Apple Going All-In on AI Chips

2026-07-14 · Source: Spyglass · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Apple is intensifying its focus on AI capabilities within its Apple Silicon efforts, particularly with the upcoming M7 Ultra chip. This chip, slated for 2028, is expected to dramatically upgrade AI performance, potentially bringing it closer to dedicated AI accelerators like NVIDIA's Blackwell for inference workloads. A key advantage for Apple is its unified memory architecture, which could allow the M7 Ultra to compete with Blackwell for LLM inference if it achieves 1.5TB of memory. The article also notes Apple's plans for a new server chip based on the M7 Ultra by 2029, suggesting a potential shift in its AI server strategy and reduced reliance on external cloud providers. However, it acknowledges NVIDIA's continuous advancements and the increasing size of AI models as significant challenges.

Key takeaway

For investors tracking the AI chip market, Apple's aggressive push into AI-optimized silicon, particularly with the M7 Ultra targeting NVIDIA Blackwell-class inference by 2028, signals a significant competitive shift. You should monitor Apple's memory capacity advancements and server chip development, as these could reduce its cloud reliance and potentially position it as a future chip vendor, impacting current market leaders.

Key insights

Apple's M7 Ultra aims for NVIDIA Blackwell-class AI inference performance by 2028, utilizing unified memory.

Principles

• Unified memory architecture benefits on-device AI inference.
• AI inference and training have distinct hardware requirements.
• Future chip roadmaps must anticipate model growth.

In practice

• Design chips with high memory bandwidth for LLM inference.
• Prioritize on-device AI to mitigate cloud compute costs.
• Develop internal server chips for AI workloads.

Topics

• Apple Silicon
• AI Chips
• M7 Ultra
• NVIDIA Blackwell
• AI Inference
• Unified Memory
• On-device AI

Best for: AI Hardware Engineer, Investor, Tech Journalist

Related on AIssential

• Apple to skip M6 Pro and Max chips, fast-tracks M7 lineup for AI focus — Apple prioritizes AI performance by fast-tracking M7 chips with enhanced Neural Engines, altering its traditional release cycle.
• Nvidia’s Share of AI Inference Chip Market Appears to Be Rising — Nvidia's share in the AI inference chip market is reportedly growing.
• Apple Wins The AI Hardware Race — Apple's hardware, particularly Mac Mini, is a surprising winner in the AI market despite software delays.
• HUGE: Apple M7 Ultra Has 1.5TB Memory and a $35,000 Price to End NVIDIA — Apple's M7 Ultra, with 1.5TB memory, targets trillion-parameter AI models and data centers, disrupting its traditional chip roadmap.
• Apple’s Big AI, Siri and Software Launch — Apple's Siri is becoming an AI co-pilot, while Nvidia and SK Hynix partner on AI memory, and SpaceX prepares an AI-driven IPO.
• Jensen Huang just put Nvidia’s Blackwell and Vera Rubin sales projections into the $1 trillion stratosphere — Nvidia projects $1 trillion in Blackwell and Rubin chip orders by 2027, reflecting booming AI hardware demand.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Spyglass.