Apple Extends Collaboration with Broadcom on Custom Semiconductors Until 2031

ASIC Collaboration Extended Until 2031
Development Underway for AI Server Chip "Valtra"

Apple has extended its collaboration with Broadcom on Application-Specific Integrated Circuits (ASICs) until 2031. This move aims to strengthen AI infrastructure to enhance the performance and reliability of Apple Intelligence-centric artificial intelligence (AI) services.

Yonhap News Agency

According to Bloomberg and other sources on July 6 (local time), the two companies will continue their partnership in the ASIC sector. ASICs are semiconductors designed for specific purposes, offering greater power efficiency and processing performance compared to general-purpose chips. As ASICs can be optimized for repetitive and large-scale AI inference tasks, major tech companies are increasingly adopting them.

There had been speculation that Apple’s collaboration with Broadcom would narrow as Apple developed its own N1 chip—a Wi-Fi and Bluetooth integrated chip embedded in the iPhone, iPad, and Mac. However, with this long-term agreement, the companies will jointly design key computational silicon assets for multiple future generations of Apple products. In particular, given Apple’s push for cloud-based Apple Intelligence, the two companies are expected to continue their collaboration in semiconductors for AI infrastructure.

Apple is also co-developing its first AI server chip, ‘Valtra,’ with Broadcom. The chip is an expanded version of the M5 Ultra chip, which Apple plans to debut this year in the Mac Studio. Following mass production in the second half of this year, Valtra is scheduled to be introduced in 2027. Valtra will replace the standard Mac chips currently used in Apple Intelligence servers to handle cloud-based AI requests, thereby enhancing the performance and reliability of AI services.

Apple’s AI strategy focuses on on-device processing while simultaneously strengthening its cloud infrastructure. At last month’s Worldwide Developers Conference (WWDC), Apple proposed a hybrid approach: personalized information and basic AI features are processed on-device, while large-scale inference and complex tasks are handled by its proprietary cloud infrastructure, ‘Private Cloud Compute.’

Since on-device computation alone cannot cover the demands of complex AI inference, analysts note that AI servers are essential to further advance Siri AI and Apple Intelligence functions. This long-term partnership with Broadcom is interpreted as a strategy to invest in custom semiconductors and to strengthen the supply chain, supporting Apple’s AI infrastructure.

This content was produced with the assistance of AI translation services.