Apple, known for designing its own chips, is now actively pursuing AI chip acquisitions after its internally designed servers failed to meet performance expectations for AI workloads.
- Apple’s AI servers underperform compared to Nvidia-powered alternatives
- Next-generation Apple AI server chip delayed until 2029
- Apple loosens cash reserves and management is more acquisition-friendly
What happened
Apple's traditionally strong chip design capability is facing a critical challenge as its current AI servers based on the M2 Ultra chip struggle to handle the heavy AI workloads required by its Gemini-powered Siri. This shortfall has forced Apple to rely on Nvidia chips hosted in Google’s cloud, marking a significant departure from its usual in-house chip reliance.
To address the slower-than-expected development of its next-generation server chip, codenamed Baltra, which Bloomberg reports may not arrive until 2029, Apple has begun exploring acquisitions of AI chip startups. The company has engaged bankers and approached potential targets, signaling a shift in strategy for the normally acquisition-averse firm.
Why it matters
Apple’s ability to design its own chips has been a major competitive advantage, underpinning its trillion-dollar valuation and product differentiation. The inability to meet AI hardware demands quickly enough risks ceding ground to competitors relying on third-party chips, particularly Nvidia, whose hardware currently powers Apple’s AI services.
This marks a strategic pivot for Apple, which has historically preferred organic growth over large acquisitions. The move to acquire AI chip companies could speed innovation and reduce external dependencies, but also highlights the growing complexity and cost of AI hardware development in the semiconductor industry.
What to watch next
Apple’s evolving hardware leadership team, with John Ternus assuming the CEO role and chip chief Johny Srouji overseeing all hardware, may drive increased willingness to invest heavily in acquisitions and strategic partnerships. Watch for potential deals, especially involving startups specializing in AI chip design or model optimization.
Additionally, Apple’s $30 billion chip supply agreement with Broadcom extended through 2031 and talks with AI model optimization startups like PrismML demonstrate a multi-pronged approach to reducing reliance on Nvidia. Observers should monitor how these investments impact Apple’s AI service performance and chip roadmap over the coming years.