The MAD Podcast with Matt Turck

Sachin Katti, OpenAI's Head of Industrial Compute, discusses the massive infrastructure buildout required to power AI systems, covering data center design, power generation challenges, custom chip development (Jalapeno), and the strategic diversification of compute sources across hyperscalers and partnerships.

The Machine Payments Protocol (MPP) is an elegant, standardized system that enables direct agent-to-service transactions without human intervention. Agents request access to services, receive payment requests, pay automatically, and complete transactions through machine-readable protocols without account creation or checkout interfaces.

AI agents consume tokens at extremely high rates, requiring a new payment model. Metronome and Tempo have partnered to enable streaming payments that charge for tokens as they're consumed in real-time, solving cash flow problems for AI companies dealing with agent buyers.

The podcast discusses how solopreneurs are driving incremental business growth in America, with 5 million people running solo companies. AI technologies, particularly domain-specific agents, are enabling these individuals to both build and operate their businesses more effectively.

AI agents can now generate complete working applications in 20 minutes through vibe coding, but deployment has become the new bottleneck. Stripe Projects aims to solve this by enabling agents to configure and integrate necessary services for deployment directly from the command line.

The speaker argues that microtransactions were previously impractical because users were unwilling to enter payment information for small amounts. However, AI agents executing transactions autonomously while drawing from stablecoin balances could make microtransactions economically viable without requiring manual user effort.

The podcast discusses how agentic commerce has evolved from theoretical concept to deployed infrastructure within a year, with real companies now building on it. The conversation explores a spectrum of agentic commerce implementations, ranging from fully autonomous agent purchasing decisions to AI-assisted shopping experiences with integrated buy buttons.

Emily Sands from Stripe discusses the rapid maturation of agentic commerce infrastructure, including the Agentic Commerce Protocol, Link wallet for agents, and shared payment tokens. She highlights token theft as an underappreciated fraud risk threatening AI company economics, and projects that agents will evolve from simple transaction executors to multifaceted economic actors running entire businesses within 12 months.

State space models achieve better intuitive understanding of sequences by compressing entire sequences into a constant-sized cache or scratchpad at each step, rather than allowing random access to full sequences like transformers. This architectural constraint paradoxically makes them smarter at tasks requiring global sequence understanding.

The speaker discusses how resource constraints in AI development necessitate more efficient computational approaches, particularly highlighting 4-bit quantization as a solution that dramatically reduces memory usage and energy consumption across various computing operations.

Moore's Law, originally an economic principle about doubling transistor density affordably every 24 months, has ceased delivering practical benefits for shrinking existing designs. In the post-Moore's Law era, success requires clever system optimization and accelerated computing rather than relying on transistor scaling.

Every company's competitive advantage relies on proprietary secrets tied to their data and platform. Open source AI technologies enable enterprises to customize and implement AI solutions that leverage their unique, valuable data more effectively than off-the-shelf alternatives.

The speaker argues that rapid progress in transformational AI technology will inevitably occur due to significant community investment, and that control of AI development cannot be concentrated in the hands of a small group because innovation is distributed across many labs worldwide with diverse ideas.

The speaker argues that while closed internets like AOL and Prodigy existed historically, the open internet has proven to be transformational for businesses. They contend that AI, as a similarly transformational technology requiring diverse applications, should likewise be developed as open technology rather than closed systems.

Bryan Catanzaro, who leads NVIDIA's NeMoTron frontier AI models, discusses how open-source AI is accelerating through community collaboration, explains the technical innovations in NeMoTron 3 (hybrid SSM-transformer architecture, mixture of experts, multi-token prediction), and argues that open technologies are safer and more aligned with how effective organizations actually work.

Matthew Prince, CEO of Cloudflare, discusses how bot traffic has surpassed human traffic on the internet as of mid-2026, driven by AI agents and LLMs. He explores how this fundamental shift is forcing a reimagining of internet infrastructure, business models, and organizational structures, with Cloudflare positioned at the center of these changes through products like Workers, AI Gateway, and edge computing solutions.

The podcast discusses how GPU depreciation costs are the largest component of cloud computing expenses, and that GPU utilization directly impacts per-hour costs. Cloud companies gain competitive advantage by building beloved products that drive high GPU utilization rates.

The transcript explains the energy-to-computation pipeline for AI systems, tracing how raw energy sources (photons or natural gas) are converted through power plants into electrical power, then processed by servers into floating-point operations, and finally transformed into AI tokens per second.