Kleiner Perkins' VC Philosophy and Practice

From the end of 2024 through 2025, the explosive adoption of generative AI created a supercycle that is being described as comparable to 'Internet 1.0'—or perhaps even ten times larger. Kleiner Perkins partner Mamoon Hamid stated on the 20VC podcast that 'it is the most exciting time to be alive,' and discussed the structural transformation brought by AI, the opportunities and challenges for investors, and the changing profile of founders from multiple perspectives. This article organizes the key points of the interview and explains the technical content in plain English. The intended readers are business professionals tracking tech industry trends and those interested in startup investment.

1. Structural transformation brought by the AI supercycle

1-1. $100 billion-class infrastructure investment and 'frontier models'

• Larry Ellison said that 'the cost of entering frontier models will reach $100 billion,' and Google, Microsoft, Amazon, and Meta continue to invest heavily in model training and GPU clusters.

• Due to the high barriers to entry, the Foundational Models layer is becoming a business similar to 'ultra-large power companies,' and VC-backed startups are forced to differentiate themselves in the Application Layer.

1-2. A new billing unit: 'Software + Labor'

Hamid emphasizes that 'AI is shifting to an era where billing is based on labor costs rather than per-seat costs.' It is not uncommon for a $30/month SaaS to jump to $300 with the integration of an AI copilot. This means that the TAM (Total Addressable Market) is expanding rapidly in two directions:

1. leveraging 'rare and high-wage' professional skills such as doctors, lawyers, and developers with AI

2. including advanced judgment tasks that could not be turned into software previously into the scope of automation.
   

2. A three-layer model for segmenting investment opportunities

2-1. Foundation models vs. middleware vs. application layer

2-2. Investment preference for 'market-creating products'

Hamid says, 'I love companies that build a new playing field from scratch, like Slack or Figma.' Market-creating startups have the advantage that they

• can design the rules themselves

• and are more likely to become platformers due to the first-mover advantage. However, since the TAM is difficult to read, investors need to scrutinize behavioral metrics such as usage frequency and retention rates to increase their conviction.

3. New common sense for valuation and fundraising

3-1. 'YOLO bucket' and tolerance for rule-breaking

'20% of strategy should be non-strategy'—At Kleiner Perkins, deals chasing extremely high-priced rounds (e.g., a $750 million valuation pre-product) are limited to a 'YOLO bucket,' keeping them to a very small fraction of annual investments to mitigate risk across the entire portfolio.

3-2. The iron rule of a '60:40' reserve ratio

Secure 60% of the equity in the initial investment and reserve the remaining 40% for Series B and beyond.

• Signaling effect: If a board-level investor does not participate in a follow-on round at all, it is likely to send a negative signal to other investors.

• Cash management: When the macro environment worsens, rescuing the existing portfolio takes priority over new investments. In the case of Box (2008), they managed to survive until their IPO through three bridge funding injections.

4. Founder Archetypes and Market Approach

4-1. Two Major Archetypes

4-2. Avoiding "Top-Down Entrepreneurship"

Calculating TAM through market research and then trying to fill a whitespace—analysis-driven founding is not Hamid's preference. He maintains the philosophy that a bottom-up approach, where "obsession with the problem to be solved" comes first, should prevail.

5. Risk Management and Learning from Failure

5-1. Large Losses at Tally and Lessons Learned

In the consumer lending company Tally, a total of $30 million was invested, but the business model collapsed due to a sharp rise in interest rates, leading to liquidation.

• Lesson: Consumer lending is vulnerable to interest rate trends. Scenarios for model changes should be rigorously verified.

• Improvement: Curb "dreamy optimism" and re-evaluate macro conditions before making additional investments.

5-2. The Pros and Cons of "Believing Too Much"

While Hamid self-critically says, "I am a dreamer, and sometimes I believe too much", he also

6. Outlook for the Next 10 Years and Investment Strategy

6-1. Technology Spending to Grow from 15% to 20% of Global GDP

• Of the current global GDP of approximately $100 trillion, technology spending accounts for $15 trillion (15%).

• If it expands to 20% ($25 trillion) by 2035, a new market worth $10 trillion annually will be created.

• AI will not just replace "software" but directly replace "white-collar labor," shifting the labor market share toward the tech industry.

6-2. Conditions for Restarting IPOs and M&A

• If large-scale IPOs on the level of Stripe or Databricks resume from 2025 onwards, capital markets will be revitalized, and exit paths for middleware companies will become clearer.

• Meanwhile, mega M&A by GAFAM is stagnating due to stricter regulations. A trend of 'unlisted unicorns acquiring unlisted startups via stock swaps' will become an alternative scenario.

Mamoon Hamid says, 'VC is often misunderstood as an easy job, but in reality, it is a series of gritty, incremental efforts.' In the AI supercycle, where incumbents with massive capital clash with startups armed with speed and tenacity,

1. Early bets on market-creating products

2. 60:40 reserve management and a moderate YOLO allocation

3. Long-term commitment to founders

These three principles will be the compass that determines success or failure over the next decade. For both VCs and entrepreneurs, returning to the fundamentals of 'designing the playing field itself and winning the game' is the key to surviving the AI era.

Related Articles