Apple and Broadcom extend the chip partnership to 2031, underscoring why custom AI hardware still matters

Apple is extending its chip partnership with Broadcom through 2031, an arrangement aimed at developing and supplying custom silicon for multiple generations of Apple products, including AI-capable server processors that Apple can use to run and scale its AI features. The extension reflects a practical reality of modern smartphone and personal-computing AI, where performance and power efficiency often depend on chips designed specifically for a company’s software stack rather than off-the-shelf parts.

In coverage of the agreement, Yahoo Finance highlighted that Broadcom will help develop custom AI server chips as part of the extended relationship, tying the hardware work to Apple’s “Apple Intelligence” initiative. The emphasis on custom AI processors suggests Apple is trying to control critical bottlenecks in the AI pipeline, from latency and throughput to energy use across devices and the infrastructure that supports more demanding workloads.

Broadcom, for its part, framed the expanded collaboration as a multi-year set of supply and development commitments. Market reporting in the last day said the companies’ new terms extend “through 2031” and cover “custom ASIC silicon” for multiple generations, describing it as a set of design wins rather than a one-off refresh. ASICs are application-specific integrated circuits, chips built for particular tasks instead of general-purpose computing.

One market read of the deal is that it reduces a key risk for Broadcom’s business model. StockStory, summarizing the market reaction and the perceived customer demand profile, said analysts estimate Apple accounts for about 20% of Broadcom’s annual revenue. Locking in Apple as a large, recurring customer through 2031, and expanding beyond wireless and connectivity into custom AI-capable chips, was described as addressing investor concerns that Apple could gradually pull more work in-house.

Broadcom’s stock reaction to the news was consistent with that interpretation. StockStory reported that Broadcom shares jumped about 4.1% early in the session after Broadcom disclosed the extended, multi-year agreements with Apple. Separate market coverage also said the update helped ease perceived 2031 revenue risk for Broadcom, reinforcing the idea that the agreement is not only about near-term product cycles but also about longer-term supply visibility.

For Apple, the logic is straightforward even when the details are technical. Building AI features that work well on-device often requires specialized accelerators for neural network computations and tight integration with the operating system and model execution software. When workloads spill over to servers, custom AI-capable chips can be used to improve cost efficiency and performance per watt, which matters for both user experience and the economics of scaling AI services.

Still, several elements of the agreement are not fully visible in the public, market-facing reporting around the announcement. It is not clear from the accessible summaries what specific chip families or performance targets are included in the “custom AI server chips” language, how quickly designs move from development into mass production, or how the mix of on-device versus server AI workloads will evolve across Apple’s product generations through 2031.

What to watch next is whether Apple provides more operational details about how these chips will be deployed for Apple Intelligence, and whether Broadcom’s disclosures give clearer timelines around design wins and shipment volumes. For the market, the key question is whether extending the relationship through 2031 translates into broader, repeatable AI hardware demand, or whether the custom silicon scope remains limited to a subset of Apple’s AI compute needs.