Apple’s self-designed AI server chip “Baltra” may be manufactured by TSMC

Apple's AI server chip 'Baltra', developed in collaboration with Broadcom, is expected to be produced by TSMC using advanced 3nm technology. The chip will utilize a Samsung semiconductor substrate and aims to enhance Apple's cloud infrastructure while reducing dependency on costly NVIDIA GPUs.

Key Points

• Apple, in partnership with Broadcom, is developing an AI server chip named 'Baltra'.
• TSMC will manufacture Baltra using the N3E process, a second-generation 3nm technology.
• Samsung Electro-Mechanics will provide the semiconductor glass substrate for the chip.
• Baltra is intended for Apple’s cloud services, improving security and reducing reliance on NVIDIA GPUs.
• The project aims to lower operating costs for data centers.

Relevance

• The trend towards in-house chip design in tech companies is accelerating in 2025, with more firms aiming to reduce dependency on third-party suppliers.
• Apple's shift to custom chips parallels industry-wide transitions towards AI and cloud solutions, emphasizing cost efficiency.
• The demand for efficient data center operations is aligned with growing environmental concerns and the push for sustainable technology.

Apple's Baltra chip signifies a strategic move towards efficiency in cloud computing while showcasing the broader trend of tech giants designing their own hardware to decrease costs and increase performance.