How tech shifts and market trends are reshaping connected ecosystems
summary
Claim that LLMs and VLMs move from cloud to on-device inference with measurable market signals: 18.6 billion devices (2024), 21B (2025), 40B (2030) and chipmakers shipping higher-TOPS modules. Concrete hardware and software examples surface: Qualcomm Cloud AI100 PCI cards (≈275 TOPS) running LLM inference, Nvidia Jetson at ~100 TOPS, and vendor stacks plus acquisitions like Qualcomm → Edge Impulse. Surprising adoption numbers show momentum: edge AI modules +77% YoY, penetration 1.2%→2.3% and a VLM demo that correctly describes an event photo.
AI-written summary · about
chapters (8)
- 0:00Tech trends in connected ecosystems
- 1:12Edge computing comeback
- 3:56Edge computing layers and devices
- 6:47AI at the edge traction
- 9:57AI models for edge deployment
- 12:05Edge AI hardware and software
- 13:30Multimodal AI at the edge
- 14:57Key takeaways on edge AI
related talks
15mAI Trends, and Running LLMs in the Enterprise
38mAI’s Evolving Landscape: Opportunities Strategy Economics and Sustainable Advantage
58mThe Next Decade of Software Development
21mScaling LLM Inference with vLLM
18mThe Cutting Edge of AI and IoT
03mClosing Remarks
47mOwn Your LLM: Running Generative and Agentic AI Locally
11mKeynote: Building a Large Model Inference Platform for Heterogeneous Chinese Chips Based on VLLM