David Cahn: Why Servers, Steel and Power Are the Pillars Powering the Future of AI | E1186

Intro (0s)

• David Cahn, a twin, believes being a twin has been the most influential factor in his life. He feels it gave him a license to be non-conformist and fostered a hyper-competitive spirit, pushing him to excel.
• Cahn's family history, with his father's family fleeing Nazi Germany and his mother's family immigrating from Syria, has deeply shaped him. He feels a sense of responsibility to make the most of his opportunities, given the sacrifices his ancestors made.
• Cahn's parents, both first-generation college graduates, instilled a strong work ethic in him. They focused on continuous improvement rather than celebrating achievements, emphasizing the importance of always striving for more.

Exploring AI's Impact: Two Critical Questions (4m41s)

• The conversation focuses on the future of AI and the massive investments being made in the field. The speaker, David Cahn, believes that AI will fundamentally change society, but he also raises concerns about the high levels of capital expenditure (capex) being poured into AI development.
• Cahn argues that while AI is a promising technology, the current level of capex may be unsustainable. He points out that the entire SaaS economy is only a $250 billion market, and the current AI investments are significantly larger. He believes that it's important to consider the financial implications of these investments and how they will play out in the long term.
• Cahn highlights a shift in sentiment within the tech industry. He notes that recent statements from Mark Zuckerberg and Sundar Pichai acknowledge the risks associated with the current level of AI investment. This suggests a growing awareness that the massive capex may not be guaranteed to yield returns, even though the potential benefits of AI are significant.

Impact of Conscious CapEx Overspending in AI (8m9s)

• The realization of overspending on capital expenditure (CapEx) in AI is not likely to significantly change the current landscape. While some believe that companies like Google and Microsoft are making calculated investments, the speaker argues that these investments are speculative and driven by the desire to maintain their oligopoly in the tech industry.
• The speaker believes that this overspending is beneficial for startups. As large tech companies overproduce compute resources, the cost of compute for startups will decrease, leading to higher gross margins and more valuable companies.
• However, the speaker acknowledges that this overspending could further concentrate power in the hands of large tech companies. By investing heavily in AI infrastructure, these companies are creating barriers to entry for new players, potentially hindering innovation and competition in the AI space.

Reconciling Compute Cost and Its Future Importance (12m2s)

• The cost of compute is decreasing, but compute itself is still crucial for the future of AI. While the overproduction of compute leads to lower costs for startups, the physical infrastructure required for compute, such as data centers and GPUs, remains essential.
• The physical reality of compute is often overlooked. The term "compute" is a euphemism that doesn't fully capture the complexity of building and maintaining data centers. The rapid pace of technological advancement means that data centers quickly become outdated, requiring constant upgrades and replacements.
• The future of AI is tied to the evolution of data centers. As models become more efficient and require more processing power, the demand for larger and more sophisticated data centers will increase. This will necessitate significant investments in infrastructure, labor, and real estate, making data centers a critical asset in the AI landscape.

Is Model Progression Outpacing Data Center Development? (17m35s)

• The text discusses the potential for model progression to outpace data center development. The question is whether research breakthroughs will occur faster than the ability to build larger data centers. Some believe that scaling laws will dominate, while others anticipate breakthroughs in areas like reasoning and data utilization.
• The text explores the impact of AI integration on software companies' margins. The example of Canva highlights how AI features can increase costs without generating additional revenue, potentially eroding margins. The speaker argues that AI should add value to products to justify higher prices.
• The text examines the relationship between value creation and pricing power. While adding value can potentially lead to higher prices, the speaker acknowledges that commoditization of features can drive prices down. The speaker suggests that industries with high barriers to entry, such as those with data moats, are more likely to retain pricing power after AI integration.

The Future of Vertical Integration in the Compute Stack (21m38s)

• Vertical Integration in the Compute Stack: The discussion revolves around the trend of companies owning the entire "vertical stack" of their AI infrastructure, from hardware to software, rather than relying on external providers. This is seen as a way to gain more control and efficiency.
• Nvidia's Dominance: While vertical integration is becoming more common, the speaker believes it's difficult to bet against Nvidia's CEO, Jensen Huang, and the company's dominance in the GPU market. Huang is seen as a visionary leader, and Nvidia has a strong track record of success.
• The Importance of Data Center Integration: The speaker argues that vertical integration between the model layer (AI models) and the data center is crucial for success. Companies like Tesla and Meta are already implementing this approach, recognizing that separate teams for data centers and models won't be efficient as models grow larger and more complex.
• The Challenge for Startups: The speaker believes that startups face significant challenges in competing with large tech companies in the AI space due to the high capital expenditure (CAPEX) required for data centers and the need for a strong cash flow source outside of the AI business itself. Companies like Facebook, Amazon, and Microsoft have existing cash machines (Instagram, AWS, Azure) that allow them to invest heavily in AI.

Facebook's Unique Position Without Cloud Revenue (24m26s)

• Meta's unique position without cloud revenue allows them to be more aggressive in AI development. Meta, unlike cloud giants like Google and Amazon, doesn't rely on cloud revenue, giving them the freedom to be more creative and take risks in AI. They can afford to play offense, while cloud companies are forced to play defense to protect their existing businesses.
• Meta's open-source approach with Llama benefits both the company and the AI ecosystem. By releasing Llama as an open-source model, Meta encourages startups and developers to build upon it, fostering innovation and potentially leading to killer consumer use cases that could generate significant revenue for Meta.
• The AI landscape is characterized by a unique dynamic of partial ownership and strategic partnerships. Companies like OpenAI and Anthropic, while not fully acquired, have strong partnerships with Microsoft and Amazon respectively. This structure allows for a balance of ownership and incentive alignment while maintaining separate business entities. This dynamic is likely to persist due to antitrust concerns and the difficulty of acquiring such large and influential AI companies.

Core Bottlenecks in AI: Compute, Algorithms, or Data? (28m51s)

• The speaker initially agreed with Alex's view that data is the primary bottleneck in AI development, but now believes that compute, models, and data have converged. He argues that major AI companies have access to similar data, compute is a commodity, and model differences are less significant due to scaling laws.
• The speaker proposes a new framework for understanding AI development, focusing on "servers, steel, and power." He believes these are the key factors driving progress in the field.
• The speaker highlights the importance of server technology, particularly the chip innovation happening with companies like Nvidia and AMD. He sees this as a key area of competition and growth. He also emphasizes the role of steel in AI development, as data centers require significant construction and infrastructure. Finally, he believes that AI's energy demands will drive an energy revolution, making power a crucial factor in the future of AI.

The Future of Chip Pools & Nvidia's Product Roadmap (30m41s)

• Nvidia's product roadmap is expected to continue to deliver impressive performance improvements at increasingly lower prices. This is based on the historical trend in Silicon Valley of Moore's Law, where chips become cheaper and more powerful over time. The B100 chip is a prime example of this trend, offering significant performance gains at a competitive price.
• Nvidia's dominance in the chip market is likely to attract competition from companies like AMD and Broadcom, as well as startups. This is driven by the high gross margins and large market size associated with the chip industry.
• The political landscape, particularly regarding import/export policies and the reliance on Taiwan for chip manufacturing, could significantly impact the future of the chip industry. The potential for a US investment in its own chip supply chain is a key factor to consider, although the extent of this investment and its impact on the reliance on TSMC remain uncertain.

Key Considerations for Steel Supply & Demand Dynamics (32m26s)

• Steel is a crucial component in the manufacturing of AI infrastructure, including servers, generators, and batteries. The demand for steel is expected to increase significantly as AI development progresses, but there are concerns about the ability of steel manufacturers to meet this demand.
• The supply chain dynamics for steel and other industrial materials are complex. Large tech companies are working with manufacturers to increase production capacity, but manufacturers are hesitant to invest heavily due to the risk of oversupply.
• The power requirements for AI are substantial and will drive an energy revolution. The demand for power will outpace current supply, leading to increased investment in renewable energy sources like solar and batteries. This will create opportunities for companies like Nextera, which are already investing in clean energy technologies.

The Risks & Benefits of Open vs. Closed AI Models (37m48s)

• The discussion revolves around the societal implications of open versus closed AI models, specifically the potential risks and benefits of each approach. The speaker acknowledges the contrasting views of figures like Mark Andreessen and Alex Wang, who hold opposing stances on the dangers of open AI models.
• The speaker expresses a relatively relaxed perspective on the potential risks of AI, suggesting that the fear of AGI (Artificial General Intelligence) is overblown. They believe that AI will likely improve our lives and productivity, but that AGI is not an imminent threat.
• The speaker advocates for the coexistence of both open and closed AI models, emphasizing the importance of having diverse options. They believe that the current landscape, with both open-source and closed-source AI models, is a positive development. The speaker also highlights the inconsistency of believing in AGI while simultaneously investing in traditional software-as-a-service (SaaS) products, as the existence of AGI would render such products obsolete.

China's AI Progress: Behind or Underestimated? (40m0s)

• China's AI Progress: The speaker believes that while the US has a strong advantage in AI development due to its business environment and culture, China is catching up quickly. He emphasizes the importance of not underestimating competitors and encourages continued innovation to stay ahead.
• The $900 Billion Question: The speaker discusses the potential for AI-related spending to reach a trillion dollars, driven by the increasing demand for data centers and the "prisoners dilemma" where companies feel compelled to keep building to stay competitive.
• Off-Balance Sheet Financing: The speaker highlights the shift towards off-balance sheet financing for data centers, where companies like Microsoft and Google lease rather than own their infrastructure. This allows them to minimize the perceived cost of their investments while still committing to significant future spending.

Lessons in Deal Selection from Leading AI Companies (45m3s)

• Capital will flow to the best opportunities in the AI infrastructure market. The speaker believes that the high demand for servers, steel, and power for AI development will attract significant investment. This is due to the attractive returns offered by these deals, which are considered "too good to pass up."
• The spread between risk-free and risk-adjusted returns will compress as more investors enter the market. As more investors recognize the potential of these deals, competition will increase, leading to a narrowing of the profit margins. However, the speaker believes there is still ample capital available for these projects.
• The speaker sees a connection between Adam Smith's "invisible hand" and human psychology. He believes that Adam Smith's understanding of human behavior is key to understanding capitalism. The speaker draws a parallel to Isaac Asimov's "Foundation" series, where a character predicts the future of humanity based on predictable human behavior.

Leading at 27 & Working with Top Companies (46m55s)

• David Cahn's experience working with top companies like Marquetta, UiPath, Snowflake, and DataBricks taught him the importance of focusing on what customers actually do, rather than what they say. He learned that customers often express dissatisfaction with products while continuing to pay for them, indicating their true value.
• Cahn emphasizes the importance of learning from others in the venture capital industry. He believes in identifying the strengths of successful individuals and striving to achieve 80% of their expertise in those areas. He sees this as a way to develop a well-rounded skillset.
• Cahn acknowledges the duality of venture capital, where there is a clear definition of success (being a "slugger" who generates billion-dollar gains) but also a need for individual development and conviction. He believes that while learning from experienced investors is valuable, ultimately, each venture capitalist must develop their own style and be willing to put their neck on the line for their investments. He sees the structure of his firm, Sequoia, as a supportive environment that encourages individual decision-making and accountability.

How Sequoia Welcomes & Empowers New Partners (54m7s)

• Sequoia fosters a culture that welcomes and empowers new partners. The company emphasizes that the next great investment could come from anyone, regardless of their experience. This creates a sense of freedom for new partners to share their ideas and contribute.
• Sequoia's high standards and demanding environment push partners to excel. The company's focus on building long-term, impactful businesses sets a high bar for success. This pressure, however, is seen as positive, as it encourages partners to strive for excellence and develop their skills.
• Sequoia's partners are known for their expertise and adaptability. The text highlights the strengths of specific partners, such as Sonia's expertise in AI and Pat's ability to adapt to changing market dynamics and invest in both early and late-stage companies.

Ranking Core Pillars of Venture: Sourcing, Selecting, Servicing (1h2m4s)

• Sourcing, Selecting, and Servicing are the core pillars of venture capital. David Cahn believes that selecting the right companies is the most important aspect, followed by winning the deal. He considers sourcing to be the least important of the three, but still acknowledges its importance.
• Cahn's experience at Sequoia has changed the value proposition he offers to founders. While previously he relied on his own skills and reputation, now the Sequoia brand itself is a significant draw for founders. This has made winning deals easier, but he still emphasizes the importance of earning the right to win by demonstrating value and building relationships.
• Cahn's deal flow has not significantly increased since joining Sequoia. This suggests that he is a proactive deal hunter who actively seeks out promising companies, rather than relying solely on inbound interest. He believes that the best founders are not actively seeking out investors, and that it takes effort to convince them to meet.
• Cahn has employed creative strategies to win deals in the past. He has used platforms like Cameo to send personalized messages from celebrities to founders, and he has developed a daily "loom video" strategy to showcase interesting products. These tactics demonstrate his willingness to go the extra mile to make a connection.
• Cahn is committed to continuous improvement and learning. He recognizes that he has a lot to learn and that he is still in the early stages of his career. He is eager to learn from others and to find the next big investment opportunity. He starts his day early and is passionate about his work, believing that every meeting with a founder could lead to a groundbreaking investment.

David’s First Deal (1h8m6s)

• David Cahn recalls his first significant deal, which involved investing in Starburst Data. He remembers this deal fondly because he was the one who found the opportunity and helped secure it.
• Cahn met Justin Borgman, the founder of Starburst Data, during a customer diligence call for DataBricks. He was impressed by Borgman's intelligence and expertise in the open-source technology Presto.
• After building a relationship with Borgman over a year, Cahn helped him secure funding for Starburst Data. He was proud to have played a role in the company's early success and remembers the experience as a significant milestone in his career.

Quick-Fire Round (1h9m23s)

• David Cahn believes in God, a belief that is not widely shared in Silicon Valley. He considers this to be a point of difference between himself and many of his peers.
• Cahn's most memorable first founder meeting was with Paul Graham, the founder of Y Combinator. He was immediately impressed by Graham's vision and the potential of his company.
• Cahn's most contrarian advice for founders is to offer a significant equity stake (up to 5%) to individuals who can transform their business. He believes that the potential return on investment from such a move outweighs the traditional equity allocation models.
• Cahn has changed his perspective on many things in the past year, particularly after learning he will be a father in two months. This has led him to re-evaluate his priorities and approach to life.
• Cahn admires Mike Volpi, a venture capitalist at Index Ventures, for his investment acumen. He believes Volpi is a fantastic investor and has learned a lot from observing his work.
• Cahn believes that OpenAI will be one of the few foundational model companies to still exist in 10 years. He attributes this to the company's strong brand, consumer distribution, and the success of ChatGPT.
• Cahn has learned that even with a strong brand and reputation, the selection process for investments remains incredibly difficult. The bar for success is high, and investors must constantly strive to identify the next YouTube, Instagram, or DoorDash.
• Cahn is currently learning to drive. He acknowledges that this is a skill he should have learned earlier, but he has been putting it off due to living in New York City and London, where driving is not essential.