Semiconductors

TSMC

S

Public

Taiwan Semiconductor Manufacturing Company, founded in Hsinchu in 1987 by Morris Chang, invented the dedicated-foundry business model and now dominates it utterly: 69.9% of the global foundry market in 2025, rising to 72.3% in Q1 2026, with Samsung a distant second at 6.5%. It is effectively the sole manufacturer of leading-edge AI silicon — Nvidia, Apple, AMD, and Broadcom all fab their flagship chips there — which made it the primary beneficiary of the AI buildout: 2025 revenue hit a record $122.4B (+35.9%), and Q1 2026 revenue grew another 35.1% to $35.9B at a 66.2% gross margin, with high-performance computing now 61% of sales. Its N2 (2nm) node, the company's first gate-all-around process, entered volume production on schedule in Q4 2025 at Fab 22 in Kaohsiung, with the enhanced N2P variant due in H2 2026 and the A16 node (backside power delivery) now slated for 2027 after slipping from late 2026. TSMC raised its 2026 capex guidance to a record $60–64B in July 2026 — the largest capital budget in semiconductor history — while globalizing its footprint: Arizona's Fab 21 has produced 4nm chips at Taiwan-par yields since Q4 2024, and on its July 16 earnings call TSMC pledged roughly $100B more to expand the Arizona campus with at least four additional 2nm-and-below fabs, lifting its total US commitment to about $265B (the largest foreign direct investment in US history); Japan's JASM is building a second Kumamoto fab upgraded to 3nm for late 2027, and the Dresden ESMC fab targets production in late 2027. A January 2026 US–Taiwan trade deal exempts TSMC from new 25% Section 232 semiconductor tariffs in proportion to its US buildout. The overhang remains geopolitical: the vast majority of its leading-edge capacity sits on Taiwan, at the center of US–China tension.

ASML

S

Public

ASML, founded in 1984 as a Philips/ASM International joint venture that built its first machine in a shed in Eindhoven, is the world's dominant supplier of photolithography systems (~83% of the market) and the only company on Earth that makes EUV machines — the $180M–$400M tools without which no leading-edge chip can be manufactured. That monopoly has made the Veldhoven company Europe's most valuable tech firm (~$693B, July 2026). 2025 revenue reached €32.7B with a record €13.2B in Q4 bookings and a €38.8B year-end backlog, and in April 2026 ASML raised its 2026 guidance to €36–40B as CEO Christophe Fouquet declared chip demand to be outpacing supply, with 60+ EUV shipments planned for the year. The next act is High-NA EUV: the first EXE:5000 shipped to Intel in December 2023, the first production-grade EXE:5200B arrived at Intel in Q4 2025 for its 14A node (risk production 2027), and Samsung, SK hynix, TSMC, and imec have early systems — fewer than a dozen exist worldwide, at roughly $380M each. TSMC has committed to High-NA from its A14 node (~2028). ASML also made an uncharacteristic strategic bet in September 2025, investing €1.3B to become the largest shareholder (~11%) of French AI lab Mistral AI. The main headwind is geopolitics: China was 33% of 2025 sales but is expected to fall to ~20% in 2026 under expanding US and Dutch export controls, with new congressional proposals targeting even DUV servicing.

Nvidia

S

Public

Nvidia, founded in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem, invented the modern GPU (GeForce 256, 1999) and then — via the 2006 CUDA platform — turned it into the engine of the AI revolution. It is now the world's most valuable company at $5.11 trillion (July 2026), having crossed $4T in July 2025 and $5T that October. Fiscal 2026 revenue hit $215.9B (+65%), and the momentum accelerated: Q1 FY2027 (ended April 2026) delivered a record $81.6B (+85% YoY) with data center revenue of $75.2B, and Q2 is guided to $91B. The Blackwell Ultra GB300 platform drove the ramp, and at GTC 2026 Huang declared the seven-chip Vera Rubin platform — Vera CPU, Rubin GPU, NVLink 6, and a newly integrated Groq 3 LPU — in full production, with partner systems arriving in H2 2026 and roughly $1 trillion in cumulative Blackwell-plus-Rubin orders projected through 2027. Nvidia has also become the AI buildout's central dealmaker: up to $100B invested in OpenAI to deploy at least 10 GW of Nvidia systems, up to $10B in Anthropic, a key role in the $500B Stargate project, and millions of GPUs committed to Meta. The China business is the big casualty — export restrictions triggered a $4.5B charge in 2025, and despite a December 2025 deal allowing H200 sales with a 25% cut to the US government, shipments remain stalled and Nvidia's 10-K calls the company effectively foreclosed from China's data-center market. Competition now comes less from AMD than from customers' own silicon: Google TPUs, AWS Trainium, and Broadcom-built custom ASICs.