Uploaded April 2026 | Updated November 2026, 16 hours ago
How do you bring generative AI into games without turning your studio into an ML research team?
In this GDC session, Qualcomm introduces the Snapdragon Game AI SDK — a developer‑focused toolkit for building on‑device AI experiences like AI NPCs, teammates, coaches, and companions, all optimised for low latency and power efficiency.
You’ll see:
- How on‑device AI enables lower latency and longer play sessions
- AI NPCs, teammates, coaches, and companions built on a shared SDK
- Running LLMs, ASR, and TTS locally using CPU, GPU, and Hexagon NPU
- Engine‑friendly workflows (Unreal plugins, drag‑and‑drop + APIs)
- Hybrid edge + cloud strategies for scalable game AI
This talk focuses on practical game development realities — power, latency, memory, and tool usability — while showing how generative AI can enhance gameplay without replacing designers or writers.
Speaker:
Andy Michel, PhD (Qualcomm Technologies)
Key sections:
00:00 Why game AI is moving on‑device
03:32 Generative AI trends in games (agents, multimodality)
08:28 AI NPCs, teammates, coaches, and companions
12:37 Edge vs cloud: latency, cost, and design trade‑offs
18:02 Snapdragon hardware roles (CPU, GPU, NPU)
23:49 Hybrid AI architecture for games
26:07 Snapdragon Game AI SDK overview
31:29 Engine integration and developer workflows
34:47 Voice AI demo: noise suppression & intent detection
41:43 LLM pipelines, RAG, memory, and model sizing
46:29 Live demo: AI NPC interaction
48:56 Key takeaways for shipping AI‑powered games
Subscribe to get news and updates on Snapdragon's game developer solutions: qualcomm.com/support/contact/forms/snapdragon-gaming
How do you bring generative AI into games without turning your studio into an ML research team?
In this GDC session, Qualcomm introduces the Snapdragon Game AI SDK — a developer‑focused toolkit for building on‑device AI experiences like AI NPCs, teammates, coaches, and companions, all optimised for low latency and power efficiency.
You’ll see:
- How on‑device AI enables lower latency and longer play sessions
- AI NPCs, teammates, coaches, and companions built on a shared SDK
- Running LLMs, ASR, and TTS locally using CPU, GPU, and Hexagon NPU
- Engine‑friendly workflows (Unreal plugins, drag‑and‑drop + APIs)
- Hybrid edge + cloud strategies for scalable game AI
This talk focuses on practical game development realities — power, latency, memory, and tool usability — while showing how generative AI can enhance gameplay without replacing designers or writers.
Speaker:
Andy Michel, PhD (Qualcomm Technologies)
Key sections:
00:00 Why game AI is moving on‑device
03:32 Generative AI trends in games (agents, multimodality)
08:28 AI NPCs, teammates, coaches, and companions
12:37 Edge vs cloud: latency, cost, and design trade‑offs
18:02 Snapdragon hardware roles (CPU, GPU, NPU)
23:49 Hybrid AI architecture for games
26:07 Snapdragon Game AI SDK overview
31:29 Engine integration and developer workflows
34:47 Voice AI demo: noise suppression & intent detection
41:43 LLM pipelines, RAG, memory, and model sizing
46:29 Live demo: AI NPC interaction
48:56 Key takeaways for shipping AI‑powered games
Subscribe to get news and updates on Snapdragon's game developer solutions: qualcomm.com/support/contact/forms/snapdragon-gaming










