Apple explores acquisitions of AI chip startups to strengthen AI server infrastructure after Baltra delay
Apple $AAPL is exploring AI chip acquisitions to fix its server bottleneck, per The Information. Its M2 Ultra servers hit performance limits and heavy Siri AI workloads run on Nvidia $NVDA hardware in Google Cloud. The silicon-first company is now shopping to close that gap.
Apple $AAPL is shopping for chip companies, a notable shift for a business that rarely makes large acquisitions. The Information reports that Apple has talked to bankers and semiconductor startups about possible deals to strengthen the server side of its AI infrastructure, and the report has been picked up by Engadget and Reuters. The background: Apple's AI servers run on its M2 Ultra chips, and they have hit a performance ceiling. The heavy lifting behind the new Siri, powered by a Google Gemini model, currently runs on Nvidia $NVDA hardware in Google Cloud. Apple's next-generation server chip, codenamed Baltra, was reportedly planned for this year but has slipped, and Bloomberg reports that a server chip based on the M7 Ultra will not be ready before 2029. Just last week, Apple committed to buying 30 billion dollars of US-made Broadcom $AVGO chips. Apple built its silicon reputation on consumer devices, from the PA Semi purchase in 2008 for 278 million dollars to the M-series Macs. Data center AI chips are a different discipline, and the acquisition search signals that Apple's in-house design cycle is not keeping pace with what its AI features require. With 45.6 billion dollars in cash and equivalents at the end of March, Apple has the balance sheet to pay chip-market premiums.