March 26, 2025 – Apple’s Commitment to Proprietary Servers Shaken as $1 Billion NVIDIA Deal Surfaces
Despite the bold assertions from Apple executives that its AI services operate exclusively on self-developed Apple Silicon servers, a recent report by Loop Capital analyst Ananda Baruah suggests a significant shift in strategy. Apple is reportedly spending approximately $1 billion to procure NVIDIA’s AI servers, potentially undermining its previously emphasized “privacy-first” technological approach.
A timeline of events, as reported in various blog posts, reveals the following developments:
April 2024: Rumors emerged that Apple would utilize its proprietary chips to build AI servers.
June 2024: Further speculation indicated that Apple’s data centers would fully adopt Apple Silicon.
September 2024: Craig Federighi, Apple’s Senior Vice President of Software Engineering, publicly confirmed that Apple Intelligence services run entirely on in-house servers, touting this as a “new industry standard for cloud processing.”
March 2025: Analysts reveal that Apple has ordered 250 NVIDIA NVL72 servers, with each unit costing between 3.7millionand4 million, totaling nearly $1 billion.
Each NVL72 server is equipped with 36 Grace CPUs and 72 Blackwell GPUs. Anticipated to hit the market on March 18, 2025, Apple may be among the first global customers to receive these servers.
Federighi previously highlighted that the proprietary servers construct privacy barriers through innovative techniques like “transparent log signature verification.” However, this procurement could imply that the computing power of Apple’s homemade chips is insufficient to meet the demands of generative AI. Alternatively, it could suggest that Apple is developing a hybrid architecture to balance performance and privacy concerns.
