TSMC sees no signs of the AI boom slowing for at least two or three years

TSMC sees no signs of the AI boom slowing for at least two or three years

Summary

Taiwan Semiconductor Manufacturing Company (TSMC) reported blockbuster Q4 2025 results and a bullish outlook driven by sustained AI demand. Q4 revenue was $33.7bn (up 25.5% year‑on‑year) and full‑year revenue reached $122.5bn (up 36%). Annual net income topped $55bn. Management forecasts Q1 2026 revenue of $34–$35.8bn and expects ~30% revenue growth for FY 2026.

CEO C.C. Wei said rising AI model adoption across consumer, enterprise and sovereign AI is creating multiyear demand for leading‑edge silicon, and customers typically engage with TSMC two to three years before production. CFO Jen‑Chau Huang said TSMC spent $101bn on capex over the past three years and plans significantly higher spending ahead — roughly $52–$56bn in 2026, with 70–80% devoted to advanced nodes (7nm and smaller).

Management signalled inevitable price rises as new nodes require heavier investment; wafer price increases of around 20% were noted and further rises are expected. TSMC will ramp its 2nm process fast in 2026, with two fabs already producing 2nm parts with good yield and a new “A16” process moving to volume in H2 2026. Around 30% of 2nm capacity will be produced in the USA. The company aims to lift gross margins from ~62.3% toward 65%+ while managing factory ramps and utilisation.

Key Points

  • Q4 2025 revenue: $33.7bn (+25.5% YoY); FY 2025 revenue: $122.5bn (+36% YoY); net income > $55bn.
  • Management expects multiyear AI demand; customers plan capacity 2–3 years ahead.
  • Capex intensity rising — $101bn spent over three years; $52–$56bn expected in 2026 alone.
  • 70–80% of capex will target advanced nodes (7nm and below), pushing unit costs higher.
  • Wafer price hikes (~20%) have begun and further price increases are anticipated.
  • 2nm will be ramped rapidly in 2026; volume A16 production slated for H2 2026; ~30% of 2nm output in the US.

Context and relevance

This matters because TSMC is the backbone of high‑end silicon for GPUs, AI accelerators and premium SoCs. Its capex plans, node ramps and pricing signals directly affect hyperscalers, ASIC designers and hardware vendors. Continued heavy investment in 2nm and advanced nodes preserves performance gains for AI workloads but raises manufacturing costs — a factor that will cascade into cloud pricing, GPU availability and vendor margins over the next few years.

Why should I read this?

Short answer: because TSMC pretty much runs the engine room for modern AI. If you buy GPUs, rent cloud AI instances, build AI products or follow semiconductor supply chains, this shapes prices and capacity for years. We read the earnings so you don’t have to — quick, sharp takeaways here to save you time and headaches.

Author style

Punchy: this is industry‑level news with real consequences — the figures and capex plans backing a multiyear AI boom are significant, so dive into the numbers if your hardware budget or product roadmap depends on chip supply or cost.

Source

Source: https://go.theregister.com/feed/www.theregister.com/2026/01/16/tsmc_q4_2025/