When AI needs to operate inside games, simulations, enterprise tools, robotics, or secure environments, cloud-based tools create latency, recurring API costs, privacy exposure, weak offline access, and limited control. Pontex solves all of it at the runtime layer.
Does not make chips. Licenses the architecture layer beneath chips. SoftBank paid $32B on $1.5B revenue. Now valued at $440B. The thesis was architectural necessity.
Does not make AI models. Licenses the runtime layer above local inference. Every game, simulation, enterprise tool, and robotics system running local AI will need a runtime. That runtime is Pontex.
Live inside TCG World, a Unity HDRP open-world game, since February 2026. Full agent loop: voice in, reasoning, memory, tool execution, environment change, voice out.
~0.3 second voice latency. ~4 second tool-call latency. Validated across 6 LLMs and 4 hardware backends in a real-time 3D environment.
Three provisional filings covering local inference/memory bridge, real-time scheduling, and multi-modal orchestration with safe execution. PCT filing in progress.
Active technical discussions with Microsoft teams including Foundry Local and Windows Runtime. Strategic meeting with NVIDIA completed. Apple contacts engaged.
Reference implementation demoed at TwitchCon Rotterdam, May 2026. Approximately 300 qualified leads generated in two days with no paid marketing.
A live demo of the full Pontex runtime running inside TCG World is available for qualified investors and strategic partners under NDA.
David is CEO and co-founder of Pontex AI and CEO of TCG World, where he led the development of a large-scale Unity-based gaming platform. He is leading Pontex's transition from internal production technology into a standalone AI infrastructure company focused on local agent execution, enterprise licensing, and strategic platform partnerships.
LinkedIn →Justin is co-founder of Pontex AI and EVP of TCG World. He leads fundraising, investor outreach, commercial strategy, IP coordination, and patent strategy. Justin has shaped Pontex's positioning as a standalone AI infrastructure platform focused on turning the production-proven local AI runtime into a licensable product.
LinkedIn →Richard designed and built the core Pontex local AI runtime architecture, including local inference integration, memory handling, structured output, safe tool execution, host-system integration, and real-time execution inside Unity-based environments. He leads the technical roadmap as Pontex moves from reference implementation to standalone SDK platform.
Submit your details and investment range below. Our team will follow up within 48 hours to arrange a call and provide access to the full investment materials.
By submitting you agree to receive follow-up communications from Pontex regarding this offering.