How Apple Plans to Reduce Its Reliance on Nvidia for AI Workloads

Expanding the Broadcom partnership to a $30B+ commitment – On July 7, 2026, Apple announced a new multiyear deal with Broadcom expected to exceed $30 billion, covering custom silicon components and wireless connectivity technology for a wide range of products. The deal extends through 2031 and will produce more than 1 billion U.S.-made chips.

Adding Intel as a U.S. manufacturing partner – In June 2026, Apple partnered with Intel to produce custom chips, a move tied to U.S. domestic semiconductor goals. Analysts note that advanced chips from this deal are still 2–3 years away from volume output, if they materialize at all.

Using Nvidia GPUs via Google Cloud in the interim – In a notable twist, Apple software chief Craig Federighi confirmed (June 2026) that Apple's advanced AI model "FM Cloud Pro" runs on Nvidia GPUs inside Google Cloud — effectively acknowledging that Apple's private infrastructure is not yet sufficient for its most demanding AI workloads.

How This Fits the Broader AI Arms Race

• Apple's approach is deliberately different from rivals. Rather than building giant GPU clusters of its own, Apple has historically minimized capital expenditure on AI infrastructure, buying compute from cloud partners instead. This frugal posture is now shifting into a more vertical-integration model.
• The broader industry is moving in the same direction: TrendForce projects custom ASIC shipments from cloud providers will grow 44.6% in 2026 vs. 16.1% for GPUs, as hyperscalers invest in their own silicon.
• Apple's in-house silicon advantage (M-series, A-series) gives it a unique foundation: it already runs AI inference for Apple Intelligence on-device using its own chips. The Baltra server chip extends this philosophy to the data center.

John Ternus's Incoming Leadership and AI Strategy

• John Ternus, Apple's longtime hardware engineering chief, was named CEO effective September 1, 2026, with Tim Cook becoming executive chairman. Apple also promoted silicon chief Johny Srouji to the newly created role of "chief hardware engineer," effectively making him Ternus's hardware successor.
• Ternus has articulated a "product-first" AI philosophy. In a recent interview, he stated: "We never think about shipping a technology" — instead, Apple asks how AI tools can serve users. This contrasts with Microsoft and Google, which have poured tens of billions into AI infrastructure and models.
• The challenge: Apple has been seen as trailing in AI, and Ternus must articulate a fuller AI strategy. His hardware expertise is viewed as an asset for Apple's silicon-driven AI push. The Baltra chip, Broadcom deal, and potential chip acquisitions are likely pillars of the revamped AI roadmap he will oversee.