Oracle、NVIDIA幹部がAIインフラ会議でキーノート

The Connected World LIVE! 2026 (TCWLIVE! 2026), held in partnership with InterGlobix, has announced its opening executive program for September 9, 2026. Oracle's Dr. Sanjay Basu, Senior Director of GPU & Gen AI Solutions & Services – Cloud Engineering, will deliver the opening keynote, followed immediately by a fireside chat with NVIDIA's Pete Coticchia, Global Head of Business Development & Go-to-Market (GTM) for Oracle Cloud.

Dr. Basu's keynote, "Designing AI Factories for Inference: Token Economics, Data Center Design, and Next-Generation Connectivity," will address the industry's pivot from AI model training to inference. The talk will center on token economics—defined as the cost, speed, and efficiency of generating AI tokens—which Basu will argue is becoming the key metric for measuring the success and viability of next-generation digital infrastructure. The keynote will also explore how AI factories at scale demand fundamentally new approaches to connectivity, networking, and data center architecture to handle unprecedented workloads.

Immediately following the keynote, the exclusive executive fireside chat, "The Evolution of NVIDIA AI Platforms and the Future of AI Infrastructure," will be moderated by Vinay Nagpal, CEO of IG Group. Basu and Coticchia will examine how successive generations of NVIDIA's AI platforms—Hopper, Blackwell, Vera Rubin, Rubin Ultra, and Feynman—have transformed and continue to reshape data center design, networking infrastructure, connectivity, power consumption, and cooling requirements. The conversation will address the growing challenge of scaling AI infrastructure, particularly the critical role of high-bandwidth, low-latency, and energy-efficient interconnects across GPUs, clusters, campuses, and cloud regions in enabling both performant and economically efficient AI inference.

The Connected World LIVE! 2026 event positions itself at the intersection of two industry shifts: the move from AI model training to inference at scale, and the infrastructure challenges that shift creates. By pairing Oracle's Dr. Sanjay Basu—whose expertise spans GPU infrastructure and cloud engineering—with NVIDIA's Pete Coticchia from the business development side, the conference signals that the conversation is no longer primarily about raw model capability but about the practical, economic, and architectural constraints of running inference at global scale.

The emphasis on token economics as a defining metric reflects a maturation in how the industry measures AI deployment success. Rather than focusing solely on model size or training performance, the keynote and fireside chat will address cost-per-token, latency, and energy efficiency—metrics that directly affect operating margins and deployability for enterprises. The progression through NVIDIA's hardware roadmap (Hopper, Blackwell, Vera Rubin, Rubin Ultra, and Feynman) suggests that interconnect bandwidth, power management, and cluster-level scalability are the technical frontiers where infrastructure vendors are competing to support the next wave of AI applications.