Eric Vishria: Where is the Value in AI - Chips, Models or Apps? | E1206
25 Sep 2024 (16 days ago)
- Foundational models are described as the fastest depreciating asset in human history. (0s)
- There is a belief that Nvidia will not be the only significant player in the AI infrastructure space. (3s)
- A major shift in AI is occurring, which could be larger than previous technological shifts combined. (6s)
- This period is considered both the most exciting and disorienting time in the last 25 years of technology. (11s)
- Despite the uncertainty, there has been more activity in AI since 2010 and 2011. (18s)
- The speaker expresses excitement about the discussion, noting it has been five or six years since the last one. (37s)
- There is a light-hearted exchange about aging and the ineffectiveness of botox. (45s)
Reflecting on CEOship at RockMelt (1m0s)
- The CEO felt they fell short of their hopes and dreams for the company, RockMelt, despite having a good team and interesting ideas. (1m1s)
- Reflecting on the experience, it was noted that having a great team and provocative ideas is not necessarily enough for success. (1m40s)
- Distribution for a browser was particularly challenging, which contributed to the company's struggles. (2m19s)
- The experience made the CEO more empathetic towards entrepreneurs, understanding the difficulties and uncertainties involved in startups. (2m37s)
- It was emphasized that evaluating the team, timing, and product at the beginning cannot determine with certainty whether a startup will succeed due to many changing factors. (3m19s)
- The CEO would have done many things differently, such as focusing more on distribution and being more flexible with compensation for key hires. (4m8s)
- As an investor, the focus is on whether a founder has thought deeply about distribution and is constantly learning and adapting. (5m1s)
- Career investors are considered better due to their extensive experience and exposure to many companies and pitches, which provides valuable data and mental models. (6m35s)
- However, career investors may not be the best board members as they might lack empathy and understanding of the practical challenges faced by companies. (7m45s)
- The CEO's learning process for new categories involves evaluating extraordinary entrepreneurs, unique insights, and market potential rather than being a sector specialist. (9m57s)
- The CEO believes that the AI era will challenge the spreadsheet-driven, banker-type approach to venture capital, requiring more fundamental assessments. (12m33s)
The Impact of AI on Markets (12m48s)
- Vertical SaaS companies with deep data reserves can leverage commoditized foundation models to build better vertical-specific solutions. (12m48s)
- Platform shifts often involve a variant of the innovator's dilemma, requiring companies to potentially disrupt their existing business models. (13m17s)
- Successful navigation of platform shifts often depends on having a learning entrepreneur who constantly rethinks their approach. (13m47s)
- Insight development is crucial, and successful companies often have unique insights that differentiate them in the market. (14m9s)
- In highly competitive markets, violent execution by a determined team can sometimes succeed even without a unique insight. (15m10s)
- Market creation can occur when a product significantly improves an existing process, as seen with examples like Uber and AI medical scribes. (17m1s)
- Evaluating market creation involves imagining if a product will become a significant part of the market in the future. (18m50s)
- In competitive markets, the bar for believing in the entrepreneur and their insight is higher. (19m59s)
- Venture capitalists must genuinely believe in the entrepreneur and the business to effectively support and attract talent. (20m20s)
- The quality of a product may be less important than the existing distribution modes of incumbents, who can leverage their scale to dominate the market. (21m24s)
- Incumbents like Microsoft and Nuance are aware and proactive, posing significant competition to new entrants. (22m0s)
Does AI Enhance Revenue or Erode Margins for Companies? (22m21s)
- AI can both enhance revenue and erode margins for companies. (22m21s)
- The impact of AI on pricing models, such as price per seat, is complex and not straightforward. (22m42s)
- In the AI coding area, significant capital has been invested in AI tools like co-pilot and AI software engineers. (22m50s)
- A fully loaded software engineer might cost $200,000, with $10,000 allocated to tools like Jira, GitHub, and development environments. (23m0s)
- The market cap created from this $10,000 tool spend is substantial, reaching hundreds of billions of dollars. (23m32s)
- If AI can capture more of the $200,000 value attributed to software engineers, it could lead to significant value creation, potentially 20 times more. (23m47s)
- This situation presents both exciting opportunities and challenges, as the potential prize is enormous, justifying high investments. (24m10s)
- However, the industry is still some distance away from fully replacing software engineers with AI, and the market value capture is not yet fully realized. (24m47s)
- Companies must navigate this tension and challenge daily as they integrate AI into their operations. (25m3s)
Analyzing Revenue Quality: Sugar High vs. Sustainable Revenue (25m13s)
- Companies are reaching $10 million in revenue faster than ever before. (25m19s)
- Many AI products are seen as almost magical by customers, leading to high demand and rapid revenue growth, even if the long-term viability of the product or its monetization strategy is unclear. (26m19s)
- While revenue is scaling quickly in the AI sector, the monetization strategies are still being figured out, similar to how it took time for search engines to find effective monetization models. (30m19s)
Value in the Stack: Compute vs. Models (30m36s)
- The value in the AI stack today is primarily in compute, as evidenced by Nvidia's success, and in models, with OpenAI being a notable example. (30m36s)
- Foundational models are considered the fastest depreciating asset in human history, which raises questions about the long-term value of companies like OpenAI. (31m2s)
- The competition among companies like OpenAI, Anthropic, Meta, and Google benefits consumers by pushing the state-of-the-art in AI. (31m23s)
- Benchmark has not invested in foundational models but has made significant investments in AI infrastructure, such as Cerebrus, a semiconductor and systems company, and other infrastructure software companies. (31m53s)
- The future of foundational model companies may involve acquisitions by larger players, but not all will be acquired. The ability to raise significant capital remains crucial for these companies. (32m55s)
- Large acquisitions in the AI space, such as $30 billion deals, are not unprecedented, but $100 billion acquisitions would be. Antitrust issues could also pose challenges. (33m49s)
- Benchmark's disciplined approach to fund size and investment flexibility allows them to write large checks if necessary, without being constrained by traditional portfolio construction considerations. (34m37s)
- The focus is on finding exceptional opportunities led by exceptional people, rather than adhering to strict portfolio construction rules. (36m21s)
- Benchmark's AI portfolio includes a mix of infrastructure software companies, a semiconductor company, and a few application companies, reflecting a balanced approach. (37m31s)
- The firm adapts to the current market conditions, choosing to be more or less active depending on the opportunities available, as seen in their limited investments during the 2021 SaaS boom and increased activity in the current AI shift. (38m21s)
Are We Overestimating AI's Impact in the Short Term? (39m43s)
- It is important to be able to identify extraordinary opportunities and companies. (41m12s)
- Venture capitalists should focus on a few key questions when making investment decisions, such as understanding the chemistry with the entrepreneur and the cogency of their insights. (43m27s)
- It is important to have a good partnership with other venture capitalists, where each partner brings their own perspectives, biases, and insights to the table. (47m16s)
Does a $3M Gross Margin Matter in the Long Run? (48m54s)
- Early-stage gross margins are not necessarily indicative of long-term financial success. (49m12s)
- Traditional financial metrics used in stable industries like SaaS may not be applicable to rapidly evolving sectors like AI. (49m34s)
- Companies can undergo significant transformations as they grow, making early-stage financial projections unreliable. (50m39s)
Balancing Time Across Sourcing, Diligence, and Servicing (51m14s)
- The speaker currently spends approximately 80% to 85% of their time working with their portfolio companies. (51m41s)
- The speaker is on the board of 12 or 13 companies, with four or five being in their early stages. (51m51s)
- The company uses a voting system from 1 to 10 to quantify feedback on potential investments, with scores of 6 and above indicating approval and 4 and below indicating disapproval. (53m40s)
Takeaways from Working with Bill Gurley, Peter Fenton & Matt Cohler (55m42s)
- Even great companies can be overvalued. (56m5s)
- Price is a mental trap. (58m25s)
- The most important skill for an investor is understanding the depth of an entrepreneur's insight. (59m35s)
Quick-Fire Round (1h0m7s)
- It is believed that Nvidia will not be the only dominant player in the AI infrastructure market. (1h0m7s)
- Jim Gats is highly respected for his success in both consumer and enterprise sectors, including significant wins like WhatsApp and Palo Alto Networks. (1h0m47s)
- The decision not to join Sequoia Capital in 2008 was influenced by a desire to be a founder, leading to the creation of Rockmelt, which was later acquired by Yahoo. (1h2m1s)
- An unmade decision that weighs heavily is not pursuing early acquisition interest in Rockmelt, although in hindsight, it may not have significantly changed the outcome. (1h3m0s)
- A story about a wise man in ancient China illustrates the unpredictability of events and their outcomes, relating to the ups and downs of founding and building a company. (1h3m52s)
- Advice before the birth of children would be to prepare for a significant inversion of personal time, with weekends becoming more exhausting than weekdays. (1h6m1s)
- A memorable moment at Benchmark was when the founders of Confluent chose Benchmark for their Series A funding, marking a significant milestone. (1h7m45s)