Apple and Broadcom Working on AI Server Chip | Cpu hardware list | Gpu vs cpu architecture | Cpu-z | Turtles AI

Apple is reportedly developing a custom server processor for AI, in collaboration with Broadcom. The project, known as “Baltra,” is expected to go into production in 2026, with a focus on advanced technologies such as 3.5D packaging. Apple already has experience with multi-die architectures and will likely leverage existing technologies.

Key points:

• Apple is designing a custom server chip for AI, with support from Broadcom.
• The codename for the project is “Baltra,” planned for 2026.
• The chip will be used for services and AI in Apple’s operating systems.
• The two companies already collaborate on 5G and could share advanced packaging technologies.

Apple is working on an ambitious project to develop a custom server processor that will power AI services in its operating systems. This processor, codenamed “Baltra,” could see the light of day in 2026, thanks to a collaboration with Broadcom, a leading semiconductor company. Although official information is still scarce, the project appears to be an integral part of Apple’s strategy to boost its AI infrastructure, a field that is becoming increasingly important in modern technologies. In particular, AI is being integrated into numerous aspects of Apple devices, from cloud services to virtual assistants, with a focus on performance and privacy. Against this backdrop, the support of a partner like Broadcom is not surprising, as the two companies have been collaborating on 5G solutions for some time. Broadcom’s advanced packaging technology, such as 3.5D, could also be used in the context of “Baltra” to optimize chip interconnection and power efficiency by leveraging high-bandwidth memory modules. Broadcom’s 3.5D technology is designed to combine chips in innovative ways, overcoming the limitations of traditional architectures. Key innovations include the use of Hybrid Copper Bonding (HBC) to reduce latency times and improve interconnect performance between chips. In addition, the use of multi-die packaging techniques and chiplet integration could enable unprecedented scalability. Although it is unclear whether Apple’s design will use exactly the same technologies as Broadcom’s, the two approaches seem complementary, as Apple itself has already implemented multi-die architectures in its chips, as in the case of the M2 Ultra. However, it remains to be seen how these developments will materialize, in a context where Apple prefers to keep any details confidential before official launches. It is not yet clear whether the partnership with Broadcom will lead to such a large-scale collaboration, but peAI interest and data center management seems to be the main driver behind the project.

Until then, we will continue to speculate on how Apple’s ecosystem, particularly its AI servers, will evolve in the near future.