The growing competition in AI chip manufacturing and cloud services has led to unexpected partnerships, with Apple revealing its use of Amazon Web Services’ custom AI chips for various services.
Key developments: Apple disclosed at AWS Reinvent conference that it employs Amazon’s proprietary AI chips for search services and is evaluating new chips for AI model pre-training.
- Benoit Dupin, Apple’s senior director of machine learning and AI, reported a 40% efficiency gain using Amazon’s Trainium and Graviton chips
- Apple has utilized AWS for over a decade to power services including Siri, Apple Maps, and Apple Music
- The company expects up to 50% improvement in efficiency with pre-training using Amazon’s new Trainium2 chip
Strategic implications: This rare public acknowledgment of a supplier relationship by Apple represents a significant endorsement for AWS in its competition with Microsoft Azure and Google Cloud.
- The partnership demonstrates AWS’s growing influence in the AI chip market, traditionally dominated by Nvidia
- Apple maintains relationships with multiple cloud providers, including Google Cloud, which it used for training iPhone AI services
- AWS is preparing to announce the availability of Nvidia Blackwell-based AI servers for rent
Apple’s AI approach: The company’s strategy differs from conventional cloud-based AI implementations, focusing on device-based processing with cloud support.
- Apple Intelligence, launched this fall, offers services like notification summarization and email rewriting
- The system will soon integrate with OpenAI’s ChatGPT and enhance Siri’s capabilities
- Processing occurs primarily on Apple devices (iPhone, iPad, or Mac), with complex queries handled by Apple-operated servers using M-series chips
Innovation landscape: The development of alternative AI chips marks a shift in the industry’s reliance on traditional processors.
- Cloud providers and startups are actively developing alternatives to expensive Nvidia graphics processors
- Apple’s adoption of custom chips could encourage other companies to explore non-Nvidia training approaches
- The focus on efficiency gains suggests a growing emphasis on cost-effective AI processing solutions
Future trajectory: While Apple’s endorsement of AWS’s custom chips signals a potential shift in AI infrastructure, the company’s hybrid approach of combining on-device and cloud processing may set a new standard for AI implementation in consumer technology.
Apple says it uses Amazon's custom AI chips