September 7, 2026 · Decision research
Crowding Positioning
Finding
The weakest assumption is that crowded ownership is itself evidence that semiconductor fundamentals are about to break. Current primary data do not support that inference: NVIDIA's latest quarter and outlook accelerated, while TSMC's latest reported sales remained exceptionally strong. The better conclusion is not "buy because demand is strong," but "do not reduce solely because the trade is crowded."
What changed
The evidence has moved from a positioning-only warning to a two-sided picture. NVIDIA reported second-quarter fiscal 2027 revenue of $96.2 billion, up 106% year over year, and data-center revenue of $89.0 billion, up 117%. It expects third-quarter revenue of $108.0 billion, plus or minus 2%, while assuming no data-center compute revenue from China. TSMC reported July 2026 revenue of NT$467,580 million, up 44.7% year over year, and cumulative 2026 revenue of NT$2,872,064 million, up 37.0%; the figures are unaudited.
Decision
Keep the current "consider" action unchanged, but make it conditional and staged. Do not add risk simply because earnings are strong, and do not exit simply because the survey is crowded. If exposure is already large, rebalance toward a position size that can survive a sharp valuation or policy shock rather than making an all-or-nothing call.
Watch next
The decision should change to reduce or defer exposure if NVIDIA cuts its forward guide, TSMC's year-over-year growth materially decelerates, hyperscalers reduce planned infrastructure spending, or semiconductor ETF flows turn persistently negative while earnings estimates also fall. A single weak stock reaction is not enough; the falsifier is a joint deterioration in demand, estimates, and positioning.