← Back to stories Detailed view of computer hard disk drive components, showcasing circuitry and connectors.
Photo by Marta Branco on Pexels
TechNode 2026-04-09

Apple’s self-designed AI server chip “Baltra” reportedly to be built by TSMC (台积电) on second‑generation 3nm process

What’s new

It has been reported that Apple is working with Broadcom on an internal AI server chip codenamed “Baltra,” and that the chip will likely be manufactured by TSMC (台积电) using its N3E second‑generation 3nm process. Reports also say Samsung Electro‑Mechanics may supply semiconductor components for the design. Apple has not confirmed the project publicly; the details remain unverified and have been characterized in reporting as preliminary.

Why it matters

If true, Baltra would mark a major push by Apple into custom data‑center silicon beyond its consumer‑device chips. Apple’s move toward vertically integrated AI hardware mirrors a broader industry trend: companies want control of both models and the silicon that runs them to squeeze out performance and efficiency gains. TSMC’s N3E node would give Apple higher transistor density and lower power consumption versus older nodes — an attractive proposition for inference and mixed‑workload server designs.

Geopolitical context

This story sits at the intersection of supply‑chain strategy and geopolitics. TSMC, based in Taiwan, is the world’s leading contract foundry and a focal point of U.S. and China semiconductor competition. U.S. export controls and broader trade policy around advanced chips complicate where and how the most cutting‑edge silicon can be produced and shipped. Will Apple’s supply decisions influence where Baltra‑powered servers can be sold or deployed? That remains an open question.

What’s next

Expect more reporting and, eventually, official confirmation or denial from the companies involved. If Apple and Broadcom proceed, Baltra could reshape competitive dynamics in AI infrastructure — and highlight how much of the next wave of compute will hinge on partnerships between design houses and a handful of advanced foundries. Who benefits: hyperscalers, chipmakers or vertically integrated device firms? The answer will matter to cloud customers and geopoliticians alike.

AISemiconductors
View original source →