[News] Apple Reportedly Pushes In-House AI Chip, Eyes 2027 Server Deployment; Foxconn to Gain

Apple is entering the race to develop in-house AI chips. According to Economic Daily News, citing sources, the company is working with Broadcom to develop its first proprietary AI chip, dubbed “Baltra,” with AI servers deployment expected to begin in 2027. The move aims to further advance Apple Intelligence services and, in turn, support sales of end-device hardware, with Apple’s key manufacturing partner Foxconn expected to be among the main beneficiaries, the report adds.

The report, citing Wccftech, indicates that Apple’s in-house AI server chip, “Baltra,” is intended for AI inference and is expected to adopt TSMC’s 3nm N3E process, with design work likely to be completed in about one year.

Wccftech also notes that Apple is unlikely to train large-scale AI models for now, as it has already reached an agreement with Google to deploy a customized 3-trillion-parameter Gemini model to power Apple Intelligence in the cloud. Under the deal, Apple is reportedly set to pay Google about US$1 billion per year for access to the model.

Apple’s AI Server Build-Out Gains Momentum, Foxconn Emerges as Key Beneficiary

Apple’s in-house AI chip development is moving forward steadily. As noted by Economic Daily News, Apple announced in late October 2025 that its Houston, Texas facility had entered production ahead of schedule and begun manufacturing AI servers. The report adds that as Apple Intelligence features roll out across new iPhone models, AI capabilities will continue to expand, driving higher demand for AI server capacity.

Therefore, institutional investors cited by Economic Daily News note that although Apple’s AI server purchases are far smaller in scale than those of NVIDIA or the four major cloud service providers (CSPs), Apple remains Foxconn’s largest customer. As the primary assembler of iPhones, Foxconn has also emerged as a key supplier of Apple’s AI servers, giving it strong positions across both cloud infrastructure and end devices.

As Economic Daily News highlights, Foxconn Group’s orders span both NVIDIA-based GPU servers and ASIC servers. The report adds that Foxconn holds a market share of more than 40% across major AI platforms, and that its ASIC server shipments in 2026 are expected to sustain rapid growth or even double. Against this backdrop, the report notes that market estimates put the split between GPU- and ASIC-based servers at roughly 8:2, a mix Foxconn is also expected to increasingly align with.

Read more

• [News] Foxconn Teams Up With OpenAI on AI Hardware; Output Hits 1,000 Server Racks a Week
• [News] Apple Rumored to be Developing 7 In-House Chips

(Photo credit: FREEPIK)