Apple explores AI chip acquisitions as Baltra chip slips past 2026

Apple's next-generation AI server chip, code-named Baltra, has slipped past its planned 2026 debut, forcing the company to explore semiconductor acquisitions as it struggles to match Nvidia's performance with its own M2 Ultra-based servers.

"We're no longer targeting a specific net cash level, which gives us more flexibility," Chief Financial Officer Kevan Parekh told analysts, freeing up Apple's $45.6 billion cash pile for potential deals.

Apple has talked with bankers about possible acquisitions and approached semiconductor startups about selling themselves, according to The Information. The company's second-largest deal ever — the nearly $2 billion purchase of Israeli AI startup Q.ai in January — shows it is already spending more. Its biggest acquisition remains the $3 billion Beats purchase in 2014, while the $278 million buyout of PA Semi in 2008 laid the foundation for its entire chip business, producing the A-series iPhone processors and later the M-series Mac chips that now power its AI servers.

The delay leaves Apple reliant on Nvidia's H100 and H200 GPUs inside Google Cloud for the most demanding AI tasks, including the Gemini-powered version of the new Siri. Bloomberg reports a chip powerful enough to rival Nvidia may not arrive until 2029, with an M5 Ultra upgrade filling the gap in the interim. For a company that built its $3 trillion market cap on vertical integration — designing its own CPUs, GPUs, and neural engines — needing to outsource AI computing to a competitor's hardware marks a strategic inflection point.

The competitive stakes are high. Nvidia's data center revenue reached $130 billion in its latest fiscal year, powered by demand for AI training and inference workloads. Google, through its Tensor Processing Units and partnerships with Nvidia, has positioned itself as the infrastructure provider of choice for companies that cannot build their own AI silicon. Apple's delay means it remains a customer of both, rather than a competitor.

A $30 Billion Bet on Broadcom — and a Shopping List

Apple has already committed $30 billion to chips from Broadcom, a partnership it extended to 2031. It is also in talks with PrismML, a startup that shrinks large AI models to run directly on iPhones, according to The Information. On the manufacturing side, Apple's chips are built by TSMC on advanced nodes (3nm and below), making the Taiwanese foundry a critical single-source supplier for any future server silicon.

The leadership transition adds another variable. Tim Cook hands the chief executive role to hardware chief John Ternus in September, while chip chief Johny Srouji now oversees all of Apple's hardware. Both are engineers with a track record of vertical integration — and both may be more willing to buy their way out of a supply constraint.

Apple's reliance on external AI infrastructure carries implications for multiple tickers. Nvidia, trading at roughly 35 times forward earnings, benefits from Apple's continued dependence on its GPUs. Broadcom, which secured a $30 billion multiyear commitment, stands to gain regardless of whether Apple's in-house chip eventually materializes. For Apple itself, the delay raises questions about its ability to maintain its silicon advantage — a key pillar of its premium valuation relative to other hardware makers.

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