RENEGADE 2026: RNGD Mass Production, Performance Benchmarks, Enterprise Deployment, Cloud Partners

On April 2, 2026, FuriosaAI held its inaugural summit, RENEGADE 2026, to celebrate the mass production of RNGD, our high-performance AI accelerator. The keynote sessions feature Furiosa’s technical leadership alongside key industry partners discussing the transition from AI development to large-scale production. The presentations provide a technical and strategic overview of Furiosa’s Tensor Contraction Processor (TCP) architecture, the integrated software stack, and real-world deployment strategies to reduce Total Cost of Ownership (TCO) and energy consumption in modern data centers. Key Topics: [ 00:02:54 - Mission & Vision ] From the "AlphaGo shock" to building a native AI digital brain for the era of inference-centric data centers. [ 00:16:31 - RNGD Technical Specs ] Achieving 512 TFLOPS and 1 PetOPS of compute power within a 180W TDP limit. [ 00:23:28 - Performance Benchmarks] Comparative data showing RNGD outperforming legacy enterprise GPUs on the EXAONE 4.0 32B and Qwen3 32B models. [ 00:31:29 - LG AI Collaboration ] Partnering with LG AI Research and LG U+ for commercial-scale foundation model optimization. [ 00:39:01 - NPU-as-a-Service ] Samsung SDS introduces subscription-based RNGD cloud instances for public and enterprise sectors. [ 00:47:22 - Global Expansion ] Exporting the "AI Full Stack" and foundation model consortium to new markets. [ 00:57:58 - TCP Architecture & Compiler ] Hardware-software co-design focusing on massive on-chip data reuse and global search-based optimization. [ 01:36:03 - Deployment Readiness ] Transitioning to RNGD with "one line of code" via PyTorch, vLLM, and OpenAI API compatibility. To learn more, visit furiosa.ai.