Industry View · Gaming

A $200bn industry rebuilds its content pipeline before it learns to charge for it

Gaming is the first creative industry where generative AI has reached near-universal adoption inside the studio: roughly 90% of developers now use it somewhere in their workflow, yet the standalone generative-AI-in-gaming market is still only $1.8bn against $201.6bn in total games spending. The disruption is real but lopsided — AI is compressing the cost of making games far faster than it is creating new ways to sell them.

Emerging ROIROI Classification

Asset costs cut 30-40% but Take-Two sees no observed savings; revenue not yet demonstrated

Key Figures

$201.6B
Global games market, 2025 (first time past $200B)
Outrun Gaming
$1.81B
GenAI-in-gaming market, 2025
TBRC
90%
Developers already using AI in workflows
Google Cloud
34,000+
Games-industry layoffs since 2021
GDC / Game Developer

Value Chain

Ideation
Brainstorming goes first

The most widespread use: a large majority of surveyed developers use generative AI for research and brainstorming, the lowest-risk entry point.

ChatGPT, Claude, in-house LLMs
Asset creation
The cost epicenter

Text-to-3D, texture and concept-art generation are cutting early-stage asset costs 30-40%, the single largest disruption vector.

Roblox Cube, Scenario, Tencent VISVISE
NPCs & dialogue
Characters that improvise

LLM-driven NPCs and co-playable companions shift dialogue and behavior from scripted to generated, live in shipping titles.

Nvidia ACE, Inworld, NetEase
Animation & QA
Quietly automating the grind

Procedural animation and AI-assisted playtesting compress two of the most labor-intensive, least-glamorous stages.

Tencent, Unity, in-house tools
Whole-game gen
The frontier bet

Real-time 'world models' generate playable environments frame-by-frame from a prompt; impressive demos, not yet products.

Decart Oasis, Google Genie 3, MS Muse

01 · The thesis

The pipeline is being automated faster than the business model is being reinvented

The center of gravity is asset and content production — concept art, textures, 3D meshes, animation, NPC dialogue — where AI tools are cutting early-stage costs by 30-40% at studios that have integrated them, and where Morgan Stanley estimates AI could nearly halve industry-wide development cost and unlock roughly $22bn in annual studio profit over time. EA's CEO has told investors AI could impact 60% of all development processes; Krafton has declared itself an 'AI-first company' and committed to a ~$70m GPU cluster. This is a genuine, durable productivity shift, not a fad. The harder question is revenue. Take-Two's CEO has publicly said the company has not yet observed AI-driven cost savings and 'may never' — a useful corrective to the hype. The defensible value is concentrating in two places: distribution platforms (Roblox, Microsoft, Tencent) that own the creator economy, and world-model labs (Decart, Google DeepMind) chasing real-time generative gameplay. Tool vendors in the middle face commoditization, and the human cost is already visible in 34,000+ layoffs and a year-long voice-actor strike fought explicitly over AI.

Brainstorming goes first

The most widespread use: a large majority of surveyed developers use generative AI for research and brainstorming, the lowest-risk entry point.

The cost epicenter

Text-to-3D, texture and concept-art generation are cutting early-stage asset costs 30-40%, the single largest disruption vector.

Characters that improvise

LLM-driven NPCs and co-playable companions shift dialogue and behavior from scripted to generated, live in shipping titles.

Quietly automating the grind

Procedural animation and AI-assisted playtesting compress two of the most labor-intensive, least-glamorous stages.

The frontier bet

Real-time 'world models' generate playable environments frame-by-frame from a prompt; impressive demos, not yet products.

02 · The two clocks

Three figures that frame how fast — and how unevenly — AI money is moving through gaming.

The total games market crossed $201.6bn in 2025, the first time past $200bn, split roughly 56% mobile (~$113.3bn) and 44% PC/console (>$88bn). Against that, the dedicated generative-AI-in-gaming market is tiny — $1.81bn in 2025 — meaning AI is a cost-side and tooling story far more than a consumer-revenue story so far. Capital is flowing fastest into world-model labs. Decart, whose Oasis model renders a playable game world in real time, raised $300m at a $4bn valuation (announced 2026, led by Radical Ventures with Nvidia participating) — one of the steepest re-ratings in the sector. The human ledger runs the other way: more than 34,000 games-industry workers have been laid off since 2021, with recent-year cuts among the heaviest on record, even as studios cite AI-driven 'efficiency.'

03 · Public players & exposure

Who routes through, who gets routed around

We plot the listed players on two editorial axes — how exposed each is to AI disruption, against how ready its data, brand and position are to be the answer. The figures in the table are sourced; the placement is our read.

04 · Private flagships

The AI-native challengers

The companies attacking this industry AI-first, with disclosed funding where available:

Decart

Its Oasis model generates a playable, Minecraft-like world frame-by-frame in real time from user input — the clearest demo of AI-generated gameplay rather than AI-assisted production.

Inworld AI

Provides a platform for contextually aware, personality-driven non-player characters and dialogue, integrated into games and immersive experiences.

Scenario

Generates game art assets that match a studio's own art style via custom-trained workflows, targeting the asset-production bottleneck directly.

Google DeepMind (Genie 3)

Generates navigable, interactive 720p environments at 24fps from a text prompt, holding consistency for minutes — research frontier rolling toward consumer access via Project Genie.

Latitude

Creator of AI Dungeon, an early proof that fully generative, open-ended text gameplay can sustain a real player base and business.

05 · Signals

What moved, and what to watch

06 · The exposure read

Who’s defensible, who’s at risk

AI rewards clean, structured advantage and punishes friction. The line runs through who owns the data, the brand and the customer — and who is merely a step the technology can route around.

Sources

Where this comes from

The spend, and the payoff

GenAI-in-gaming market: small base, fast climb

Source: The Business Research Company (TBRC) / OpenPR, 2025 — generative-AI-in-gaming market size and forecast. (USD billions)

Who's defensible, who's at risk

Defensible vs At Risk

Defensible

  • Platform and distribution owners — Roblox, Microsoft and Tencent monetize a creator economy, so cheaper content creation expands their supply without threatening their take rate.
  • Infrastructure vendors — Nvidia sells the GPUs and the ACE runtime regardless of which studio or model wins, capturing value across the whole stack.
  • World-model labs with capital — Decart and Google DeepMind are building the one capability nobody else has, attracting the steepest valuations in the sector.
  • Lean and mid-size studios — the teams that can cut asset costs 30-40% gain the most relative leverage against AAA budgets that now start near $100m.

At Risk

  • Performers and production artists — the SAG-AFTRA strike and 34,000+ layoffs show voice, concept-art and animation roles are the first absorbed by generative tooling.
  • Mid-stack tool vendors — companies selling generic asset or NPC generation risk commoditization as engines (Unity) and platforms (Roblox) fold the same features in for free.
  • AAA publishers banking on cost cuts — Take-Two's own CEO warns the promised savings may not materialize, leaving budgets above $1bn without an AI offset.
  • Studios chasing 'AI slop' — output that floods storefronts with low-quality generated content risks player backlash and platform de-ranking.

The signals — how it unfolded

Mar 2025

Roblox opens its Cube model

Roblox releases its Cube 3D/4D generative foundation model open-source on GitHub and HuggingFace, betting on platform lock-in over IP secrecy.

Jul 2025

Voice-actor strike ends with AI terms

SAG-AFTRA ratifies a deal requiring consent and comparable pay for AI digital replicas, setting the first labor template for generative AI in games.

Aug 2025

Genie 3 raises the world-model bar

Google DeepMind unveils a general world model generating interactive environments in real time, intensifying the race with Decart and Microsoft Muse.

Q3 2025

Krafton goes 'AI-first'

Krafton formalizes an AI-first strategy with a ~$70m GPU cluster and an SK Telecom consortium 500B-parameter model — a major publisher reorganizing around AI.

Oct 2025

EA partners with Stability AI

EA signs Stability AI to co-develop models and tooling, the clearest sign a Western AAA publisher is industrializing generative content.

Challengers to watch

Decart

Its Oasis model generates a playable, Minecraft-like world frame-by-frame in real time from user input — the clearest demo of AI-generated gameplay rather than AI-assisted production.

Inworld AI

Provides a platform for contextually aware, personality-driven non-player characters and dialogue, integrated into games and immersive experiences.

Scenario

Generates game art assets that match a studio's own art style via custom-trained workflows, targeting the asset-production bottleneck directly.

Google DeepMind (Genie 3)

Generates navigable, interactive 720p environments at 24fps from a text prompt, holding consistency for minutes — research frontier rolling toward consumer access via Project Genie.

Latitude

Creator of AI Dungeon, an early proof that fully generative, open-ended text gameplay can sustain a real player base and business.

Exposure table

CompanyStanceThe sourced fact
Roblox RBLXPlatform ownerIts open-sourced Cube 3D/4D foundation model had generated over 1.8M assets since launching in March 2025, with creators using AI tools showing a 31% lift in publishing output.
Microsoft MSFTInfra + modelMicrosoft Research's Muse (WHAM), trained on Ninja Theory's Bleeding Edge, is a generative gameplay world-and-action model for prototyping; Microsoft also owns Azure gaming AI and Xbox studios.
Nvidia NVDAPicks & shovelsIts ACE platform powers shipping co-playable characters — Krafton's PUBG Ally and inZOI Smart Zoi — running small models locally on player GPUs.
Electronic Arts EAEmbracing hardCEO Andrew Wilson told investors up to 60% of EA's development processes could be impacted by generative AI; EA partnered with Stability AI in October 2025 to co-develop models and tools.
Tencent TCEHYTooling + scaleTencent's VISVISE AI suite claims to compress game-art production time dramatically; the firm is pivoting capital from studio investments toward AI and user-created games.
NetEase NTESPragmatic adopterNetEase has reported AI-driven pipelines materially boosting production efficiency, deployed across dynamic NPCs (Justice Mobile), asset creation and procedural animation.
Krafton 259960.KSAI-first pivotDeclared itself an 'AI-first company' in late 2025, committed ~$70m to a GPU cluster and joined an SK Telecom consortium developing a 500-billion-parameter foundation model.
Take-Two TTWOShow-me skepticCEO Strauss Zelnick said the company has not yet observed AI-driven cost savings and 'may never,' even as GTA 6's budget is widely estimated north of $1bn.
Unity USqueezed middlePushes AI via Sentis (on-device model inference) and engine tooling, but sits between platform giants and free model vendors after deep 2024-25 layoffs and a strategy reset.
Voice / art labor SAG-AFTRADisplacement riskA year-long video-game performers' strike fought primarily over AI replicas was suspended in June 2025 and ratified in July, winning consent-and-pay terms for digital replicas.

Sources