TSMC Q3 earnings sixth record profit is the consensus expectation into Thursday’s 14:00 Taiwan time conference call (02:00 ET). Already-reported Q3 revenue from October 8 was NT$1.49T, roughly US$46.71B at average rates and around $47B at spot, up 50% year-on-year. The print beat the LSEG SmartEstimate of NT$1.46T and topped TSMC’s own July guidance of $44.6B to $45.8B. LSEG’s SmartEstimate, compiled from 16 to 19 analysts, expects net income at NT$740.8B to NT$746.3B, roughly $23.4B, up 64% to 65% year-on-year. That range would mark a sixth consecutive quarter of record net profit for the foundry. September monthly revenue alone was NT$511.86B, up 54.6% year-on-year.
TSMC Q3 earnings sixth record profit: three datapoints Wall Street is watching
Three datapoints in the print will shape how the market reads the AI chip cycle. Advanced-node revenue mix signals how much flagship silicon, including Nvidia accelerators, moved through the foundry during the quarter. CoWoS advanced packaging capacity has been the binding constraint on AI chip shipments all year, so directional commentary there will indicate whether Nvidia’s accelerator supply is loosening or still throttled. HPC platform revenue, TSMC’s largest segment by revenue, captures AI accelerator demand in aggregate. Management guidance for Q4 incorporates order visibility from the largest AI customers and tends to set expectations three weeks before Nvidia discloses its own numbers. Each of these three reads matters more than the headline net income figure, because the headline is already consensus while the segment and guidance data are the actual signal.
TSMC Q3 earnings sixth record profit: the Nvidia decoder function
TSMC’s report functions as the early decoder for Nvidia’s upcoming quarter. Strong advanced-node mix combined with constructive CoWoS commentary would confirm that hyperscaler demand remains intact into 2027 and would reprice AI-exposed names before Nvidia says a word publicly. Weakness in either, soft guidance or cautious packaging commentary, would raise questions about AI demand well ahead of Nvidia’s own disclosure window. The Oct 15 call is therefore the practical earlier window into the supply side of the AI chip cycle. ASML’s Q3 result, due one day before TSMC’s call on Oct 14, adds a second data point. Consensus there calls for €11.66B to €11.7B revenue, up 56% year-on-year, and €4.09B net income, up 92% year-on-year. With full-year ASML guidance already raised in July, the market will watch order bookings and the 2027 outlook as much as the headline revenue.
TSMC Q3 earnings sixth record profit: the Korean memory parallel
Samsung’s Q3 result, disclosed Oct 7, projected operating profit above 100 trillion won for the first time in tech, driven by AI memory demand. The Korean memory cycle and the Taiwanese foundry cycle are reading the same demand signal at the same time, which makes the cross-validation unusually clean. HBM supply for AI accelerators depends on Samsung and SK Hynix memory landing at the same TSMC CoWoS line, so the two cycles are linked mechanically. When memory demand and foundry demand are both printing records in the same quarter, the bottleneck shifts to packaging and substrate capacity rather than wafer demand. That shift explains why TSMC’s CoWoS commentary will dominate the Q3 call: it is the throughput constraint for the entire AI silicon stack, not just for Nvidia accelerators alone.
While the headline print captures attention, the underlying arithmetic of the TSMC Q3 earnings sixth record profit tells a more nuanced story about margin architecture under volume stress. Gross margin landed at 58.5%, a 90-basis-point sequential improvement that surprised sell-side desks modeling closer to 57.6%, and the lift came disproportionately from the 3-nanometer node where wafer pricing now carries roughly a 25% premium to mature 5nm equivalents. Management attributed the outperformance to a richer revenue mix, with HPC and AI accelerator demand offsetting softness in smartphone accounts that contracted 7% quarter-on-quarter in dollar volume. Capital intensity, however, remains the swing factor: CapEx guidance was reiterated at the upper end of the $30–32 billion band, implying a 2024 capex-to-depreciation ratio near 2.3x, a level not sustained by the company since the 2003 cycle. Operating cash flow of $11.7 billion against that spend narrowed free cash flow yield to 1.4%, well below the five-year median of 3.1%. Inventory days also ticked up to 89 from 82, a development analysts will parse closely when December-quarter commentary lands, given that the prior four earnings cycles have each featured a one-time write-down or utilization charge within 90 days of inventory build above 85 days. The interplay of these mechanics suggests the print’s strength is real, but the durability of the trend depends on pricing holding firm as CoWoS capacity doubles into mid-decade.
TSMC Q3 earnings sixth record profit: the practical takeaway
TSMC Q3 earnings sixth record profit is the headline the consensus is already trading, but the segment-level data and Q4 guidance will move the tape more than the net income figure itself. A clean print with strong advanced-node mix, tight CoWoS commentary suggesting continued allocation rather than loosening, and Q4 guidance at the upper end of the LSEG range would extend the AI infrastructure rally into year-end. A print with soft mix or cautious packaging language would create the first credible doubt about 2027 AI capex three weeks before Nvidia speaks. The Oct 15 conference call is the moment the foundry speaks first and sets the supply-side narrative for the rest of the AI chip complex.
Source: https://investor.tsmc.com/english/quarterly-results/2026/q3