Four CEOs on the Future of AI: CoreWeave, Perplexity, Mistral, and IREN

Overview

This episode features conversations with four AI industry leaders—Michael Intrator (CoreWeave), Aravind Srinivas (Perplexity), Arthur Mensch (Mistral AI), and Daniel Roberts (IREN)—live from NVIDIA's GTC conference. The discussions explore the evolution of AI infrastructure, specialized models, the orchestration of AI tools, and the future of data centers and energy.

Notable Quotes

- The decommoditization of compute is about building supercomputers that can change the world. – Michael Intrator, on scaling AI infrastructure.

- AI is the operating system now. You start with objectives, not instructions. – Aravind Srinivas, on the future of AI orchestration.

- Open source models save costs, provide control, and allow enterprises to leverage decades of intellectual property. – Arthur Mensch, on the advantages of open AI models.

🚀 AI Infrastructure and Compute Scaling

- Michael Intrator shared CoreWeave's journey from crypto mining to hyperscale AI infrastructure, emphasizing their early adoption of GPUs and risk management expertise.

- CoreWeave transitioned from crypto to CGI rendering, medical research, and neural network training, eventually focusing on building scalable supercomputers for AI.

- Intrator highlighted the importance of long-term contracts and dismissed concerns about GPU obsolescence, noting their six-year depreciation model and continued profitability.

🧠 Multi-Model AI Orchestration

- Aravind Srinivas discussed Perplexity's focus on accuracy and orchestration, enabling users to leverage multiple AI models for specialized tasks.

- Perplexity's Computer product integrates AI tools, browsers, and local hardware to create a seamless user experience for both consumers and enterprises.

- Srinivas emphasized the importance of trust, privacy, and hybrid solutions, including local and server-side orchestration for sensitive data.

🌍 Open Source AI and Enterprise Customization

- Arthur Mensch explained Mistral AI's commitment to open source models, which allow enterprises to customize AI for specific verticals like manufacturing, finance, and healthcare.

- Mistral's Forge product enables enterprises to train models on proprietary data while maintaining strict data segregation and security.

- Mensch highlighted the importance of human expertise in refining models and the need for context engines to manage sensitive enterprise data.

⚡ Energy and Data Center Innovation

- Daniel Roberts detailed IREN's strategy of building massive data centers powered by 100% renewable energy, leveraging wind and solar in West Texas.

- Roberts described the challenges of scaling infrastructure, including workforce shortages and supply chain constraints, while emphasizing community investment and local hiring.

- He addressed the growing importance of nuclear energy and modular reactors as potential future solutions for powering AI data centers.

📈 The Future of AI Demand and Accessibility

- All four CEOs agreed on the relentless growth of AI demand, driven by advancements in software efficiency and the proliferation of use cases.

- Roberts noted that increased compute availability will further accelerate AI adoption, likening it to the induced demand seen in other industries.

- Discussions touched on the democratization of AI tools, enabling individuals and small businesses to leverage AI for automation and innovation.