What This Report Is
Andreessen Horowitz's State of Generative Media 2026 is the most comprehensive annual survey of how AI-generated images, video, audio, and 3D assets are actually being used in production. Not demos. Not vibes. Production deployments across gaming, advertising, e-commerce, and creative agencies.
This is the report that separates what's working from what's hype.
The Headline Numbers
14 models per enterprise: The average production deployment uses 14 different models for image and video generation. This is radically different from the LLM landscape, where OpenAI, Gemini, and Anthropic control 89% of enterprise wallet share. In generative media, fragmentation is the norm because no single model handles every use case well enough.
Multi-model workflows are standard: The typical production pipeline chains models together — generate an image, remove the background, upscale it, recolor it, apply a style-consistent LoRA. One-shot prompting delivers demo-quality output. Production-quality requires orchestration.
58% prioritize cost: More than half of organizations identified cost optimization as their primary criterion when selecting model infrastructure. Not quality. Not speed. Cost. This tells you where the market actually is versus where the marketing says it is.
Industry-Specific Adoption
Gaming: Studios use generative models to prototype concept art, populate environments, and produce in-game assets at speeds traditional pipelines can't match. The creative directors still make the decisions — the AI handles the iteration volume.
Advertising: Campaigns that took weeks of production now generate hundreds of personalized variations in hours. The shift isn't from human to AI — it's from one campaign to hundreds of micro-campaigns.
E-commerce: Product photography — which required teams of photographers, weeks of shoots, and long editing cycles — now compresses into prompts and a library of production-ready assets. This is where generative media has the clearest ROI.
The Video Frontier
Video generation is where the report gets most interesting. Seedance 1.0 topped generation leaderboards in mid-2025, and previews of Seedance 2.0 show coherence and control improvements that suggest video generation is approaching the inflection point that image generation hit in 2023.
The implication: if your business involves video production — marketing, training, content — the economics are about to shift dramatically.
Why This Matters for Tool Selection
The report's core insight for practitioners: stop looking for one model to rule them all. The winning strategy is orchestration — picking the right model for each step in the pipeline and building workflows that chain them together. The companies doing this well aren't the ones with the best AI — they're the ones with the best plumbing.
Read the full report: [a16z.com/the-state-of-generative-media-2026](https://a16z.com/the-state-of-generative-media-2026/)