The company that manufactures the world's most advanced AI chips is still struggling to keep up with demand. TSMC reported June 2026 revenue of about NT$398 billion — up 67.9% from a year earlier — in monthly figures released on July 13, 2026. The result capped a quarter that beat Wall Street's expectations and underscored that, for now, the AI buildout is still tightening, not loosening.
The numbers
June's haul pushed second-quarter revenue to roughly NT$1.27 trillion, ahead of the analyst consensus of about NT$1.264 trillion. For the first half of 2026, TSMC has now booked NT$2.4 trillion in revenue, a 35.6% increase over the same period last year. The world's largest contract chipmaker cited AI demand it described as "extremely robust."
Sold out where it counts
The scarcity is concentrated exactly where the AI race is decided. TSMC says it remains sold out on its leading-edge N3 process node and on CoWoS, the advanced packaging technology used to stitch together high-bandwidth memory and logic in AI accelerators. That packaging line has been the single biggest bottleneck constraining shipments of Nvidia and AMD data-center GPUs, and TSMC has been racing to add capacity for more than a year.
A read on the whole trade
Because nearly every frontier AI chip is fabricated at TSMC, its revenue is the cleanest available proxy for real, shipped AI-hardware demand — as opposed to announced spending plans. A 67.9% monthly jump argues that orders from hyperscalers and chip designers are still climbing, not plateauing, even as investors grow more skeptical about when the industry's roughly $1.5 trillion in 2026 infrastructure spending will pay off.
What's next
The monthly print sets up TSMC's full second-quarter earnings on July 16-17, where management's guidance on capacity, pricing and full-year growth will carry outsized weight. With chip and memory stocks wobbling on AI-spending doubts, TSMC's outlook has become one of the most important tells for whether the hardware cycle keeps running hot into the second half.