TSMC accelerates 1.4 nm plans, targets 2027 pilot runs

TSMC has accelerated construction on fab 25 in the Central Taiwan Science Park to produce 1.4 nm wafers, aiming for risk production by the end of 2027 and high-volume ramp in the second half of 2028.

TSMC has quietly stepped up work on a major new complex, fab 25, in the Central Taiwan Science Park near Taichung. the planned site is expected to include four production buildings and carries an initial investment estimate in the low-trillion New Taiwan dollar range. suppliers and construction partners have been told to accelerate deliveries so specialized tools and materials arrive on a compressed schedule to meet the tighter timeline.

the company is targeting the first plant to begin risk or production by the end of 2027, with high-volume production planned for the second half of 2028 to serve major customers including Apple, NVIDIA and AMD. TSMC describes the node as an Angstrom-era technology and expects it to deliver more than 15 percent performance gains and significant power reductions reportedly up to 30 percent. the report also warns that wafer costs will be substantially higher than at the 2 nm node because of greater complexity and elevated operational expenses.

infrastructure and regional planning are part of the accelerated buildout, with negotiations under way on recycled water and desalination commitments to protect local supplies. some Hsinchu facilities are being refitted to host pilot lines and research and development work that will support the 1.4 nm program, and planners in the south are already considering sites for future, even smaller-node fabs. if TSMC meets the revised schedule, fab 25 would further extend the companys foundry lead and alter the capital intensity and competitive dynamics of advanced chipmaking.

85

Impact Score

Anthropic launches Project Glasswing for cyber defense

Anthropic has introduced Project Glasswing to address mounting cybersecurity risks tied to increasingly capable Artificial Intelligence models. The initiative brings major technology and finance companies together to use Claude Mythos Preview as a defensive tool for critical software.

Intel and SambaNova pitch modular inference architecture

Intel and SambaNova are positioning a mixed-hardware inference design as an alternative to GPU-only deployments. The approach splits prefill, decode, and orchestration across different processors for demanding Artificial Intelligence agent workloads.

Global Artificial Intelligence governance pulls back

A broad pullback in Artificial Intelligence regulation is taking shape across Colorado, the European Union, Canada, the United Kingdom, and the United States. The shift reflects implementation gaps, competitive pressure, and resistance to heavy compliance burdens rather than the end of governance efforts.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.