Analysis
TSMC is investing an additional $100 billion in its Arizona operations, pushing its total commitment to the site to $265 billion, as the world's largest contract chipmaker reported a 77% jump in net income and record quarterly revenue driven by what it now calls sustained, multi-year AI chip demand rather than a cyclical spike.
The Arizona build-out is progressing faster than skeptics expected when TSMC first broke ground: the first fab is operational and hitting yields comparable to TSMC's flagship facilities in Taiwan, a milestone that matters because early US fab attempts by other chipmakers have historically struggled with yield parity. A second fab is moving in equipment now, and a third is under construction -- meaning TSMC is committing to US manufacturing depth, not just a single symbolic plant.
“A second fab is moving in equipment now, and a third is under construction -- meaning TSMC is committing to US manufacturing depth, not just a single symbolic plant.”
The financial numbers back up the infrastructure bet. TSMC raised its 2026 capital expenditure guidance to a range of $60-64 billion, among the most aggressive in its history, and is projecting revenue growth of slightly above 40% for the year. Its top customer, Nvidia, posted $81.6 billion in quarterly revenue, up 85% year-over-year -- a reminder that TSMC's fab economics are now directly tethered to a small number of AI chip customers whose own demand curves determine TSMC's capacity utilization.
The timing lands the same week Microsoft is expanding its AI silicon relationship with AMD, a signal that hyperscalers want supply diversity even as they keep placing enormous orders with Nvidia and, by extension, TSMC. For TSMC, more fabs and more customers reduces concentration risk on both counts -- geographic (less exposure concentrated in Taiwan) and customer (serving both Nvidia and AMD's AI silicon).
What to watch: whether TSMC's Arizona yields hold as it scales beyond the first fab, and whether the $60-64 billion capex guidance proves conservative if AI chip demand keeps outrunning even TSMC's aggressive planning assumptions.
