Monday, August 31, 2026

TSMC Boosts 2026 Capex to $10B+ as Global AI Chip Race Spans Taiwan, US, Germany

Taiwan Semiconductor Manufacturing Company raised its 2026 capital expenditure guidance on January 1, joining Intel's Malaysia expansion and Infineon's Dresden facility in a coordinated global buildout. The semiconductor industry is betting on sustained AI hardware demand through 2027, with TSMC targeting 3nm and 2nm chips for Nvidia GPUs while Marvell pursues optical interconnect technology to solve data bottlenecks in training clusters.

ViaNews Editorial Team

February 23, 2026

TSMC Boosts 2026 Capex to $10B+ as Global AI Chip Race Spans Taiwan, US, Germany
Image generated by AI for illustrative purposes. Not actual footage or photography from the reported events.

Taiwan Semiconductor Manufacturing Company raised its 2026 capital expenditure guidance on January 1, part of a global semiconductor buildout spanning Asia, North America, and Europe. Intel committed funds to a Malaysia packaging facility in December 2024, while Infineon Technologies began construction on a Dresden power semiconductor plant targeting late 2026 production.

The coordinated expansions follow a strategic shift in chip design priorities. Marvell Technology entered acquisition talks with Celestial AI in December 2024 for optical interconnect technology—photonics-based chips that move data between GPUs in training clusters. Current copper-based links max out at 800 Gbps, while next-generation language models require 1.6 Tbps interconnects. Celestial AI's silicon photonics promise 10 Tbps bandwidth.

TSMC's increased guidance targets advanced node capacity for 3nm and 2nm chips used in Nvidia's Blackwell and future Rubin GPU architectures. Intel's Malaysia facility addresses packaging substrate shortages that have constrained AI accelerator production since mid-2024. Infineon's Dresden site will manufacture power semiconductors for data center cooling systems serving AI workloads across European and Middle Eastern markets.

Microchip Technology reported "broad-based recovery" in revised guidance, citing AI infrastructure alongside automotive and industrial demand. The recovery signals spread beyond hyperscale data centers in the US and China to regional facilities in Southeast Asia, India, and the EU.

Investment timelines suggest structural confidence in AI demand. Infineon's Dresden project extends to 2028, well beyond typical semiconductor cycle peaks. TSMC, Intel, and Infineon spending patterns create a testable hypothesis: if AI chip revenue growth correlates with optical computing acquisitions through 2026, manufacturers view demand as permanent rather than cyclical. Early signals point to long-term bets—optical interconnects solve physics problems that pure compute scaling cannot address.


Sources:
1 Yahoo Finance, "Is It Time To Reassess Infineon Technologies (XTRA:IFX) After The Recent Share Price Pullback" (March 22, 2026)
2 Yahoo Finance, "OTCQX Best 50 Virtual Investor Conference: Presentations Now Available for Online Viewing" (March 20, 2026)
3 Yahoo Finance, "China’s secret weapon for smuggling AI chips might be hairdryers" (March 20, 2026)
4 Yahoo Finance, "How Investors May Respond To Microchip Technology (MCHP) Expanding Into High-Reliability SiC Power M" (March 20, 2026)
5 Yahoo Finance, "How The AI Data Center Cycle Is Reframing The Story For Nvidia (NVDA)" (March 22, 2026)

In this story · Knowledge Files

What we know · the intelligence behind this page
Live from the substrate
What we're seeing
Enterprise AI's Trust Gap: Microsoft-Mistral Ecosystem Expansion Meets a Governance Deficit in Agentic Adoption
Microsoft is deepening its AI platform bet through simultaneous moves — expanding its Mistral partnership (Copilot Studio, Foundry, European infrastructure capacity) and deepening enterprise AI governance ties with Manulife — just as independent research (Google Cloud, VentureBeat, Box) shows enterprises racing toward agentic AI adoption (100% planned within two years) while data access and trust in agent decisions lag badly (average 45% data access, only ~half trust agent outputs). The result is a structural mismatch between platform-vendor momentum and enterprise readiness to actually govern and trust the agents being deployed.
Our read on the data ›
Signals we're tracking
EPKINLY Regulatory-Clinical Success Cascade
High probability of expanded label indications, additional combination approvals, and competitive positioning strength in follicular lymphoma market. Predicts positive commercial uptake and potential accelerated review for related indications.
Patterns we're watching ›
Where sources disagree
Morgan Stanley & Co. LLC
Same entity (Morgan Stanley & Co. LLC), same metric (net_income), same fiscal period (Q1 2026), same observation date (2026-03-31), but vastly different values: $5.567 billion vs. $5.57. These cannot coexist for the same time period. Fact B appears to be a data entry error (possible missing decimal placement: 5.57 should likely be 5,567,000,000 or a per-share figure incorrectly entered as total).
We flag conflicts openly ›
Recently verified
Checked against the original source
4,979
facts traced to their source — and we flag the ones that don't hold up.
101 entities tracked4,979 facts checked against source5,257 source documents archived
Query this data → isubstrate.com