Apple shares rise 4% to record on AI chip acquisition talk

Apple Inc. has held discussions with bankers about potential chip acquisitions in recent months and approached semiconductor startups to gauge interest in a sale, according to a person familiar with the matter. The iPhone maker's M2 Ultra chips — built on TSMC's 5nm process — couldn't run Google's Gemini models efficiently, forcing Apple to rent Nvidia Corp. GPUs hosted in Google Cloud for more demanding workloads.

"Apple has held discussions with bankers about potential chip acquisitions in recent months and has also approached semiconductor startups to gauge their interest in a sale," the person said.

The company's future AI server chip, code-named Baltra, has been delayed, according to the report. Apple's current infrastructure relies on internally designed M2 Ultra processors for some workloads, but those chips have struggled with larger AI models. The company had to turn to Nvidia GPUs hosted in Google Cloud for more demanding tasks.

Apple shares rose more than 4% Wednesday and hit a fresh all-time high on July 15 as investors priced in potential M&A upside. A confirmed acquisition would reduce Apple's dependence on Nvidia and strengthen its in-house AI capabilities, potentially reshaping the competitive dynamics in the AI semiconductor space.

Apple's server chip struggles mirror a broader industry push. Amazon.com Inc., Alphabet Inc. and Microsoft Corp. have all developed custom AI chips to reduce reliance on Nvidia, whose H100 GPU — with 990 TFLOPS of FP16 performance and 80GB of HBM3 memory — dominates the data center market. The M2 Ultra, designed for Mac Pro workstations rather than data center-scale inference, lacks the high-bandwidth memory and parallel compute capacity needed for large language models. Running models like Gemini requires sustained memory bandwidth that consumer-grade chips cannot provide. Nvidia's H100 delivers 3.35 TB/s of memory bandwidth through its HBM3 stack, while Apple's M2 Ultra offers 800 GB/s — a gap that makes inference on large models impractical at scale. This performance shortfall forced Apple to rely on Google Cloud's Nvidia-powered infrastructure during the development of its revamped Siri.

Apple agreed to acquire Israeli startup Q.ai for $2 billion earlier this year, its second-largest acquisition after the $3 billion purchase of Beats Electronics in 2014. The deal signaled a greater willingness to pursue larger transactions as competition in AI intensifies. Acquiring a chip startup could accelerate development of the delayed Baltra server chip. Apple needs processors capable of running large AI models efficiently on its own infrastructure — a capability that would reduce cloud rental costs and give Apple greater control over its AI roadmap. The company has approached bankers and semiconductor startups in recent months, suggesting it is actively evaluating targets rather than passively monitoring the market.

Apple shares trade at roughly 30x forward earnings, a premium to the S&P 500's 21x but below Nvidia's 35x multiple. The AI chip acquisition speculation has added momentum to a stock that has already gained more than 20% this year. If Apple successfully acquires a chip startup and delivers Baltra on schedule, it could save hundreds of millions annually in GPU rental costs from cloud providers. Nvidia faces a potential long-term headwind if Apple reduces reliance on its GPUs, though any competitive threat remains years away. Chip development cycles from acquisition to production typically span 18 to 24 months, and Apple has yet to name a specific target. For now, Nvidia's dominance in AI data center chips remains unchallenged — the company controls an estimated 80% of the market for AI training and inference processors.

This article is for informational purposes only and does not constitute investment advice.