Apple has quietly started using Nvidia chips to power some of its artificial intelligence work, a move that runs counter to the company’s long-standing preference for designing its own silicon. The shift, confirmed by people familiar with the matter, underscores the technical challenges Apple faces in scaling AI without relying on the chipmaker it has historically avoided.
Why the decision matters
Apple’s internal chip teams have built a reputation for creating custom processors—from the A-series in iPhones to the M-series in Macs—that tightly integrate hardware and software. But for training large AI models, the company apparently found that existing in-house solutions couldn’t match the raw computing power of Nvidia’s graphics processing units. The GPUs are widely used across the industry for machine learning because they handle parallel calculations efficiently.
That dependency is a notable departure for a company that has tried to minimize reliance on outside chip suppliers. Apple’s relationship with Nvidia has been fraught in the past; the two companies clashed over pricing and power consumption in earlier product cycles.
Internal resistance
Not everyone inside Apple was comfortable with the move. Some engineers and executives argued that using Nvidia chips could undermine the company’s long-term hardware strategy and expose it to supply chain risks. The reluctance wasn’t just ideological—it also reflected concerns about sharing technical details with a competitor. Nvidia supplies chips not only to Apple’s rivals in cloud computing but also to other big tech firms building AI systems.
Despite the pushback, the teams working on AI projects eventually won approval to procure Nvidia hardware. The decision was framed as a pragmatic necessity: Apple’s AI ambitions, including features for Siri, image recognition, and autonomous systems, require massive compute clusters that Nvidia’s GPUs can deliver today.
What’s at stake for Apple’s AI push
Apple has been slower than peers like Google and Microsoft to integrate generative AI into its products. The company has invested heavily in machine learning research, but rolling out consumer-facing AI tools at scale demands a different kind of infrastructure. Using Nvidia chips gives Apple a shortcut to build that infrastructure without waiting for its own silicon to catch up.
Still, the arrangement is likely temporary. Apple is believed to be developing its own AI accelerator, codenamed something that would eventually replace the Nvidia hardware. Whether that project can deliver the performance needed—and do so on a timeline that satisfies Apple’s product road map—remains an open question inside the company.
The next test will come later this year, when Apple is expected to preview new AI features at its annual developer conference. If those features rely on models trained with Nvidia’s help, the company will have to acknowledge using a rival’s tech—or stay quiet and let the chips speak for themselves.
