Skip to content
Source · Daily Brief

AI Daily Brief — 9 September 2025

A triple-marquee Tuesday. Apple Awe Dropping. Mistral + ASML official €1.7B. NVIDIA Rubin CPX. A day that simultaneously redefined the consumer AI client, the European AI champion, and the inference data centre.

Top stories

  • Apple holds “Awe Dropping” event. Tim Cook called iPhone Air “biggest leap ever for iPhone” — Apple’s thinnest (5.6mm) with grade-5 titanium frame. iPhone 17/17 Pro/17 Pro Max, three new Apple Watches, AirPods Pro 3. All ship Sep 19. via Apple
  • A19 Pro chip with neural accelerators in every GPU core. iPhone 17 Pro/Pro Max ship with A19 Pro on TSMC N3P, 6-core CPU (2P+4E) up to 4.26GHz, 12GB LPDDR5X (9600 MT/s). Neural accelerators added to each GPU core — up to 40% better sustained perf and ~4x peak GPU compute vs A18 Pro for Apple Intelligence workloads. via Apple
  • Apple debuts in-house C1X 5G modem + N1 Wi-Fi 7 chip. iPhone Air gets Apple-designed C1X cellular modem (2x faster than C1, 30% less energy, sub-6GHz only — no mmWave). All iPhone 17 models get N1, Apple’s first in-house wireless chip handling Wi-Fi 7, Bluetooth 6, Thread — displacing Broadcom. Major silicon vertical-integration step. via 9to5Mac
  • NVIDIA unveils Rubin CPX — million-token-context inference GPU. New GPU class for prefill/compute-bound phase of long-context inference: 30 PFLOPS NVFP4, 128GB GDDR7, 3x faster attention vs GB300 NVL72. Implements “disaggregated inference” — CPX chews context, standard Rubin handles decode in same rack. Targets 128K → 1M+ scaling wall. Availability: end of 2026. via NVIDIA
  • Vera Rubin NVL144 CPX rack — 8 exaFLOPS, 100TB memory. Integrated rack pairs 144 Rubin CPX + 144 Rubin GPUs + 36 Vera CPUs for 8 exaFLOPS NVFP4 (7.5x GB300 NVL72), 100TB fast memory, 1.7PB/s bandwidth. NVIDIA pitches $5B in token revenue per $100M invested — inference as money-printing primitive. via NVIDIA Tech Blog
  • Mistral AI closes €1.7B Series C at €11.7B. Europe’s most valuable AI company. ASML led €1.3B, taking ~11% fully diluted + Strategic Committee seat (CFO Roger Dassen). Existing investors DST Global, a16z, Bpifrance, General Catalyst, Index, Lightspeed and NVIDIA participated. via Mistral

Who shipped

Apple shipped iPhone Air + A19 Pro + C1X + N1. Mistral + ASML shipped a partnership. NVIDIA shipped Rubin CPX.

Open-source pulse

AirPods Pro 3 add Apple Intelligence Live Translation + heart-rate sensing — in-ear live translation for face-to-face conversations (on-device via Apple Intelligence on iPhone, triggered by pressing both stems or “Siri, start Live Translation”). 2x more noise cancellation than AirPods Pro 2. via Apple

Money, infra & hardware

Apple Watch Series 11 ($399) adds blood-pressure monitoring/hypertension detection and 5G. Ultra 3 ($799) adds satellite connectivity in 49mm titanium case with 3000-nit display, 42h battery (72h Low Power). SE 3 ($249). iPhone 17 Pro Max ships with 6.9-inch ProMotion display, Ceramic Shield 2, A19 Pro at $1,199 (same as 16 Pro Max). ASML + Mistral sign long-term strategic partnership — deploy Mistral models across ASML’s product portfolio, R&D and operations. Positioned as flagship of European tech sovereignty by French digital minister Clara Chappaz.

Quiet corners

Apple keeps Siri’s AI overhaul out of iPhone 17 launch — Visual Intelligence is the AI headline. AI talking points largely WWDC rehash. via TechCrunch Mistral Compute infrastructure — 18,000 NVIDIA Grace Blackwell Superchips deployed across 40 MW data center in Essonne, France.

By the numbers

  • 5.6mm / 4x GPU / A19 Pro — iPhone Air thinness / peak compute vs A18 Pro / chip
  • C1X / N1 — Apple in-house modem / Wi-Fi 7 chip
  • 30 PFLOPS / 128GB / 3x / 2026 — Rubin CPX NVFP4 / GDDR7 / attention vs GB300 / availability
  • 8 exaFLOPS / 100TB / 1.7PB/s — NVL144 CPX rack specs
  • €1.7B / €1.3B / 11% / €11.7B — Mistral round / ASML / stake / valuation

Compiled by AI Feed’s editor from verified web sources for 9 September 2025.