{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"TechSurge: Deep Tech Podcast","title":"From Chips to Agents: Rebuilding the AI Stack","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/f61c038e\"></iframe>","width":"100%","height":180,"duration":3661,"description":"\nBy May, Google alone was processing more than 3.2 quadrillion tokens a month, seven times what it processed a year earlier. Keeping pace with that demand means rebuilding the entire technology stack, from compute and networking to power and security. Few companies sit closer to every layer of that rebuild than Cisco.\nIn this episode of TechSurge, host Sriram Viswanathan speaks with Jeetu Patel, President and Chief Product Officer of Cisco, who runs the organization behind the company’s silicon, optics, networking, security, and Splunk businesses. Last fiscal year, Cisco booked more than $9 billion in AI infrastructure orders from the largest cloud providers, part of a buildout Jeetu compares to the dot-com boom of the early 2000s, when Cisco briefly became the world’s most valuable company.\nPatel argues that trust, not raw compute, is now the gating factor on agentic AI adoption: agent adoption is actually slower than chatbot adoption was, largely because people don’t yet trust giving agents access to their email, calendar, or credit card. He walks through Cisco’s transformation from a 250-acquisition “holding company” into a single integrated platform spanning silicon (Silicon One), photonics (Acacia), systems, and software, and explains why token efficiency has become a security discipline in its own right, an adversary that exhausts an agent’s token budget can leave it defenseless, “like running out of ammunition in the middle of a war.”\nThe conversation also covers a real incident in which an AI agent tried to escape its sandbox during a Hugging Face security test, Cisco’s goal to have the vast majority of its own code written by AI within two years and who is accountable when that code breaks, and Jeetu’s own leadership philosophy on hunger, humility, and publicly owning mistakes. The episode closes on Cisco’s silicon roadmap for scale-up, scale-out, and scale-across data center architectures, and Andy Grove’s observation that while the laws of physics are...","thumbnail_url":"https://img.transistorcdn.com/jQPF4l9NFf0J8GE3ySmQUhKsdFs-I1vwVANYFaBaoL0/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9kOGI1/OGFhMjdjOWMzMDhj/MGY4MGFiMDMyMmIx/Y2M4ZS5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}