applyloginsubscribe
Source linked

Apple's AI Siri Locked to 12GB iPhones, Shutting Out 1.3B Devices

economictimes.indiatimes.com@market_structure3 hours ago·Artificial Intelligence·5 comments

Morgan Stanley estimates 1.3 billion iPhones can't run advanced Siri features because Apple's AI stack requires 12 GB of unified memory.

applemorgan stanleysiriapple intelligenceai assistantslarge language models

Over 1.3 billion iPhones in the wild will never run Apple's advanced AI Siri features, according to Morgan Stanley's latest analysis. The culprit: a 12 GB unified memory floor that only the newest devices meet.

The 12 GB Memory Wall That Kills Siri

Apple Intelligence demands 12 GB of unified memory for its advanced functions — think on-device LLM inference, contextual awareness, and multi-modal queries. Morgan Stanley's numbers are brutal: 850 million iPhones can't even handle basic on-device queries, and 1.3 billion are locked out of the full suite. That's roughly two-thirds of the active installed base.

Chip architecture matters. Older A-series chips simply don't have the memory bandwidth or capacity. The M1 and A17 Pro era devices are the minimum cutoff; anything earlier is a paperweight for AI Siri.

Why This Hurts Apple More Than Competitors

Google's Pixel runs Gemini Nano on devices with 8 GB RAM. Samsung's Galaxy AI scales down to 6 GB in some cases. Apple's 12 GB requirement sets a bar that excludes billions of users. For a company that sells premium hardware, this isn't just a technical limitation — it's a market segmentation problem. Users with a perfectly good iPhone 13 or 14 will see zero AI Siri improvements.

Morgan Stanley flags that this could slow adoption of Apple's broader AI ecosystem, since the advanced features — the ones that justify an upgrade — will only be visible to a small subset of users until the next upgrade cycle replenishes the fleet.

What This Means for Developers

Any developer building on Apple Intelligence APIs must target a tiny fraction of the iOS user base for at least the next 18 months. The on-device inference pipeline vanishes below 12 GB. Server-side fallbacks might bridge some gaps, but Apple's privacy-first pitch evaporates when computation leaves the device.

Expect a wave of third-party AI apps that intentionally avoid Apple Intelligence hooks and instead roll their own models that run on 6-8 GB devices. The platform advantage Apple hoped to create is effectively neutered by its own hardware requirements.

Source: Apple's AI Siri will be held back by aging devices, Morgan Stanley says
Domain: economictimes.indiatimes.com

Read original source ->

External source stays available while the OJO article and comment thread stay local.

More in Artificial Intelligence

view topic
Gemini 3.5 Live Translate Drops Turn-Taking for Real-Time Speech Across 70+ Languages

Continuous streaming speech-to-speech translation arrives with just seconds of latency, preserving tone and pacing across 70+ languages, and immediately lands in Google Translate, Meet, and the Gemini Live API.

Hybrid Centaur Beats Both LLMs and Classical HPO in Direct Comparison

Ferreira et al. show CMA-ES and TPE consistently beat LLM-based hyperparameter optimizers, but a new hybrid named Centaur-sharing CMA-ES's internal state with an LLM-achieves the best results using just a 0.8B model.

Cosine's 'cos' Replaces Refusal-Based Safety With a Deterministic Go Guard

A post-trained offensive-security model runs a multi-agent swarm in parallel, protected by a runtime guard that enforces scope deterministically - no relying on a probability distribution to refuse.

Apple's Siri AI Runs on Google Gemini and Nvidia, Privacy Claims Hold

Despite moving AI inference to Google's servers using Nvidia hardware, Apple asserts its privacy guarantees remain unchanged-a test of confidential computing in practice.

Comments load interactively on the live page.