Newswrap: TSMC Earnings, Nvidia I'm 'No Loser' CEO on Dwarkesh Podcast

Tae Kim · Key Context by Tae Kim · April 16, 2026 at 14:41 · ⏱ 11 min read  | Read on Substack ↗
Summary
The newsletter argues that AI demand is accelerating at an unprecedented pace, supported by TSMC's stellar earnings and raised capex guidance, and Jensen Huang's confident outlook on agentic AI and Nvidia's ecosystem dominance. Markets should expect continued leadership from TSMC and Nvidia, while custom ASIC challengers face structural headwinds and memory pricing pressures weigh on lower-end electronics.
  • TSMC's Q1 revenue rose 41% y/y, with Q2 revenue guided above consensus at $39-40.2B vs. $38.1B estimate.
  • TSMC raised its 2026 revenue growth view to 'above 30% y/y' from prior 'close to 30%', and guided 2026 capex to the high end of $52-56B.
  • Advanced nodes (7nm and below) accounted for 74% of TSMC wafer revenue; HPC (AI) drove 61% of Q1 revenue, growing 20% qoq.
  • TSMC is expanding N3 capacity for AI demand, a historic move since they normally do not add capacity after a node reaches target.
  • Jensen Huang stated agents will 'grow exponentially', driving tool usage (e.g., Synopsys software) and causing software companies to 'skyrocket'.
  • Huang argued custom ASICs like TPU and Trainium owe their growth solely to Anthropic, not a broad trend, and that building something better than Nvidia 'is not easy'.
Read time 11 min
Length 11,883 chars
Category finance
Trade Ideas
Tae Kim Senior writer, Barron's; author of The Nvidia Way
Huang claimed TPU (Google's custom chip, designed by Broadcom) and Trainium growth are '100% Anthropic' and not a trend, and that many ASICs get canceled. This casts doubt on Broadcom's custom AI chip
Huang claimed TPU (Google's custom chip, designed by Broadcom) and Trainium growth are '100% Anthropic' and not a trend, and that many ASICs get canceled. This casts doubt on Broadcom's custom AI chip pipeline beyond Anthropic, which could pressure growth expectations. Risk: Broadcom's networking and connectivity business provides diversification; if custom ASIC revenue disappoints, valuation may adjust.
Tae Kim Senior writer, Barron's; author of The Nvidia Way
TSMC's steep capex increase for N3 expansion and the three-year outlook of significantly higher spending directly implies rising demand for EUV lithography systems, ASML's core product. TSMC is the la
TSMC's steep capex increase for N3 expansion and the three-year outlook of significantly higher spending directly implies rising demand for EUV lithography systems, ASML's core product. TSMC is the largest buyer of EUV tools. Risk: Export controls on China, technology transition risks (High-NA EUV adoption), and potential order timing shifts.
Tae Kim Senior writer, Barron's; author of The Nvidia Way
TSMC reported strong earnings, raised revenue and capex guidance, and explicitly cited accelerating AI demand driving N3 capacity expansion. The three-year capex outlook is 'significantly higher' than
TSMC reported strong earnings, raised revenue and capex guidance, and explicitly cited accelerating AI demand driving N3 capacity expansion. The three-year capex outlook is 'significantly higher' than the past $101B, signaling sustained revenue growth. Risk: Geopolitical tensions over Taiwan, overbuilding risk if AI demand decelerates, and high customer concentration.
Tae Kim Senior writer, Barron's; author of The Nvidia Way
Jensen Huang predicted that agentic AI will cause tool usage to 'skyrocket', explicitly naming 'Synopsys Design Compiler' and other EDA tools. This suggests increased license and royalty revenue for S
Jensen Huang predicted that agentic AI will cause tool usage to 'skyrocket', explicitly naming 'Synopsys Design Compiler' and other EDA tools. This suggests increased license and royalty revenue for Synopsys as agents automate more design tasks. Risk: Cyclicality in semiconductor R&D spending; competition from Cadence and open-source EDA tools.
Tae Kim Senior writer, Barron's; author of The Nvidia Way
Jensen Huang defended Nvidia's ecosystem stickiness, dismissed ASIC competition as limited to Anthropic, and argued GPU programmability enables new AI algorithms. He also stated Nvidia can handle $100
Jensen Huang defended Nvidia's ecosystem stickiness, dismissed ASIC competition as limited to Anthropic, and argued GPU programmability enables new AI algorithms. He also stated Nvidia can handle $100B orders and has secured CoWoS supply, reinforcing long-term demand visibility. Risk: Potential export restrictions on China, energy bottlenecks limiting US compute expansion, and eventual competition from ASICs if they improve.
More from Key Context by Tae Kim

This newsletter, published April 16, 2026, features Tae Kim discussing AVGO, ASML, TSM, SNPS, NVDA. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Tae Kim  · Tickers: AVGO, ASML, TSM, SNPS, NVDA