Cisco's AI Infrastructure Drive: Founder Mindset and Center Buildout

•

 43 min video

•

 8 min read

YouTube video ID: ofxAtQhV-xA

Source: YouTube video by Y Combinator — Watch original video

PDF

G2 Patel, Cisco's President and Chief Product Officer, discusses Cisco's transformation into an AI infrastructure company, the massive data center buildout, and the importance of a "founder's mindset" in a large organization. He also shares his personal journey and advice for aspiring entrepreneurs.

Cisco's Transformation into an AI Infrastructure Powerhouse

Patel describes Cisco as the "picks and shovels company during the gold rush" of AI. Four years ago, this description would have been unimaginable, highlighting the significant transformation the company has undergone. Cisco is now focused on providing the infrastructure for the AI era, ensuring that AI runs securely and safely at scale.

Patel emphasizes that this transformation was a team effort, with leaders setting the direction and hiring the right people. He views his role as ensuring the development of the best products for customers at scale.

The "Founder Mode" at Cisco

Patel's journey to Cisco involved running his own company, working at EMC, and then at Box. He identifies with a "founder mode" approach, where individuals act as owners of the business, comfortable with interdepartmental collaboration. He believes this mindset is crucial during times of market shift and rapid change.

He sees himself as a "late-stage co-founder" of Cisco, deeply invested in the company's success. This involves taking responsibility for all aspects of the business, from sales to marketing, and expecting others to do the same.

To balance the benefits of this founder mode with the need for seasoned operators, Cisco employs a "rule of thirds" for its leadership team: - One-third: Individuals who understand how to navigate Cisco at scale. - One-third: External hires from competitors or the market who are systems thinkers and have operated at scale. - One-third: Co-founders and CEOs of acquired companies, given charters larger than their original businesses.

Patel believes Cisco is currently in "wartime," having achieved a successful product turnaround and now squarely in the AI infrastructure game. The goal is to operate like "the world's largest startup," combining speed with scale to capitalize on the immense opportunities ahead.

Key Challenges and Opportunities in the AI Era

Patel identifies two critical problems Cisco aims to solve:

  1. Infrastructure Shortage: With less than 2% of humans currently using AI agents, the existing infrastructure is already supply-constrained. As AI agent usage grows to 25-50%, the demand for infrastructure will scale dramatically. Cisco aims to provide the necessary infrastructure to meet this demand.
  2. AI Safety and Security: The increasing sophistication of AI agents raises concerns about them "going rogue" or being influenced maliciously. Cisco is focused on building safety, security, and observability capabilities to ensure AI is safe for everyone. This includes:
    • Networking: Providing the underlying infrastructure.
    • Security: As one of the largest security companies, ensuring defenses at machine scale.
    • Observability: Leveraging Splunk to provide critical machine data insights.

This combination positions Cisco as critical infrastructure for the AI era's critical infrastructure companies.

AI Adoption Within Cisco: A Mandate for Innovation

Cisco has gone "all in" on AI, not just in external investments but also in integrating AI internally. Patel clarifies that the goal isn't efficiency, but rather to leverage AI for original insights and to solve problems previously unimaginable. He believes AI will lead to a need for more engineers, not fewer, as new human bottlenecks emerge with each AI advancement.

To drive AI adoption among its 32,000 engineers, Cisco took a bold approach: employees were told that not using AI would lead to job loss. This was framed not as a threat, but as a necessity in a secular shift, akin to the expectation that knowledge workers use PCs or the internet.

Cisco provided unlimited tokens for employees to experiment with AI, emphasizing familiarity and proficiency before efficiency. This top-down mandate, though initially met with resistance, led to significant breakthroughs, such as refactoring half a million lines of code into 120,000 in just two weeks. Cisco became a design partner with OpenAI and Codex, and Patel regularly meets with his engineering leadership to review AI progress.

AI's Impact Across Cisco's Full Stack

Cisco develops software, hardware, and silicon, each with different AI adoption cycles: - Software: AI's impact is rapid, with fast clock speeds. - Hardware: Design patterns are emerging, but adoption is slower than software. - Silicon: Adoption has been slower due to concerns about data and IP security. However, with improved models and open-weight models, this is expected to accelerate.

Cisco's full-stack approach involves co-designing across different planning cycles: - Silicon: 5-year planning cycle. - Hardware: 18-30 month planning cycle. - Software/Operating Systems: 12-18 month planning cycle. - Models/Agents/Apps: 3-month to weekly planning cycles.

The ability to harmonize these diverse cycles creates a significant strategic moat, making it difficult for others to replicate. Patel believes AI will compress all these cycles.

The Massive Data Center Buildout

Patel highlights the unprecedented scale of the current data center buildout, driven by AI. Cisco plays a central role in this, particularly in networking. The evolution of GPU-driven AI workloads has led to: - Scale-up networking: Interconnecting GPUs within a server and then within a rack. - Scale-out networking: Connecting racks of servers within a data center. - Scale-across networking: Connecting data centers hundreds of kilometers apart to function as a single logical unit, addressing power constraints.

Cisco provides the silicon, switch trays, and photonics (optics) for these networking solutions. In just two years, Cisco's AI workload business with hyperscalers grew from zero to $9.3 billion in orders, significantly exceeding initial projections.

Patel dismisses concerns about overbuilding, arguing that unlike the dot-com boom, current AI infrastructure is being consumed almost instantly. He points out that an AI agent is 450% more consumptive of network bandwidth than a human performing the same task, indicating a massive future demand.

Personal AI Usage and the "Bubble" Debate

Patel shares his recent experience with an AI agent called "Instinct," which processed over 146,000 unread emails in his personal inbox, reducing them to 200. This demonstrates the immense token consumption and potential of AI.

He addresses the "AI bubble" debate, stating that while there might be frothiness in valuations for some companies, the underlying secular shift driven by AI is undeniable. The two conditions can coexist: a transformative shift requiring massive infrastructure, and some overvalued companies in a new, experimental market.

AI Safety and Trusted Delegation

Regarding AI safety, Patel emphasizes its critical importance, especially after recent incidents like AI agents compromising Hugging Face. He believes the ability to trust AI systems for delegation is crucial for market leadership.

Cisco is developing an "agentic security platform" called AI Defense, which was launched a year and a half ago and is performing exceptionally well. This platform provides: - Visibility into model inputs. - Identity management for non-human identities (agents). - Inventorying of agents in an environment. - Model validation, including algorithmic red teaming to detect and prevent "jailbreaking." - Dynamic enforcement of guardrails at runtime to prevent unwanted agent actions.

This platform, initially developed for the chat era, has been extended to agents, including dynamic monitoring to prevent issues like agents issuing excessive refunds.

G2 Patel's Personal Journey and Advice

Patel's early life was marked by significant challenges, including a con man father, a dangerous upbringing in India, and working as a waiter at Sizzler after moving to the US at 19. Despite being a "terrible student" and having an inferiority complex, these experiences shaped his mental framework.

He became an intern at a company called Docu, and at 19, took out a loan to buy out an investor, eventually running the business for 17 years. He considers this his "biggest mistake" because the business model didn't align with his ambition to "change the world" through software and Silicon Valley.

A pivotal moment came when he joined EMC, where his mentor, Rick Devine, told him he had "one year of experience 17 times over." This experience taught him the value of continuous learning over ego-driven decisions.

His advice for entrepreneurs and those transforming organizations: 1. Ambition is the Limit: Your upper limit is bound by your ambition, not necessarily your ability. Think big and have bold ambitions. 2. Team Sport and Relationships: Business is a team sport. Take pride in building long-lasting relationships and making others successful. 3. Power of Compounding: Strive to be 1.27% better each day, leading to significant growth over time. Think in decades, not just months. 4. Align Ambition with Business Model: Ensure your ambition aligns with the business you're in. Don't grind it out in a business that doesn't fulfill your long-term goals. 5. Leverage Your Platform: If you are afforded a platform (education, location, tech), don't squander it. This is a magical time to be in tech, offering immense opportunities to make an impact. 6. Embrace Curiosity: Don't be cynical about large companies; learn from them. They were once startups, and understanding their operations can be invaluable.

Patel emphasizes that the current era of AI offers unparalleled opportunities for impact, urging young people to seize this moment with dedication and curiosity.

  Takeaways

  • Cisco has repositioned itself as the "picks and shovels" provider for the AI gold rush, delivering secure, scalable networking, security, and observability infrastructure that powers AI workloads at enterprise scale.
  • Patel champions a "founder mode" culture and a "rule of thirds" leadership mix—internal veterans, external hires, and acquired founders—to act like a massive startup in wartime.
  • Cisco is targeting two AI-era challenges: a looming infrastructure shortage as AI agent usage surges, and the need for safety, security, and observability to prevent rogue or malicious AI behavior.
  • The company mandated AI usage for its 32,000 engineers, offering unlimited tokens, which sparked rapid code‑refactoring breakthroughs and deepened its partnership with OpenAI as a design partner.
  • Patel’s personal story stresses ambition, continuous learning, and leveraging platforms, urging entrepreneurs to think big, build lasting relationships, and compound small daily improvements for long‑term impact.

Frequently Asked Questions

Why does Patel describe Cisco as the "picks and shovels" company during the AI gold rush?

Patel calls Cisco the "picks and shovels" company because it supplies the foundational networking, security, and observability hardware and software that AI developers need to run models at scale, much like tools that miners need to extract gold. By focusing on infrastructure rather than AI models themselves, Cisco positions itself as essential to the AI boom.

What is the "rule of thirds" that guides Cisco's leadership composition?

Cisco's "rule of thirds" splits its senior leadership into three equal groups: one third are long‑tenured Cisco insiders who understand the company's scale, one third are external hires from competitors who bring systems‑thinking and fresh market experience, and one third are founders or CEOs of acquired firms who retain entrepreneurial drive. This mix aims to blend deep institutional knowledge with innovative thinking.

Who is Y Combinator on YouTube?

Y Combinator is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.

Does this page include the full transcript of the video?

Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.

Helpful resources related to this video

If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.

Links may be affiliate links. We only include resources that are genuinely relevant to the topic.

Full transcript is not shown on this page

This page focuses on the summary and original notes. For full verification, refer to the original YouTube video.

PDF