pricing uplift as AI and HPC dominate the mix.
Advanced node mix and pricing power at NYSE:TSM
The node distribution you gave is blunt: about 74% of wafer revenue comes from technologies at 7nm and below. Inside that advanced bucket, 3nm contributes around 23% of wafer revenue and 5nm accounts for about 37%. This is an extraordinary concentration in bleeding-edge processes that only a handful of players can even attempt at volume.
Because AI accelerators, high-end CPUs, and premium smartphone SoCs all crowd into these nodes, this mix has two direct consequences. First, it lifts gross margins because these nodes command much higher average selling prices per wafer and per die. Second, it gives NYSE:TSM room to raise prices without losing business, since the alternative for most customers is either delayed roadmaps or sub-par performance.
The articles you provided mention announced and expected price hikes of roughly 3–10% in 2026 across nodes, with some specific discussion of 8–10% increases on advanced nodes. On top of that, there is the natural ASP uplift as the mix shifts from 5nm to 3nm and then to 2nm and A16. Taken together, that means revenue growth is not purely a function of volume; it also has a clear margin-accretive pricing layer.
This is why at $319.61, investors are paying for a company that is effectively an oligopoly supplier on the most profitable slices of logic manufacturing. The risk is that this pricing power attracts political, regulatory or competitive push-back, but in the near term the scarcity of alternatives keeps that risk mostly theoretical.
Foundry 2.0, CoWoS and packaging turning NYSE:TSM into a double bottleneck
The second level of moat is the move from pure wafer foundry to an integrated front-end and back-end platform. Management’s “Foundry 2.0” framing in your articles is not marketing language; it reflects a deliberate strategy to pull advanced packaging inside the walls.
In the material you gave, advanced packaging revenue – largely CoWoS and other high-end integration – is described as approaching 10% of total revenue, with growth faster than the rest of the company. CoWoS is the enabling technology that attaches logic dies to high-bandwidth memory stacks, effectively creating the single GPU or accelerator unit that powers AI training and inference servers. Capacity there is tight, demand is growing rapidly, and TSM is raising prices on packaging just as it is increasing wafer prices.
Competitors like Intel and Samsung would need to match not only TSM’s transistor technology at 5nm, 3nm and 2nm, but also its CoWoS yield, thermal performance and overall packaging economics at scale. Given that TSM’s capex plan calls for $40–42B in 2025 stepping up toward $50B in 2026 and 2027, while Samsung and Intel are spending less overall and across more diversified businesses, the probability of anyone matching this full stack in the next few years is low.
The result is that NYSE:TSM controls two separate choke points in the AI hardware supply chain: advanced logic nodes and advanced packaging. That is exactly why Nvidia and the hyperscalers pre-book capacity and accept price increases; they are paying for access to a constrained resource that they cannot easily source elsewhere.
Capex intensity, Arizona fabs and geographic diversification for NYSE:TSM
The capex numbers in the articles are aggressive by design. Management is guiding for $40–42B in capital expenditures for 2025 and mapping out a path to around $50B per year for 2026–2027. This spending is being deployed into incremental 3nm and 2nm capacity in Taiwan, into advanced packaging lines, and into the multi-fab complex in Arizona.
The second article you pasted highlights an important shift in timing. The second Arizona fab is being accelerated: equipment is expected to start moving in around summer 2026, targeting 3nm production in 2027, rather than the earlier expectation of 2028. That pulls forward US-based advanced-node revenue by roughly a year and positions TSM to benefit directly when US domestic content rules and subsidies start to bite harder.
At the same time, TSM has already committed more than $100B to the Arizona build-out, while the first fab ramps 4nm. When you overlay that with new US tariffs on Chinese chips scheduled from June 2027, the strategic logic is obvious. For products destined for the US market, especially in AI and high-end compute, Chinese foundries like SMIC become structurally disadvantaged. NYSE:TSM becomes the default advanced foundry for US-aligned customers with both Taiwan and US capacity.
The constraint is Taiwan’s own national-interest policy. The material you included references proposals for an N-2 export rule, which would only allow export of technologies two generations behind the domestic leading edge. If Taiwan is running A16 or 1.4nm as the frontier, overseas fabs might be capped at 3nm or 2nm. That slows the pace at which the absolute leading edge transfers abroad and keeps the most valuable process technology concentrated inside Taiwan, which is both a strategic strength and a geopolitical risk.
2nm, A16 and yield performance supporting NYSE:TSM’s margin plateau
The node roadmap in your texts is clear. N3 has already ramped to about 23% of wafer revenue. N2 (2nm) began volume production in Q4-2025, and supply-chain commentary in your articles indicates that yields are coming in ahead of internal expectations. At the same time, TSM is planning an A16 (1.6nm) technology featuring backside power delivery (“Super Power Rail”), scheduled around H2-2026 with meaningful financial impact in 2027.
When 2nm yields are better than expected this early, cost per good die falls faster than planned and the node becomes economically compelling for a broader set of customers sooner. Because 2nm wafers can command significantly higher prices than 3nm and 4nm, that yield advantage directly feeds through into gross margin and operating margin as volumes ramp.
Combine that with the price increases you have in the articles – roughly 3–10% across nodes and higher uplifts at the bleeding edge – and you get the foundation for what several analysts in your text call a new structural gross margin band above 60%. TSM already reported gross margin near 59–60% with guidance for 59–61%, and the mixture of 3nm scale, 2nm ramp and A16 on the horizon should keep that margin elevated for several years rather than dropping back to the low-50s band of past cycles.
A model in your sources takes the 28.8% four-year revenue CAGR assumption, pairs it with about 300 basis points of gross margin expansion, and produces a four-year net income CAGR around 30.9%. At that pace of earnings growth, a $319.61 share price and a low-30s forward P/E no longer look stretched; they look like a growth multiple on a structural compounder.
Valuation frameworks and upside potential for NYSE:TSM from $319.61
The first valuation layer is simple multiples. At $319.61, NYSE:TSM trades on a forward P/S around 10.3x, a forward P/B around 8.7x, and a forward P/E around the high-20s to low-30s, depending on the EPS source. Those numbers are multiples of the sector median metrics, which sit closer to 3.3x P/S, 4.4x P/B and 24x forward P/E in the datasets quoted in your articles. On that view, the stock is expensive and priced for perfection.
Then you layer in growth. The same data show a forward EPS long-term growth estimate around 21.8% and a forward PEG near 1.33, roughly 20% below sector median PEG. In other words, the stock is expensive versus the average semiconductor name, but not expensive relative to its own earnings trajectory.
One of the more detailed articles in your package runs a discounted earnings framework using the internal 28.8% 4-year revenue CAGR, assuming that some of the gross margin expansion flows through to net income and landing on 30.9% four-year earnings CAGR. Using a terminal 1-year forward P/E of 22.2x (equal to the peer median) and discounting back at 8.5% as an equity opportunity cost, it arrives at an implied fair-value market cap of about $2.43 trillion, versus $1.31 trillion today. That implies roughly 85% upside from current levels. The sensitivity grid in that piece still shows upside north of 50% even if earnings CAGR is cut and discount rates are raised to reflect higher geopolitical risk.
The other valuation shortcut from your material takes consensus FY2026 EPS around $12.72, applies a modest surprise factor of 0.75% and a target P/E of 29.5x near the top of the historical band. That yields a fair value per share near $378, which was 26% above the roughly $303 share price at the time of that article and still around 18–20% above today’s $319–$320 area.
The conclusion from these internal models is consistent: NYSE:TSM is not a bargain on raw P/E, P/S or P/B, but once you adjust for realistic revenue and margin trajectories, today’s price is still below a conservative fair-value range. The risk, and the reason the stock is not trading at 40x earnings, is that this entire structure is vulnerable to external shocks.
Risk profile – geopolitics, China’s response and AI-cycle sensitivity for NYSE:TSM
The first and most obvious risk is geopolitical. TSMC is physically concentrated in Taiwan, the Taiwan government is the largest direct shareholder, and the company is at the centre of global attention. A direct conflict between China and Taiwan, even short of a full invasion, would instantly reprice the stock, disrupt supply chains and potentially impair assets. The N-2 export rule discussion reinforces that semiconductors are treated as strategic assets in Taipei, not just commercial products.
China is responding economically as well. The first article you provided mentions Chinese incentives of up to $70B aimed at supporting domestic chipmakers. In the short and medium term, these subsidies are unlikely to erase the multi-year lead of NYSE:TSM, but they increase the probability that some future trailing-edge and mid-range nodes face pricing pressure and that China eventually narrows the gap on certain technologies.
The second cluster of risks is cyclical. The bullish case assumes that AI accelerator demand, hyperscaler capex and high-end compute consumption continue to grow at elevated rates through at least 2029. If the AI investment cycle were to cool faster than expected in 2026–2027, or if overcapacity emerges because too many players invest simultaneously, TSM would be left with very high depreciation and a temporarily under-utilised advanced-node and packaging footprint. Given the planned $50B per year capex levels, even modest utilisation dips can compress margins and free cash flow in a way that would trigger a derating from today’s 30x type multiples.
The third risk bucket is executional. At current valuation levels, NYSE:TSM is effectively priced for consistent beats or at least meets. Any meaningful earnings miss, yield issue on a new node, delay in a major node ramp, or cost overrun on US and other overseas fabs could prompt a sharp reset. The stock is “priced for high performance,” which means the margin for operational error is thin.
Because you specifically emphasise insider flows in your rules, it is worth stating how to frame them here. For a company of this size, insider transactions are more about signalling than fundamentals. The core drivers in this story are capex, node yields, customer demand and geopolitical risk; that said, keeping an eye on insider buying or heavy selling around big inflection points still makes sense. For that angle, the references stay your own tools at TSM stock profile and the specific TSM insider transactions page, rather than trying to read too much into any one form 6-K or 20-F disclosure.
Technical backdrop and positioning view on NYSE:TSM
On the price action side, the technical description in your second article is straightforward. Relative to the S&P 500, NYSE:TSM is in a clear uptrend and has just pushed out of a consolidation phase with a strong bullish engulfing pattern, the kind of move that technicians label a clean re-accumulation breakout. The stock has more than doubled from the $134 area to above $300 within a year, yet the relative chart still looks constructive rather than exhausted; that kind of pattern usually signals institutional accumulation rather than distribution.
With the stock trading at $319.61 near its record high, there is no valuation “floor” to protect short-term traders. Volatility is the price of admission. But for investors working on a multi-year horizon who are willing to average entries around this zone, the combination of strong fundamentals, structural moat, margin profile and AI leverage still aligns with a bullish technical backdrop rather than a topping pattern.
Buy, sell or hold – clear stance on NYSE:TSM from here
Putting all your data and the external numbers together, NYSE:TSM at around $319–$320 is a high-quality, high-risk, high-reward AI infrastructure core holding. The company controls the advanced-node and advanced-packaging bottlenecks, is compounding revenue at roughly 30%, net income around 30%, and sustaining net margins north of 40% with realistic potential for 60%+ gross margins as 2nm and A16 ramp. Capex intensity is extreme but coherent, the balance sheet is strong, and hyperscaler AI capex plus US tariff policy are powerful tailwinds.
The stock is not cheap on simple multiples and is absolutely priced for continued execution, which means any macro shock, AI-cycle slowdown or geopolitical event will hit hard. But when you run through the earnings and valuation models embedded in your articles, the fair-value range consistently sits well above the current $319.61 level, with upside scenarios between roughly +25% on the conservative one-year target and +50–85% on more aggressive multi-year CAGRs.
Given that profile, the stance is not neutral. On the information you provided and the numbers on the table, NYSE:TSM is a Buy, skewed toward Strong Buy, with a clearly bullish view for 2026 and beyond, as long as position sizing respects the real geopolitical and cycle-risk embedded in the name.
That's TradingNEWS
