Roundtable
Anthropic's $30T Assumption & OpenAI Confirms IPO | Why Customer Service & Robotics are Overinflated
- A $9 billion exit is deemed insufficient for seed fund returns in 2026, as a 50x return with dilution would require an exit valuation of approximately $63 billion; however, this bar is only cleared if the definition of "seed" is expanded.
- Venture investors are predicted to succeed by securing 15x returns on failures, while hyper-growth markets may see significant numbers of acquisitions where standalone DCF is negative but the strategic value to the acquirer remains high.
- NVIDIA's acquisition of Poolside involves licensing the model factory for $6 billion, investing an additional $1 billion at a $12 billion pre-money valuation, and transferring 109 engineers to Nemotron.
- The market is expected to hit a capital intensity wall across all players, with only $1 to $2 trillion outcomes per decade likely to occur, concentrated primarily within frontier models.
- Mercury (Merkur) is forecast to reach $2 to $2.5 billion in ARR, currently led by General Catalyst at a $20 billion valuation with rumors of NVIDIA participation.
- NVIDIA's strategy involves deploying $75 billion of free cash flow toward customers, the ecosystem, and its strategy team, with growth projected to continue at 60% to 70% quarter-over-quarter.
- Second-time founders are predicted to generate $1 billion each in the current cycle, while the largest data providers are expected to become $200 billion companies.
- OpenAI aims to go public in 2027, but faces competitive pressure to accelerate growth or risk irrelevance; if it fails to outpace Anthropic's platform positioning, its IPO pricing may be forced lower than competitors.
- Anthropic's gross margins improved from negative 91% to positive 30% within one year, contrasting with predictions that OpenAI may exit at a lower price if it does not significantly alter its trajectory.
- A shift toward enterprises developing their own models is anticipated as IT sectors realize frontier models could displace their work, leading to a reckoning on AI spending and ROI control by 2027.
- Hugging Face could be acquired for $13 billion as a strategic asset to counter closed weights, with buyers advised not to alter the platform for 24 to 36 months to avoid breaking the ecosystem.
- The current transition to open weights is described as a temporary moment to sell assets, as the market is expected to move past this phase, potentially rendering public software companies below AI leaders obsolete.
- Personal productivity tools are forecast to remain mediocre due to Silicon Valley's misalignment with average American efficiency goals, while customer support software is predicted to be eliminated within 24 months.
- Humanoid robotics is viewed as a difficult investment with overestimated potential, and accounting or law firms are considered poor categories due to a lack of expected exits despite high cash injection.
- Stripe is accelerating with a 41% growth rate and billing up 71%, while security and hallucination issues currently prevent agents like GrokBot and Instinct from being trusted for critical tasks.
- Token addiction is expected to trigger a societal backlash next year, while a future environment of unbounded creativity is predicted to allow founders to rewrite industry rules.
- Risks include overextending credit with unrealistic projections, running out of capital or demand to end the current cycle, and the concentration of 10% market cap value in data providers if frontier model spend grows from $100 billion to $400 or $500 billion.