Apple pursues semiconductor acquisitions to close AI server chip gap

Apple’s in‑house M2 Ultra server chips have struggled to handle heavy AI workloads such as Google’s Gemini models used for Siri, prompting the company to rely on Nvidia GPUs hosted in Google Cloud. The shortfall has delayed the launch of Apple’s own Baltra AI server chip, now expected no earlier than 2029.

To address the gap, Apple is actively discussing the purchase of AI‑focused semiconductor startups with investment banks and chip designers. The talks follow recent large‑scale moves, including a $2 billion acquisition of Israeli AI startup Q.ai and a $30 billion multi‑year supply agreement with Broadcom for custom chips. CFO Kevan Parekh indicated Apple is moving away from its cash‑neutral policy to fund such deals, while new hardware chief John Ternus is reportedly more open to multibillion‑dollar purchases.

Apple also plans a longer‑term solution with its Project ACDC data‑center inference chip, slated for the second half of 2026, but any further delays could extend dependence on Nvidia and Google Cloud for AI services.