Skip to content
Source · Daily Brief

AI Daily Brief — 18 April 2025

Good Friday with three live storylines: Jensen continuing his Beijing diplomacy after the H20 shock, Perplexity landing its first smartphone OEM partnership, and OpenAI’s Flex pricing details circulating widely. The Section 232 semiconductor probe overhang stays in place.

Top stories

  • Jensen Huang visits Beijing after H20 export ban. Two days after the indefinite export ban and Nvidia’s $5.5B writedown, Jensen made his second Beijing trip of the year. Met Chinese officials and reaffirmed Nvidia’s long-term commitment to China, vowing to adapt its product lineup to new US regulatory constraints. via TechNode
  • Perplexity inks Motorola Razr distribution deal. Perplexity AI’s voice assistant to be embedded in upcoming Motorola Razr foldables (Razr / Edge 60 series) — its first major smartphone OEM distribution win. New device owners get three months of free Perplexity Pro. Motorola formally unveiled the partnership at the Razr 60 Ultra launch on Apr 24. via Phandroid
  • OpenAI Flex pricing details surface — 50% off o3 and o4-mini for non-prod. Lower-priority compute queue that halves API prices in exchange for higher latency and occasional resource unavailability. o3 Flex: $5/$20 per M tokens (vs $10/$40 standard). o4-mini Flex: $0.55/$2.20 (vs $1.10/$4.40). Tiers 1-3 developers must complete ID verification to access o3. via WinBuzzer
  • Section 232 semiconductor probe formally underway. Commerce Department formally launched the Section 232 investigations into semiconductors and semiconductor manufacturing equipment imports on national-security grounds (Apr 1 effective date, Federal Register notices published Apr 16). Commerce reports due within 270 days (~Dec 2025). The Section 232 plus H20/MI308 license requirement remains the dominant macro storyline for NVIDIA and AMD investors.

Who shipped

No major model releases. Perplexity made the day’s product move with the Motorola deal. NVIDIA‘s Jensen made the diplomatic move in Beijing. OpenAI rounded out Flex pricing for o3/o4-mini.

Open-source pulse

Wikipedia x Kaggle structured training dataset (Apr 16 launch — ~25 GB zipped, English + French, with summaries, descriptions, image links, infobox data and JSON-formatted article sections) continued to draw uptake — a soft counter to AI scraper-bot load that drove a ~50% surge in Wikimedia bandwidth over the prior year.

Money, infra & hardware

Markets stayed cautious on chip stocks ahead of the Easter weekend. Section 232 + H20/MI308 license shock the structural overhang.

Quiet corners

Anthropic, Google DeepMind, Meta, xAI, DeepSeek, Alibaba/Qwen, Mistral all quiet on the model side. No major funding round closed publicly.

By the numbers

  • $5 / $20 per million tokens — o3 Flex pricing
  • $0.55 / $2.20 per million tokens — o4-mini Flex pricing
  • ~3 months — Perplexity Pro free trial for new Razr owners
  • ~270 days — Section 232 report deadline from Apr 1 start
  • Most-mentioned company: NVIDIA

Compiled by AI Feed’s editor from verified web sources for 18 April 2025.