Apple and Broadcom have extended their chip supply partnership through 2031, with the renewed agreement centred on custom application-specific integrated circuit (ASIC) silicon designed to power Apple’s next generation of AI server infrastructure. The deal, announced on July 6, 2026, cements Broadcom as a long-term strategic partner in Apple’s ambition to build proprietary cloud computing hardware for its artificial intelligence services.
What Happened
Apple and Broadcom confirmed a multi-year extension that will see Broadcom supply custom silicon across multiple generations of Apple products through at least 2031. The centrepiece of the renewed deal is ASIC chip development for Apple’s AI server programme, internally codenamed Baltra. Apple is designing AI server chips with Broadcom technology integrated at the silicon level, targeting deployment as early as 2027 to support the cloud-side processing required by Apple Intelligence features. Broadcom’s established role as Apple’s supplier of radio-frequency chips for cellular connectivity, as well as Wi-Fi and Bluetooth networking semiconductors, continues under the renewed terms.
Why It Matters
The partnership extension signals that Apple is accelerating its plans for proprietary AI cloud infrastructure at a time when competitors including Google, Microsoft, and Amazon are spending aggressively on data centre AI hardware. By locking in Broadcom as a long-term ASIC partner, Apple reduces its dependence on third-party inference platforms and builds the silicon-level foundation for Apple Intelligence’s cloud services without ceding chip design control to a general-purpose hardware vendor.
For Broadcom, the deal secures one of the most consequential customer relationships in the global semiconductor industry. Apple’s ASIC orders represent significant sustained revenue and deep engineering collaboration, and the 2031 horizon gives Broadcom a long planning window for dedicated research and development aligned with Apple’s product roadmap. Broadcom shares rose on the news as investors priced in the scale of the extended commitment and the strategic importance of the AI server chip programme to both companies.
Background and Context
The trend toward custom silicon among frontier technology platforms is intensifying across the industry. Anthropic is currently in talks with Samsung to build a custom 2nm AI inference chip, while DeepSeek has moved to design its own inference silicon to reduce dependence on Nvidia GPU infrastructure. Apple’s Broadcom deal fits squarely within this broader shift: as frontier AI workloads grow in scale and specialisation, owning the full hardware stack from chip design to software runtime is increasingly a competitive necessity rather than an optional efficiency gain.
Apple has already demonstrated the advantages of vertical silicon integration with its M-series and A-series chips, which handle the on-device components of Apple Intelligence. The Baltra AI server programme extends that philosophy into the cloud, where Apple offloads computation too demanding for a smartphone or laptop to private server infrastructure, rather than relying on Amazon Web Services or other third-party platforms. This approach gives Apple tighter control over performance, latency, and data security for AI features — all factors that carry considerable weight with Apple’s privacy-focused consumer base.
What Comes Next
Broadcom’s Baltra AI server chips are expected to enter production as early as 2027, with broader deployment to Apple’s data centre infrastructure in the years that follow. Observers will be watching whether the AI server buildout meaningfully expands the capability and responsiveness of Apple Intelligence services — particularly for tasks that require more compute than a smartphone can handle on-device, such as advanced natural-language reasoning, real-time image analysis, and complex cross-app automation.
The deal also raises questions about the future shape of Apple’s relationship with Nvidia, which currently supplies GPUs for some of Apple’s AI data centre workloads. A successful Baltra programme could progressively reduce Apple’s reliance on Nvidia for inference compute, following a trajectory similar to Apple’s transition away from Intel processors for Mac hardware. For the semiconductor industry more broadly, the Apple-Broadcom extension is further evidence that custom ASIC design — purpose-built chips optimised for a single company’s specific workloads — is becoming the dominant architectural choice for AI infrastructure at scale.
