AMD CEO Lisa Su Live on TBPN | Oliver Cameron, Mohammad Norouzi, Anjney Midha, Lisa Su

Overview

This episode dives into AMD's pivotal role in the AI revolution, featuring a live interview with AMD CEO Lisa Su. The discussion spans AMD's latest advancements, partnerships, and the broader implications of AI on hardware and software ecosystems. Guests like Oliver Cameron (Odyssey), Mohammad Norouzi (Ideogram), and Anjney Midha (AMP) share insights on AI's transformative potential across industries, from robotics to enterprise design.

Notable Quotes

- The world is a heterogeneous place. There is no one-size-fits-all in compute. – Lisa Su, on AMD's diverse product portfolio.

- Driverless cars today look like NLP systems in the 2010s—intelligent, but in isolated ways. – Oliver Cameron, on the future of world models in autonomy.

- Only 15% of tenant data center capacity in the U.S. is being utilized. That 85% waste is a national security crisis. – Anjney Midha, on optimizing AI infrastructure.

🖥️ AMD's Role in the AI Boom

- Lisa Su highlighted AMD's unveiling of Helios, a rack-scale AI system powered by the MI450 accelerator, and its partnership with Anthropic to deploy up to 6 gigawatts of AMD Instinct MI450s.

- AMD's focus on open ecosystems and collaboration was emphasized, with Su noting that partnership is not just a word and citing collaborations with Meta, OpenAI, and Anthropic.

- The company is investing heavily in heterogeneous computing, leveraging CPUs, GPUs, and FPGAs to meet diverse workload demands.

🚗 World Models and the Future of Autonomy

- Oliver Cameron discussed Odyssey's work on foundational world models, which aim to replace hand-tuned systems in robotics and driverless cars.

- He argued that world models, trained on diverse data sources, can adapt to both virtual and physical environments, enabling applications in gaming, education, and autonomy.

- Cameron also highlighted the inefficiency of current driverless systems, likening them to early NLP models, and predicted a shift toward unified, physics-aware world models.

🎨 Generative AI in Design and Enterprise

- Mohammad Norouzi of Ideogram explored how generative AI is transforming enterprise design, from product photography to creative campaigns.

- He emphasized the importance of consistency and data sovereignty, noting that Ideogram's compact, open-weight models allow enterprises to host AI on-premises for cost and IP protection.

- Norouzi also discussed AI's potential in generating synthetic data for training models in manufacturing and defense.

⚡ Optimizing AI Infrastructure

- Anjney Midha from AMP highlighted inefficiencies in U.S. data centers, where only 15% of capacity is utilized. He called for output maxing to optimize compute resources.

- Midha stressed the need for smarter infrastructure investments, including nuclear energy, to meet the growing demand for AI compute.

- AMP is building 2 gigawatts of capacity in the U.S. and exploring partnerships to address long-term energy and compute constraints.

🌍 AI's Broader Impacts and Future Directions

- The panelists discussed AI's role as a force multiplier across industries, from accelerating hardware development to enabling new scientific discoveries.

- Lisa Su underscored the importance of local AI, with AMD investing in AI PCs and robotics form factors to democratize access.

- The conversation touched on the next wave of AI adoption, with reinforcement learning and formal verification emerging as key drivers of innovation.