Treat Frontier AI leaders such as OpenAI, Anthropic, and SpaceX as foundational core holdings rather than speculative bets to capture an estimated $30 trillion economic transformation. Over the next decade, target AI's primary physical bottlenecks by investing directly in energy infrastructure, power grid modernization, and high-density data centers. Expand long-term growth allocations into physical-world technologies, focusing on enterprise robotics, autonomous vehicles like Waymo, and AI-enabled healthcare services. Reduce exposure to legacy seat-based SaaS and debt-heavy 2021–2022 vintage private equity software funds, rotating only into AI-native software platforms. In private markets, avoid median venture capital funds with extended holding periods and concentrate capital exclusively in top-tier managers holding concentrated stakes in category winners.

Artificial Intelligence & Frontier Models (OpenAI, Anthropic, SpaceX)

• Frontier AI model companies represent between $3.5 trillion to $5 trillion in potential enterprise value, scaling faster than any prior technology cycle.
• The AI sector reached $100 billion in revenue in just 4 years, compared to SaaS which took 15 years to hit the same milestone.
• Unlike traditional software startups where excess capital creates overhead friction, frontier AI benefits from throwing capital directly into compute to compound product and business advantages.
• The total addressable market (TAM) spans across $30 trillion of GDP, shifting AI from a satellite portfolio allocation into a core or super-core holding.
• In late-stage venture investing, individual category-defining outcomes are expected to reach $100 billion+, enabling late-stage funds to achieve fund-returning returns on single positions (5–10%+ fund sizing).

Takeaways

• Treat AI exposure as a foundational core asset rather than an experimental allocation, prioritizing top frontier model leaders and infrastructure providers.
• When investing in venture or growth vehicles targeting AI, seek managers with concentrated position sizing in top-tier category winners rather than broad, diluted exposure.

Legacy Software-as-a-Service (SaaS) & Private Equity Software

• Public SaaS valuations have sharply corrected, with only 15 to 20 SaaS companies trading above 10x revenue, down from dozens in previous years.
• Software companies acquired during the 2021–2022 leveraged buyout (LBO) boom at 25x to 32x EBITDA (backed by over $200 billion in debt) are facing severe valuation compression, now valued at roughly half their purchase price.
• Public markets are heavily penalizing slow-growth legacy software; data shows that 1% of growth acceleration is being valued as equivalent to 3% of EBITDA margin.
• Software firms without native AI integration face extreme terminal value risk and may lack exit liquidity, as private equity buyers and public markets reject low-growth, seat-based workflow software.

Takeaways

• Exercise caution with legacy seat-based SaaS companies and private equity software funds heavily exposed to 2021–2022 vintages.
• Look for software businesses capable of executing radical pivots to AI-native architecture (such as Intercom) or those accelerating organic growth through AI agent deployment.

Physical AI Infrastructure, Robotics & Energy

• The primary bottleneck in the AI ecosystem has shifted from demand to the physical supply side: power transmission, grid capacity, specialized high-density data centers, and advanced chips.
• Emerging massive TAM opportunities over the next 10 years are expected in physical-world applications, including robotics, autonomous vehicles (e.g., Waymo), manufacturing, defense, and healthcare delivery/drug discovery.
• AI data center power density requirements are roughly 10x+ higher than legacy data centers, rendering older facilities unsuitable without complete redesigns.

Takeaways

• Broaden AI investment horizons beyond model developers to include physical bottlenecks: energy infrastructure, high-density AI data center construction, grid modernization, and specialized hardware.
• Position for multi-year growth in physical AI applications such as enterprise robotics, autonomous transport, and AI-enabled healthcare services.

Venture Capital as an Asset Class

• Venture capital exhibits extreme power-law distribution: historical data across 3,000 U.S. VC firms reveals that only 20 firms (less than 1%) have consistently generated 3x net returns over a 20-year period.
• The average 10-year venture capital net return across the industry is 1x to 2x net, which underperforms top-tier public equities and private equity after accounting for illiquidity.
• Top-performing early-stage venture funds routinely accept a ~60% loss rate to capture category-defining companies that deliver 10x+ returns.
• The venture ecosystem is seeing the "death of the middle," favoring either massive lifecycle platforms with extensive operational support and follow-on capital, or highly agile, specialized pre-seed/seed funds.

Takeaways

• Individual and institutional allocators should avoid median or broadly diversified venture funds; venture exposure only generates premium returns if invested with the top 1% of managers who have direct access to category winners.
• Expect extended liquidity timelines (often 10+ years), requiring patient capital capable of allowing category winners to compound private gains.