• Lore Brief
  • Posts
  • Issue #165: Nvidia Makes Its Biggest Bet Yet on AI Inference

Issue #165: Nvidia Makes Its Biggest Bet Yet on AI Inference

PLUS: Meta buys ManusAI to push agents, Gemini’s traffic surges, and AI coding keeps accelerating

Author

Nathan Lands
December 31, 2025

Good morning, welcome to this week’s Lore Brief.

Over the last year, we’ve gained 25,000 new readers. Thank you for the support. My goal is to make Lore Brief the best executive summary in AI.

Wishing you all a Happy New Year!

This issue is brought to you by Factory, the fastest way to ship software with autonomous engineering agents.

  • NVIDIA licenses Groq inference tech in $20B deal → The non-exclusive agreement brings Groq's assets to Nvidia, with founder Jonathan Ross, President Sunny Madra, and key team members joining to scale AI inference globally. Valued at about 3x Groq's recent mark and Nvidia's largest deal ever. Read more here →

  • Meta acquires ManusAI in $2B+ deal to boost AI agents → The Singapore-based startup, focused on bridging human intent to AI actions, was acquired to advance Meta's agent technology amid rising competition. Manus raised $75M earlier in 2025 at a $500M valuation. Read more here →

  • Gemini sees soaring web traffic → Market share jumped from 5.4% to 18.2% over 12 months, driven by distribution in Google products. ChatGPT fell from 87.2% to 68.0% amid overall GenAI growth. Read more here →

Really cool way to see how much you’ve coded with AI this year in Factory:

A simple but powerful prompting idea by Jason Fried:

Awesome demo of Antigravity, Google’s Agentic Development Platform:

  • Andrej Karpathy reflects on AI's impact on programming → He’s amazed how fast AI is reshaping coding, with agents and tools now refactoring entire projects autonomously. He calls it an "alien artifact without a manual" that's triggering a profound shift in the field, leaving many programmers feeling they're falling behind the curve. Read more here →

  • SoftBank completes massive $40B investment in OpenAI → The Japanese giant sent the final ~$22B tranche, fully funding one of the largest private rounds ever after syndicating $10B and direct commits. This pushes SoftBank's stake above 10% and fuels OpenAI's huge infrastructure push, including the Stargate data center project. Read more here →

  • Qwen launches two Flash TTS models → Alibaba-backed Qwen released Qwen3-TTS-VC-Flash for 3-second voice cloning in 10 languages like Chinese and English. The models target voice styling and rapid multilingual cloning. Read more here →

  • Kapwing publishes AI Slop Report → The blog analyzes the rise of low-quality AI videos, finding 21-33% of YouTube Shorts are AI-generated. Key stats include Spain leading subscribers at 20M and South Korea views at 8.45B. Read more here →

That’s it for this week’s Lore Brief.

See you next week!