Apple and Broadcom have collaborated to design a server processor tailored specifically for Apple’s AI features in its operating systems, known as Baltra. Mass production should begin sometime during 2026.
Apple may take this step to lessen their dependence on AI chip manufacturer Nvidia, although they have found considerable success designing and producing their own chips such as M-series processors that replaced Intel chips in Mac computers.
The Information Reports
Apple and Broadcom have joined forces to develop the first chip designed specifically for artificial intelligence processing, according to The Information. This initiative aligns with other tech giants attempting to design their own AI chips in order to reduce dependence on Nvidia processors that are high cost and limited supply. Baltra, Apple’s planned AI processor is expected to enter mass production by 2026 using Taiwan Semiconductor Manufacturing Company’s advanced 3nm process technologies.
This chip will enable Apple to accelerate AI tasks such as inferencing – where large language models interact to generate outputs – more quickly, as well as offload processing of its on-device Apple Intelligence features to the cloud – recently made available through iOS 18.2, iPadOS 18.2, and macOS 15.2.
Broadcom’s efforts in AI hardware could further bolster its market position after the rise in demand for generative AI systems. Already designing server chips for leading cloud providers, its partnership with Apple could strengthen its leadership in ASIC design for AI-powered ASICs.
The Information cites three sources
Apple already produces processors for its iPhones and Macs that contain some on-device artificial intelligence features, yet reports indicate the company is working on developing its own custom server processor using Broadcom to accelerate AI tasks such as photo analysis or voice command interpretation, according to The Information. TSMC will reportedly manufacture Baltra using their advanced 3nm process technology.
Apple’s move is expected to further the trend of technology companies seeking to reduce their dependence on Nvidia’s expensive and limited supply of AI chips, leading to greater demand for specialty AI chips produced by chipmakers like Broadcom that produce them for modern tech applications such as smartphone sensors. On hearing of this news, Broadcom shares rose; currently trading around $250. Founded in 1991 in San Jose, California – Broadcom currently supports various computer applications with both semiconductor technologies and software products that provide support.
The Information cites Taiwan Semiconductor Manufacturing Company (TSMC)
Apple Intelligence features require AI processing power that must be brought closer to home for processing purposes, which has necessitated designing a custom processor known as Baltra which should begin mass production sometime around 2026.
Engineers in Herzliya, Israel who worked on Apple’s M-series Mac chips are leading this project. Fabrication will likely occur using TSMC’s advanced N3P process; while Broadcom will play an instrumental role in providing networking technology essential for enabling devices to connect to cloud services for processing.
TSMC is one of the world’s premier microchip producers and is well known for its commitment to social responsibility, such as environmental protection, employee welfare and community outreach. Additionally, this award-winning corporation has received praise from both domestic magazines as well as global publications for its corporate governance practices and expansion in America beginning with Trump administration’s passage of CHIPS Act in 2019.
The Information cites Broadcom
Apple needs servers in order to scale its AI processing, and while it has designed processors specifically tailored for AI processing on its devices, Broadcom may soon provide its first server chip designed specifically for AI processing.
Codenamed Baltra, Broadcom’s new chip will likely go into mass production by 2026 and focus on its networking technology – essential for connecting devices and networks for AI processing. According to The Information, Broadcom will focus specifically on improving this aspect.
Apple’s silicon design team previously focused on producing its M-series processors that replaced Intel chips in Mac computers, however with this partnership marks a shift in priorities as engineers now devote themselves to designing AI chips with Apple silicon instead of creating its M4 Extreme chips glued together as planned. In order to free up engineers for working on AI chips from other vendors; Apple cancelled plans for such an M4 Extreme chip featuring multiple M4 Extreme chips glued together (intended as an alternative GPUs from Nvidia that dominate AI chip markets), reports The Information. At its developer conference earlier this year Apple Senior Vice President of software Engineering Craig Federighi explained how some AI features would run on devices while others would use servers powered by Apple silicon servers powered by servers powered by Apple silicon servers powered by Apple silicon servers powered by Apple silicon.