Niels Bross

“Epic isn't trying to sell you a model.” Interesting statement from @UnrealEngine at #UEFESTTOKYO26. But I think we're ignoring a massive elephant in the room. WHO IS GOING TO PAY FOR ALL THIS AI? Epic is going all-in on AI-assisted development, while UE6 is moving toward Verse and Scene Graph, eventually removing traditional Actors, Blueprints, and the C++ Gameplay Framework. We're talking about replacing development systems that people have spent decades learning, extending, and building production pipelines around. And apparently AI-assisted migration is part of the answer. But who exactly is this designed for? AAA studios with enormous budgets can afford expensive AI subscriptions, frontier models, API usage, and dedicated AI infrastructure. Now imagine you're a small indie studio. You have 2–5 developers and limited resources. Your $20 AI subscription hits its usage limits during a serious development session. Suddenly you're looking at $200–$500 plans, multiple subscriptions, or expensive pay-as-you-go API usage. And please don't mention “just use cheaper models” argument. Anyone who has seriously worked with AI on complex Unreal Engine C++ codebases knows that the difference between budget models and frontier models is substantial. MCP supporting every model sounds fantastic on a presentation slide. But model freedom means very little when the models you actually need are financially out of reach. @TimSweeneyEpic, I think Epic needs to answer some questions: 1. Can indie developers realistically use UE6 without expensive AI subscriptions? 2. Who pays for AI-assisted migration of large existing projects? 3. Will conversion tools work without continuous third-party API spending? 4. What happens when AI quotas are exhausted halfway through development? 5. How does Epic intend to keep its ecosystem accessible to developers who cannot afford frontier models or high-end local AI hardware? @sama recently argued that it's “quite terrible” to have scarce DRAM sitting in personal computers, supposedly being used only 5% of the time, when that memory could be utilized more efficiently in cloud data centers. So the CEO of one of the world's largest AI companies is effectively arguing that scarce computing resources would be better centralized in the cloud. @nvidia : Rumors spread that largest consumer GPU dies may increasingly be prioritized for far more profitable professional hardware. RTX 5090 availability has become non existant. And RTX 5080 is going same line. And let's not forget @AnthropicAI. Dario Amodei keeps warning about the dangers of powerful open-weight models while advocating tighter controls over their release. Meanwhile, Anthropic's increasingly restrictive Claude usage policies and account enforcement show exactly how much power centralized AI providers have over their users. Developers are being pushed toward cloud AI services where the provider controls the pricing, usage limits, acceptable use, and ultimately whether you retain access at all. And local AI requires substantial RAM, VRAM, and compute (that is not affordable or doesn't exist) to deliver comparable results on serious coding tasks. So independent developers get squeezed from BOTH directions. @UnrealEngine may remain free to license, but Epic risks turning affordable game development into an increasingly expensive ecosystem of AI subscriptions, API tokens, and hardware requirements. Basically replacing one barrier with another, potentially much more expensive one. Until then, I'll be keeping my UE5.8 fork alive. #UnrealEngine #UE6 #IndieDev #GameDev #LocalAI