TSM Holds $404 as Chip Sector Sheds $1T — Q2 Revenue $40.2B, Capex Lifted to $64B, Analysts Target $540.20

Taiwan Semiconductor closed at $404.25 after a session that ran from $398.80 to $420.35, a 5.4% intraday range that tells you more about positioning than about the company. The stock sits 15.61% below the 52-week high of $479.00 printed June 30 and 80.71% above the 52-week low of $223.70 set August 20, 2025. Market capitalization runs between $1.94 trillion and $2.10 trillion depending on the share count convention. The Taipei listing changed hands at NT$2,370.00.

Monday's broader tape offered no help in either direction. The Nasdaq Composite ripped 1.77% to 25,822.62 as President Trump called off strikes against Iran, sending crude 6.21% lower to $79.41. Cloud infrastructure exploded higher, with CoreWeave up 14.97% and Oracle up 5.6%. Memory got carved up in the same hour: Micron fell 3.5% to $794.59 after ChangXin Memory Technologies pushed its DRAM share to 8%, South Korea's KOSPI dropped 4% with Samsung Electronics and SK Hynix each down nearly 8%, and the DRAM-focused ETF has now sunk 32% amid margin calls.

TSM sits in neither camp cleanly, which is the entire problem with owning it right now. It is the sole manufacturer of the compute that CoreWeave and Oracle are buying, and it is a Taiwan-listed cyclical trading in the same index bucket that just got liquidated across Seoul and Taipei.

The sector damage over the past two weeks is quantifiable. Chip stocks shed more than $1 trillion in market capitalization inside a single week in late July. Nvidia led the rout with $238 billion erased since the prior Friday's close. SK Hynix lost $176 billion, Samsung $173 billion, Micron $113 billion, AMD roughly $110 billion, and TSMC $119 billion. The Philadelphia Semiconductor Index still shows a 92% gain over twelve months despite a near-20% drawdown across the recent stretch.

That combination — a 92% twelve-month advance sitting alongside a 20% correction — describes a group that went vertical and is now being repriced rather than one where demand broke. The decline has been driven largely by sentiment rather than fundamentals, which is the polite way of saying confidence cracked. Growth stocks derive most of their value from cash flows expected far into the future, and that requires faith the market has temporarily withdrawn.

TSM's beta reads 1.39. In a tape this violent, that number does the work.

A Record Quarter That Got Sold Anyway

TSMC reported second-quarter revenue of $40.20 billion on July 16, landing at the absolute top of its own $39.0 billion to $40.2 billion guidance range and up 36% year over year. Consolidated revenue in local currency ran NT$1.27 trillion. Gross margin printed 67.7%, a record, up 150 basis points sequentially from 66.2% and above the 65.5% to 67.5% guided band. Net income reached NT$706.56 billion, roughly $22 billion, a 77.4% year-over-year increase. Earnings per share came in at NT$27.25.

The company then raised its full-year 2026 revenue growth outlook to slightly above 40% in US dollar terms, up from above 30% on the April call. That is a nine-point upward revision to a $130-billion-plus revenue base delivered mid-year, which almost never happens at this scale.

The stock fell more than 5%. It then dropped another 3% on Friday July 17, and Asian chip names followed it down.

The mechanism is straightforward. TSMC lifted its 2026 capital expenditure budget to $60 billion to $64 billion from the $52 billion to $56 billion range set in January, a jump of nearly $10 billion and at least $4 billion above any prior forecast. It simultaneously announced an additional $100 billion investment in Arizona for 2-nanometer and below fabs plus advanced packaging, lifting total US commitment to $265 billion.

Higher capex from a foundry normally reads as a demand signal and lifts the entire equipment complex. It did the opposite. Even large TSMC suppliers including Japan's Lasertec and Taiwan's Globalwafers dropped in the wake of the results. That reaction says the market stopped pricing the AI trade on growth and started pricing it on perfection, with any incremental spending read as free cash flow compression rather than capacity conviction.

Chairman and CEO C.C. Wei framed it in the opposite terms, describing the AI megatrend as stronger and stronger and putting the company's multi-year conviction as very high. He cited agentic AI reviving CPU demand alongside accelerator demand, and stated that capex across the next three years will significantly exceed the past three.

TSMC has now beaten estimates for eight consecutive quarters. The eighth beat cost shareholders 8% inside two sessions.

Decoding the Q3 Guide: $44.6 Billion and a Margin Step-Down

Third-quarter guidance calls for revenue of $44.6 billion to $45.8 billion, a 12% sequential jump at the midpoint and a 37% gain from a year earlier. Gross margin is guided to 65% to 67%, a midpoint of 66% against the 67.7% just delivered. Operating margin is guided to 56% to 58%.

The 170-basis-point margin step-down at the midpoint is the number that matters, and management explained exactly where it comes from. The 2-nanometer node ramp is expected to dilute gross margin by three to four percentage points during the second half of 2026. That drag is temporary in the sense that every node ramp dilutes and every node ramp eventually accretes, but it is arriving in the same window as a second, structurally permanent drag.

Chief Financial Officer Wendell Huang stated that profit dilution from overseas fabs has begun to surface. As those facilities gradually reach mass production over the coming years, gross margin gets eroded by two to three percentage points initially, widening to three to four points in later stages. Stack the N2 ramp on the overseas dilution and the theoretical combined drag runs six to eight points against a 67.7% base.

That math is why the stock trades at $404 rather than $479. A foundry running 67.7% gross margins on $40 billion of quarterly revenue is one of the most profitable industrial businesses ever assembled. A foundry running 60% gross margins is still exceptional but supports a materially different multiple.

The offsetting levers are real. Utilization stays high while the AI order book fills. Cost improvement programs have been running continuously. Pricing on advanced nodes has already moved, with reports that TSMC plans to raise prices on both advanced and mature process chips by as much as 10% starting in 2027. Better yields on N2 compress the ramp drag on their own timeline.

Applied to the guidance, a midpoint of $45.2 billion at 66% gross margin and 57% operating margin produces roughly $25.8 billion of operating income for a single quarter. That is a run rate above $100 billion annualized from a company carrying a debt-to-equity ratio of 15.27%.

Arizona at $265 Billion Is the Largest Foreign Investment in US History

The Arizona commitment has been built in three tranches. The original $65 billion was expanded under the Biden administration alongside a $6.6 billion CHIPS Act grant. A second $100 billion was announced with Trump in March 2025, covering three fabs, two packaging plants, and an R&D center. The third $100 billion was confirmed on the July 16 earnings call. Total: $265 billion, the largest foreign direct investment in American history, with $200 billion of it added since January 2025.

Eight sites have been announced in Arizona including advanced packaging plants, with reported plans for 12 fabs and four packaging facilities in total. Commerce Secretary Howard Lutnick has said Taiwanese chip firms that decline to build in the United States risk a 100% tariff, and has indicated exemptions are available to companies that file US construction plans overseen by an auditor. A White House spokesperson attributed the trillions in semiconductor commitments directly to the administration's trade policy.

The cost differential is the shareholder problem. US-made chips run 20% to 50% more expensive than equivalent Taiwan production depending on subsidy timing, tax credit recognition, and other variables. Industry estimates on the manufacturing cost gap alone cluster at 30% to 50%. A leading-edge fab requires far more than a building and lithography tools: experienced engineers, specialist suppliers, reliable utilities, and thousands of tightly sequenced production steps that took Hsinchu three decades to perfect.

There are genuine strategic offsets. Building in Arizona reduces geographic concentration risk for a company whose entire leading-edge capacity currently sits 100 miles from mainland China, and it places production adjacent to Nvidia, AMD, Apple, and Broadcom. For a supply chain that Washington has designated a national security priority, that de-risking carries a valuation premium the market has not yet paid for.

TSMC is also expanding at home. Two advanced chip packaging plants are planned for the Chiayi Science Park in southern Taiwan, with the first facility already in mass production and the second expected to begin shortly. Advanced packaging capacity, specifically CoWoS, remains the physical bottleneck on Nvidia accelerator output.

Capex allocation runs 70% to 80% toward advanced process technologies, 10% to specialty nodes, and 10% to 20% toward advanced packaging, testing, and mask operations.

HPC Is Now 66% of the Business

High-performance computing generated 66% of second-quarter revenue, up 20% from the prior quarter. That single figure has reorganized the entire investment case. TSMC was a smartphone-cycle company for fifteen years. It is now an AI infrastructure company with a smartphone attachment.

The revenue math confirms it. AI chip revenue is tracking above $40 billion for 2026, close to 25% of total revenue on its own. Advanced nodes below 7 nanometers now generate 77% of wafer revenue, which is where every accelerator, every hyperscaler custom silicon program, and every flagship mobile SoC gets fabricated.

The monthly cadence shows no deceleration. June revenue reached NT$442.68 billion, a 67.9% year-over-year increase and 6.2% higher than May, which broke a four-year seasonal pattern for the month. First-half 2026 revenue totaled NT$2,404.48 billion, or $74.99 billion, up 35.6% from the same period in 2025. May revenue ran NT$416.98 billion, up 30.1% year over year, so the acceleration through June was genuine rather than a base effect.

Full-year 2025 provides the comparison base: revenue of NT$3,809.05 billion, up 31.61%, with net income of NT$1,717.88 billion, up 46.55%, and diluted EPS of NT$66.25. Growing revenue slightly above 40% on top of a year that grew 31.61% compounds to a two-year expansion above 84%.

Q1 2026 set the pace. Revenue of $35.90 billion rose 40.6% year over year and 6.4% sequentially, with net income up 58.3%, gross margin at 66.2% against 58.9% a year earlier, operating margin at 58.1%, and net profit margin at 50.5%. Diluted EPS came in at NT$22.08, or US$3.49 per ADR unit.

Trailing twelve-month figures now read $138.387 billion of revenue, 64.46% gross margin, and 50.69% net margin. A company converting half of every dollar of revenue into net income at $138 billion of scale has no meaningful comparable in the industrial economy.

The dependency runs the other direction too. Apple's chief executive attributed the company's supply constraint directly to strong demand rather than to any production failure.

The Moat Is 73% Share and It Just Got Poked

TSMC commanded 73% of the global pure-foundry market in the first quarter of 2026, and roughly 70% of dedicated foundry across 2025. No competitor holds double digits at the leading edge. That structural position is what supports gross margins competitors cannot approach, driven by economies of scale and premium pricing justified by process leadership.

The vulnerability is the same one that has always existed: each generation of process technology matures and commoditizes quickly, which forces continuous reinvestment and periodic pricing pressure. The A14 node is targeted for volume production in 2028, and the capex commitment for the next three years significantly exceeding the prior three is what buys that roadmap.

The competitive news that moved the tape came on July 31, when reports surfaced that TSMC is developing an advanced AI chip packaging technology internally described as EMIB-like, similar to Intel's Embedded Multi-die Interconnect Bridge approach. Intel shares jumped more than 14% on Thursday on the report, which reads as an odd market reaction: the incumbent copying your technique is confirmation the technique matters, not confirmation you will keep the business.

Intel's own turnaround under Lip-Bu Tan has gained momentum through another strong quarter, which reintroduces a credible second source at advanced nodes for the first time in years. Samsung Foundry remains the third option. Neither has demonstrated yield parity at N2 or below.

China is the other structural overhang. Beijing is weighing tighter export controls on AI models and chips, which would run in the opposite direction from Washington's restrictions and squeeze the addressable market from both ends. The memory side already shows what domestic Chinese capacity does to Western share, with ChangXin taking DRAM share to 8% and driving the Micron and Korean equity destruction of the past week.

Insider behavior has been constructive. TSMC disclosed insider share purchases across late July, with executives buying into the drawdown, alongside NT$11.6 billion of fixed-income investments detailed in the June filing. Executives buying a stock down 15.61% from its high is not a thesis, but it is the opposite of the signal that preceded the July decline.

Valuation: Roughly 22 Times a Number That Is Still Rising

At a $1.94 trillion market capitalization against annualized second-quarter net income of approximately $88 billion, TSM trades near 22 times current earnings power. Apply the third-quarter guide — $45.2 billion of revenue at the midpoint with a 57% operating margin — and the forward multiple compresses further as the denominator grows through the back half.

The dividend is a rounding error at a 0.87% yield, with $1.1136 per ADR and an ex-dividend date of September 16. This is a reinvestment story, not an income story, and the $60 billion to $64 billion capex figure makes that explicit.

Sell-side positioning is one-directional. Eighteen analysts rate the stock a buy with zero sells, producing a Strong Buy consensus and an average twelve-month target of $540.20, implying 33.63% upside from $404.25. The high estimate sits at $700, the low at $430, which still clears spot by 6.37%. A separate poll of 19 analysts puts the consensus at $537.43 for 32.95% upside. A five-analyst subset averages $544.00 with a $650 high and $440 low. Other aggregations land at $496.25 and $502.60.

Every one of those targets sits above the price. When a name carrying zero sell ratings falls 15.61% from its high, the risk is not that the analysts are wrong on the business. The risk is that positioning was already maximally long and there is no marginal buyer left among institutions who have already sized the position.

The bear case has a specific mechanism rather than a vague valuation complaint. Peak margins and peak utilization make profitability and free cash flow highly sensitive to even modest demand fluctuations. Capex near $64 billion against a forward multiple that compresses toward 24 times drags fair value toward the $425 area on one model. Taiwan geopolitical risk sits underneath all of it as a discount factor no spreadsheet prices consistently.

Semiconductor industry earnings grew 131% in the second quarter of 2026 across the sector, and global chip sales are projected to reach a record $975 billion this year. The top three chip stocks account for 80% of the combined $9.5 trillion market capitalization of the top ten, which is exactly the concentration that produces $1 trillion drawdowns in a week.

Technical Structure: $397.70 Is the Floor Being Defended

The chart broke down after the July 16 print and has been repairing since. The $410 to $419 support zone was breached on the earnings reaction and flipped to resistance. Price is now working inside a range with $419 as the first ceiling and the 200-period exponential moving average above it as the second.

Downside is defined. The immediate support sits at $397.70, the low buyers defended during the post-earnings week, and Monday's session low of $398.80 landed within a dollar of it. Losing $397.70 on a closing basis opens $386.20 as the next reference. From $404.25, that is 4.5% of downside to the second shelf.

Recovery requires two confirmations in sequence. A four-hour close above $419 flips former support back to neutral and cancels the breakdown. A subsequent four-hour close above the 200 EMA restores a constructive stance and puts the $440 to $450 area back in play, which is where the low end of the analyst target range sits.

The longer structure remains intact despite the damage. TSM has been trading inside a 14-month ascending channel and is currently pulling back after establishing a new higher high at $479.00. The 50-day moving average has provided support during comparable bearish legs across that channel, though it has broken temporarily on each occasion before reasserting.

Monday's $398.80 to $420.35 range is the informative data point. A 5.4% intraday swing in a $2 trillion company, on a day with no company-specific news, describes a market with no consensus on fair value and thin depth on both sides of the book. The daily technical signal reads neutral across moving averages and oscillators, which is what a genuine standoff looks like.

The gap between the 52-week high of $479.00 and the 52-week low of $223.70 spans 114%. Spot at $404.25 sits at the 71st percentile of that range, so the stock has given back a fraction of a move that more than doubled. Year to date the shares are up 43.06%, and the twelve-month gain runs 90.79% even after the drawdown.

The next scheduled catalyst is the July monthly revenue release, followed by third-quarter earnings on October 15.