Keynote: State of the Union
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
Claims massive infrastructure spend — $3 trillion by 2028 on AI data centers and a single GPT‑5 training run costing ~$1 billion — and finds inference and power are primary constraints. Notes openweight models are roughly 90% as good and ~6× cheaper, yet proprietary APIs capture ~80% of usage, wasting ~$24.8 billion/year. Points to an open-source “park stack” (PyTorch, Ray, VLM, DeepSpeed, Kubernetes), agent protocols MCP/A2A, Anthropic multi‑agent gains ~90%, and new Linux Foundation projects like OpenInfra/OpenStack and Dockling.
AI-written summary · about
chapters (8)
- 0:00Linux Foundation AI focus
- 2:35AI bubble debate and investments
- 6:53Open source in AI infrastructure
- 9:05Open models commoditizing LLMs
- 14:49Rise of PAIR stack in AI
- 17:39Open source in agentic AI
- 23:03Enterprise AI adoption 2026
- 25:13Open source catalyst for AI
related talks
27mKeynote: Open Source in the Age of AI
45mKeynote: Welcome
16mKeynote: Open Source for AI and AI for Open Source
33mKeynote: Fine-tuning our way towards openness in AI
11mKeynote: Open Source Stack Empowers the AI Era
54mThe Future of AI: Building the Most Impactful Technology Together
52mKeynote: State of the Union
17mOpen Innovation: Unlocking AI's Potential