Skip to content
Source · Daily Brief

AI Daily Brief — 26 November 2025

Wednesday closed the pre-Thanksgiving cycle with light news but two structural threads. Perplexity + PayPal Instant Buy extended Black Friday coverage — in-chat shopping flow, stored PayPal credentials, MacRumors + Dataconomy noted PayPal-enabled merchants are supported broadly (not just initial Fabletics + Adorama + Newegg). Free for all users; remembers user preferences across sessions. TSMC: advanced-node capacity ~3× short of AI demand — CEO C.C. Wei joked he’d considered wearing a “No more wafers” T-shirt. 2nm booked through 2028; NVIDIA + Broadcom reportedly turned down for additional capacity. Microsoft Copilot Studio retires GPT-4o as default — Nov 26 cutoff for generative-orchestration agents to continue using GPT-4o (via “Continue using retired models” toggle); new default GPT-4.1 for all non-GCC customers. Trump’s Genesis Mission EO from Mon still dominating policy coverage — DOE supercomputers + AI + quantum across national labs for 20+ energy/materials/biotech challenges. BAT giants commit hundreds of billions CNY to AI infra — DigiTimes reported Baidu + Alibaba + Tencent capex commitments pushing back against “AI bubble” narratives; Alibaba Cloud holds 35.8% domestic AI cloud share vs ByteDance Volcano Engine 14.8%. Microsoft-NVIDIA-Anthropic $30B partnership from Nov 18 continued to drive coverage — Claude across Foundry + Copilot + 365 Copilot + Copilot Studio. EU Digital Omnibus from Nov 19 still anchoring EU AI-policy debate — defers AI Act high-risk obligations past Aug 2, 2026. Claude Opus 4.5 from Mon remains centerpiece of frontier-lab coverage as US labs wind down for Thanksgiving. Replit Design Mode + Fast Mode (free until Dec 1) shipped earlier in week. No major arXiv breakthrough this date — academic firehose quiet pre-holiday.

Top stories

  • Perplexity Instant Buy + PayPal extends Black Friday. In-chat shopping flow, stored PayPal credentials. Free for all users; remembers preferences. via PayPal
  • Dataconomy + MacRumors: broad PayPal merchant support. Not just initial Fabletics + Adorama + Newegg. via Dataconomy
  • TSMC: advanced-node capacity ~3× short of AI demand. Wei: considered wearing “No more wafers” T-shirt. 2nm booked through 2028. NVIDIA + Broadcom turned down for additional capacity. via Tom’s Hardware
  • Microsoft Copilot Studio retires GPT-4o as default. Nov 26 cutoff. New default GPT-4.1 for non-GCC customers. via Microsoft Learn
  • Trump Genesis Mission EO continues to drive policy coverage. DOE supercomputers + AI + quantum across national labs for 20+ challenges in energy + materials + biotech. via White House
  • BAT (Baidu + Alibaba + Tencent) commit hundreds of billions CNY to AI infra. Push back against “AI bubble” narratives. Alibaba Cloud 35.8% domestic AI cloud share vs ByteDance Volcano Engine 14.8%. via DigiTimes
  • Microsoft-NVIDIA-Anthropic $30B continues to drive coverage. Claude across Foundry + Copilot + 365 Copilot + Copilot Studio. Motley Fool: “possibly Microsoft’s most important AI deal yet.” via Microsoft
  • EU Digital Omnibus still anchoring EU policy debate. Defers AI Act high-risk obligations past Aug 2, 2026 deadline. via European Commission
  • Claude Opus 4.5 remains centerpiece. Best-on-coding/agents/computer-use at 1/3 cost of Opus 4.1. via Anthropic
  • Replit Design Mode + Fast Mode (free until Dec 1). Gemini 3-powered design generation in <2 minutes. via Replit

Who shipped

Perplexity + PayPal continued Instant Buy rollout. Microsoft retired GPT-4o default. TSMC + BAT capacity/capex narratives drove infra coverage. OpenAI, Anthropic, Google, Meta, xAI, DeepSeek quiet on launches.

Open-source pulse

No major drops. Olmo 3 + HunyuanVideo-1.5 + Fara-7B continued community uptake from prior days.

Money, infra & hardware

TSMC 3× capacity shortfall is the structural infra story of the week — 2nm booked through 2028 reshapes 2026-2027 chip availability. BAT capex commitments reframe China-side AI-bubble debate.

Quiet corners

Microsoft Copilot Studio GPT-4o → GPT-4.1 default switch is a quiet but consequential enterprise-AI default change.

By the numbers

  • ~3× / 2028 — TSMC advanced-node demand:capacity gap / 2nm book-through year
  • 35.8% / 14.8% — Alibaba Cloud / ByteDance Volcano Engine domestic AI cloud share
  • GPT-4o → GPT-4.1 — Microsoft Copilot Studio default switch
  • Aug 2, 2026 → Dec 2, 2027 — EU AI Act high-risk deferral target
  • Most-mentioned company: TSMC
  • Quietest segment: fresh model launches

Compiled by AI Feed’s editor from verified web sources for 26 November 2025.