Google Cloud's 'living games' pitch, read through its six named partners

What Google is actually selling under 'living games'

The headline concept is games that 'adapt, grow and evolve themselves' — worlds that learn a player's style and generate content mid-session. But the post is written by Jack Buser, Director for Games at Google Cloud, and the pitch is an infrastructure one, not a game-design one.

Google's offer is described plainly: the same services and infrastructure behind Search and YouTube, plus Gemini models and Vertex AI, plus a partner ecosystem of engines and tools. The 'living games' language is the marketing wrapper; the product is Vertex AI, Gemini, and Google Kubernetes Engine sold into studios.

The six deployments do different jobs

The examples span distinct stages rather than one shared use case. Capcom uses Vertex AI and Gemini to generate 'hundreds of thousands of ideas' for items and environments — an ideation and iteration tool aimed at lowering cost and shortening the path to a new game.

Klang Games is the closest to the 'living' promise: its MMO SEED is meant to run 'hundreds of thousands of autonomous virtual humans' called Seedlings that form relationships and emergent societies through persistent interaction, built on GKE, Vertex AI and Gemini. This is the runtime, in-game use of AI, not a production-side tool.

The remaining four are pipeline services: ElevenLabs for character voices, translation and sound effects; nunu.AI building a multi-modal Gemini agent for continuous QA testing; Common Sense Machines generating game-ready 3D assets from text and images with Cosmic Lounge; and Series Entertainment, a startup that claims cutting development time by 90%.

Where to be cautious with the claims

Series Entertainment's 90% development-time reduction is the only hard number in the post, and it is presented without a baseline, methodology, or comparison — it is a vendor-supplied figure, not a measured benchmark. Klang's Seedlings are described in the future tense ('will enable'), meaning the flagship 'living game' example is a stated capability, not a shipped feature at the time of writing.

The post also repeats a familiar productivity argument — that gen AI frees studio staff to focus on the 'most important and interesting' creative work as development costs rise. That framing sidesteps the harder question of what happens to the roles doing the ideation, asset creation, voice work, and QA that these six tools now automate.

The implication: runtime generation is the real bet, and it's the least proven

Five of the six examples make existing production faster or cheaper — ideation, art, audio, testing, 3D assets. Those are established, measurable wins that studios can adopt today. The single genuinely novel claim, games that generate and adapt content while players are in them, rests almost entirely on Klang's not-yet-realized Seedlings.

For teams evaluating this, the practical takeaway is to separate the two. Production-side gen AI on Vertex AI and Gemini is a here-and-now efficiency play with named references. 'Living games' as a live, in-session experience remains a forward-looking bet — the part of the announcement with the most ambition and the least evidence.