Apple just committed to buying custom silicon from Broadcom for the rest of the decade and beyond. Bloomberg reported on July 6, 2026, that the two companies extended their chip supply relationship through 2031, locking in years of joint development work on processors built specifically for Apple’s product line. The length of the commitment signals how central custom chips have become to Apple’s AI strategy, not just its phones and laptops.
The agreement covers ASICs (application-specific integrated circuits), chips engineered for one narrow job rather than the flexible, general-purpose computing that CPUs and GPUs handle. That specialization is precisely why ASICs have become a favored building block for running AI workloads at scale: stripping out unneeded flexibility cuts power draw and cost per operation. Under the extended terms, Broadcom and Apple will co-develop this silicon across several future generations of Apple products, not a single chip cycle.
Apple’s own roadmap explains the urgency behind locking in supply this far out. The company intends to bring its first in-house AI servers online in 2027, shifting inference work for its AI features onto hardware it designed rather than continuing to depend entirely on chips bought from outside vendors. That marks a meaningful expansion of Apple’s silicon ambitions beyond the A-series and M-series chips that power its consumer devices, into the data center gear that actually runs AI models.
Broadcom already sells custom AI accelerator chips and networking silicon to major cloud providers, a business line that made it a natural partner for Apple’s shift toward owned infrastructure. Extending Apple’s contract to 2031 deepens that relationship on both fronts Apple cares about: the custom silicon inside its consumer devices and the infrastructure behind its data center AI push.
Bloomberg’s report does not disclose the contract’s dollar value, the specific ASIC designs involved, or which Apple products receive the new silicon first. Those omissions matter. Whether the chips target consumer devices, server inference, or both will determine how much of Apple’s AI compute actually shifts away from third-party suppliers, and how much of Broadcom’s own AI accelerator revenue depends on a single customer through the end of the decade.
A six-year supply lock also reads as insurance against the GPU shortages that have periodically squeezed AI infrastructure buildouts since 2023. Securing Broadcom’s engineering capacity now protects Apple’s 2027 server launch date from allocation risk years before it needs to ship. For any team building AI products on Apple’s platform, 2027 is the date to watch: that is when Apple’s own silicon, not licensed third-party chips, starts setting the pace for on-device and cloud AI features.
Bloomberg first reported the extended Broadcom-Apple chip partnership on July 6, 2026.
