GTC DC '25 Pregame - Chapter 5: AI for Robotics and Manufacturing

The episode opens with Foxconn’s shift from labor‑intensive lines to what CEO Yong Liu calls "Industrial 5.0," a manufacturing model that blends generative AI, robotics, and extensive sensor networks. By pairing physical factories with digital twins, Siemens CTO Peter Kurta explains how U.S. plants can be simulated, optimized, and stress‑tested before a single bolt is tightened, delivering unprecedented speed, productivity, and energy efficiency. This digital‑first approach is positioned as a national security asset, promising to restore competitive, AI‑native factories across the United States.

Figure AI’s founder Brett Adcock dives into the technical heart of humanoid robotics, describing a 40‑joint platform whose state space exceeds the number of atoms in the universe. Solving such complexity demands end‑to‑end deep‑learning pipelines running on NVIDIA GPUs, both for massive pre‑training data collection and on‑device inference without network connectivity. Figure’s recent commercial deployment demonstrates a ten‑hour autonomous shift, while the company simultaneously builds a high‑scale manufacturing hub in California to refine testing, integration, and volume production. The discussion underscores that scaling general‑purpose bipedal robots hinges more on mastering a horizontal AI stack than on traditional cost or supply‑chain concerns.

Aki Jain of Palantir frames the broader ecosystem as an ontology‑driven integration layer that unifies ERPs, MRPs, and edge AI models. By translating disparate data into a common semantic fabric, Palantir enables secure, privacy‑aware orchestration of both open‑source and proprietary models—ranging from large language models to specialized machine‑programming tools—across commercial and government domains. This capability is critical for the reindustrialization of America, where talent pipelines, compute infrastructure, and AI‑augmented decision making must converge. Collectively, the panel argues that the next wave of U.S. manufacturing competitiveness will be defined not by cheaper parts, but by the ability to embed adaptable, general‑purpose intelligence into every physical process.