The most consequential argument on institutional desks right now is not about any individual stock. It is about whether the AI capital spending cycle is compounding or cresting. This morning, TSMC handed one side of that debate a very large piece of evidence.
TSMC reported a 53.3 percent year-on-year rise in August revenue to NT$514.81 billion, roughly $16.3 billion, with monthly revenue also increasing to NT$514.81 billion from NT$467.58 billion in July. The year-on-year figure alone would be notable. The sequential acceleration is what matters more to the investment committee.
August revenue was up from NT$467.58 billion in July, when TSMC had reported a 44.7 percent year-on-year increase. Growth did not just hold. It widened by nearly nine percentage points in a single month. The company’s monthly revenue has now risen for four straight months. That is not a seasonal blip. That is a trend line moving in one direction.
Why This Number Carries Weight
TSMC is not a company that benefits from any single product cycle or any one customer’s quarterly budget. The group makes chips for customers across the AI supply chain, so its figures are closely watched as a sign of AI semiconductor demand. When TSMC’s revenue accelerates, it reflects orders placed months earlier by the hyperscalers, the cloud builders, and the custom silicon programs at Broadcom and Marvell. The August read is cleaner than most, because it incorporates no one-off quarterly pull-forward. It is simply what the world’s dominant foundry billed its customers last month.
TSMC, which is a key manufacturer for Nvidia’s leading AI accelerators, now expects capital expenditures of $60 billion to $64 billion in 2026, $8 billion above its prior $52 billion to $56 billion range. That revised capex commitment was made in July. The August revenue figure is the first clean monthly confirmation that the demand justifying that investment is still present.
The Bull Case, and Where It Gets Complicated
During its second-quarter earnings release in July, TSMC forecast third-quarter revenue between $44.6 billion and $45.8 billion. Today’s August figure is tracking well above the monthly run rate needed to reach that guidance. The company also raised its 2026 revenue growth projection to slightly more than 40% in U.S. dollar terms, up from its prior view, citing strong AI-related demand.
The bear case has never really been about TSMC’s own execution. It has been about whether the hyperscalers spending the money are building ahead of actual enterprise demand, and whether that gap closes on its own. Ongoing expansion in cloud capacity has driven elevated capital expenditures and financing needs, creating market scrutiny over free cash flow sustainability and execution risk in converting massive order backlogs. That concern does not disappear because TSMC had a strong August. But it becomes harder to sustain as the primary thesis for going underweight the semiconductor stack.
What Investors Are Missing
The debate is often framed as a binary: either AI capex peaks and the chip names correct, or it doesn’t and they run further. The more interesting question is where in the supply chain the constraint migrates next. TSMC dominates leading-edge logic manufacturing; Samsung, SK Hynix and Micron dominate advanced memory, particularly the HBM stacks that sit directly alongside Nvidia’s newest accelerators. A capex increase at TSMC does not pull forward a single additional wafer of HBM4 capacity at SK Hynix, and the two companies’ capacity-expansion timelines are not coordinated. TSMC solving its half of the equation does not solve the system-level bottleneck. That distinction matters for anyone building a position in Applied Materials or ASML on the assumption that wafer-fab equipment demand is uniform across the stack.
Then there is the Oracle dimension. Oracle is releasing its first-quarter fiscal 2027 results today, Thursday, September 10, 2026, after the close. Analysts are forecasting revenue of about $19.14 billion, with investors focused on the company’s $638 billion in remaining performance obligations. If Oracle’s cloud infrastructure commentary aligns with what TSMC’s revenue implies, the market gets two independent confirmations in a single session that the AI build is still compounding, not plateauing.
Stocks to Watch
TSM is the direct read. Revenue growing faster in August than July, against a backdrop of rising capex, means the foundry is absorbing that investment with demand rather than hoping for it.
NVDA remains the most leveraged expression of the same thesis. The capex increase at TSMC mirrors ASML’s own guidance raise, reinforcing that semiconductor capital equipment names are riding the same demand wave. Broadcom and AMD both sit inside the positive read-through, since custom silicon demand tied to agentic AI does not run through one vendor.
ASML and Applied Materials are the equipment layer. ASML reported Q2 2026 net sales of €9.3 billion and raised its full-year 2026 guidance to €43 to €45 billion. A sustained acceleration in TSMC’s revenue supports that equipment spend continuing into 2027.
AVGO is the name most likely to be underappreciated here. Custom silicon programs at Google and Meta run through Broadcom’s design teams but get manufactured at TSMC. Accelerating foundry revenue without a corresponding Nvidia-level headline means some of that volume is landing in custom ASIC programs, and Broadcom is the primary beneficiary of that shift inside the hyperscaler infrastructure budgets.
Rising Star Stocks