Fireside Chat, Conference Presentation
Why VC Today is Worse than 2021
- The 2025 investment landscape is expected to favor doubling down on the very largest companies as the most profitable strategy, while vertical AI sectors face an elevated risk of rapid Total Addressable Market (TAM) exhaustion due to excessively high market expectations.
- Market de-acceleration or saturation may render current industry estimates regarding future outcomes incorrect, with the "2020 playbook" potentially leading to catastrophically wrong growth extrapolations over the next four to five years if applied to the B2B AI sector.
- Venture capital firms anticipate a 15-year cycle extension due to equity market conditions, though returns could be 40% to 50% lower than anticipated, resulting in fund multiples of only 2x to 3x, while five-year returns have lagged public markets, pressuring LPs to demand improvement.
- Benchmark is projected to remain stable despite partner departures, leveraging strong traditions and portfolio to recruit adjacent talent, while offering new partners generative backdated carry arrangements involving entities like Fireworks and McCaw.
- Ambitious professionals are expected to prioritize rapid career acceleration within an eight-year timeframe at firms such as Vista, Bond, Founders Fund, or Benchmark, despite AI engineering roles at companies like Meta offering earnings that are projected to exceed venture capital compensation in the foreseeable future.
- The trend of companies remaining private longer, exemplified by Revolut's $75 billion valuation, signals a continued preference for private markets, with investors likely to shift strategy toward startups generating $100 million in ARR but holding less than 1% market share to mitigate TAM exhaustion risks.
- Investment outcomes are expected to derive best from companies starting in small markets and expanding TAM sequentially, similar to Spotify or Deel, whereas founders starting in circumscribed markets may fail to overcome size constraints regardless of execution skill.
- Vertical SaaS companies other than the largest are expected to struggle to scale for fund returns due to insufficient notional TAMs, though AI could increase deal sizes in legal software from $100–200k to $1 million and raise general contract values from $10,000 to $100,000 annually.
- The AI sector is predicted to shift from a software-centric game to a fixed-asset intensive one, requiring billions in capital expenditures to reach cash-flow break-even, with Poolside building a two-gigawatt data center due to the inability of existing providers to supply sufficient compute scale.
- Capital intensity in AI is expected to rise from a $500 million to a $5 billion break-even requirement, with investors concentrating on three pillars: absolute winners like OpenAI/Anthropic, de facto public entities like Revolut/Deel, and very early-stage assets, while Microsoft may step back to let partners like Oracle assume significant balance sheet risk at ratios reaching 4.6x.
- A market correction or retrenchment is expected to occur inevitably, potentially triggered by over-extrapolation of one-year adoption rates leading to a mild glut in data center capacity, with demand for inference anticipated to grow by three orders of magnitude as apps shift to 24/7 continuous operation.
- The current enthusiasm for B2B sectors like legal software is viewed as temporary and driven by exogenous AI hype, potentially mirroring the 2020 phenomenon where adoption rates could revert to only 5% of companies "in market" within 24 months.
- Consumer AI tools are expected to maintain continuous demand for new generations, contrasting with B2B tools that face cyclical adoption and high onboarding costs, while investors are likely to make more mistakes in B2B AI due to delusional application of previous playbooks.
- The speaker predicts that fund returns will be lower than expected and that the "infinite spigot" of venture capital may close if AI fails to deliver returns relative to public market benchmarks, forcing a shift toward selling companies via M&A if TAM is not accelerating.
- Extended periods of non-payment for venture carry, potentially lasting 10 to 12 years following an initial payout, are expected, alongside a projected 5x multiple on paper for a 2017 fund by the end of the current year, though the speaker does not expect to sell winners in the current privatized market environment.
- Legal and AI markets face moral and operational risks, including potential copyright infringement lines and increasing difficulty in content moderation over the next five years, while Lovable is not expected to hit $1 billion in ARR by the end of next year due to its smaller prosumer market position.
- Early-stage investing is expected to become more difficult as AI tools reduce the engineering barrier to entry for non-technical founders, while the "time to exit" in venture capital is anticipated to elongate, complicating the management of investments ranging from $1 million to $100 million ARR within a single fund structure.
- In a bull market, aggressive risk-takers are predicted to appear the smartest immediately prior to a crash, accumulating significant wealth during the upswing before suffering the greatest losses, while the most profitable strategy involves focusing on the very largest companies.
- The speaker anticipates that competition for compute capacity has become so intense that companies must build their own data centers, reflecting a move toward economic rationality where OpenAI is expected to spend more capital with Oracle than Microsoft.
- Investors are expected to be more successful allocating capital to "anointed winners" and post-public eligible private companies rather than early-stage assets, while international payroll markets are expected to see the emergence of a few large companies due to global talent access expansion.