The Biggest Bottlenecks For AI: Energy & Cooling

Market Macro Trends

• US-based technology companies now constitute 5 to 8 of the top 10 most valuable firms globally, with tech having "swallowed the whole market."
• Companies are staying private significantly longer; the average time to IPO has extended from a 5–10 year window to approximately 14 years.
• Aggregate market capitalization of private companies valued over $1B has grown 7x over the last decade to roughly $3.5 trillion.
• Only ~5% of software/internet public companies now forecast growth exceeding 25%, pushing high-growth opportunities almost exclusively into the private market.
• The firm expects the trend of delayed IPOs to persist due to private market adaptations like tender offers and SPVs.

AI Infrastructure and Investment Thesis

• AI infrastructure build-out is characterized by massive scale, with big tech companies (Google, Meta, Amazon, Microsoft) targeting ~$400B in annualized CapEx dedicated to AI.
• Input costs for AI models have declined >99% (approx. 100x) in the last two years, outpacing Moore's Law.
• Model frontier capabilities are improving by a double-digit factor every seven months.
• Unlike the early internet cycle, the current build-out is funded primarily by large, profitable tech giants rather than leverage-heavy private debt, reducing systemic risk.
• AI adoption is 5.5x faster than historical precedents; ChatGPT reached 365B searches in two years versus Google's 11 years.
• Current user base estimates: >1B monthly active users for major AI tools, with 1.5–2B active users globally.
• Energy and Cooling Bottlenecks:
  • Energy capacity is the primary near-term constraint on AI scaling.
  • The firm is optimistic about nuclear power (including restarting Three Mile Island) and localized natural gas solutions.
  • Cooling infrastructure is identified as the subsequent major bottleneck once energy supply is resolved.

Business Models and Monetization

• Value Distribution: The firm's view is that 90% of AI value will accrue to end-users (surplus), while companies capture the remaining 10% as market cap.
• Pricing Evolution: Unlike previous cycles, AI enables granular price discrimination (e.g., $3/mo in India vs. $200–$300/mo in the US) based on usage intensity and features.
• Consumer Stickiness: Consumer usage is highly sticky (avg. 20–30 mins/day on ChatGPT), creating a durable user base that is less price-sensitive than expected despite free alternatives.
• Gross Margins: The firm is more lenient on gross margins for AI-native apps compared to mature SaaS, anticipating input costs will continue to fall due to model competition.
• Retention Metrics: Investment screening prioritizes gross retention rates (>90% stickiness) and ease of customer acquisition over short-term profitability.
• Task Monetization: While "seat-based" pricing remains dominant, the long-term opportunity lies in monetizing specific completed tasks (e.g., resolved customer support tickets), though this is currently in early stages.

Portfolio Strategy and Allocation

• Investment Baskets:
  • High-Momentum: Undeniable companies with rapid growth (e.g., Cursor, Decagon, 11 Labs, Abridge).
  • Top-Tier Teams: Early bets on the top 5 AI research teams globally, accepting high business variance for asymmetric capital returns due to team quality.
• Sector Focus: Primary allocation to AI infrastructure and AI apps; secondary focus on American Dynamism (autonomy, vision) and AI-enabled health.
• Crypto: Investments made in coordination with the firm's crypto team, focusing on stablecoin enablement and high-conviction growth-stage opportunities.
• Follow-ons vs. New: The firm aims to maximize follow-on investments (ideally 100%) as a proxy for early-stage success, but will pursue new deals where the best opportunities lie.
• Exits and Liquidity:
  • Portfolio includes companies with long private tenures; the firm supports tender offers to provide liquidity for talent competition.
  • DPI (Distributed to Investors) remains a key metric; the firm does not plan to force unnatural exits for companies that benefit from remaining private.

Competitive Landscape and Disruption

• Disruption Risk: Public software companies are vulnerable if they lack new UI/UX (agent-based workflows), unstructured data access, or novel business models.
• Salesforce Vulnerability: Incumbents with rigid form-based UIs and structured databases face disruption from agents that can proactively execute tasks and query unstructured data.
• Research Teams: No single alternative exists to current top-tier researchers (e.g., Ilya Sutskever), making investments in these specific teams unique and high-stakes.

Operational and Team Dynamics

• Early/Late Stage Integration: 80% of growth investments involve pre-existing relationships with the firm's early-stage teams, providing superior access and product insights.
• Collaborative Alpha: The firm relies on early-stage teams to identify market shifts 12–24 months ahead of the growth phase.
• Team Culture: The growth and early-stage teams share a high employee engagement score (91/100), facilitating deep collaboration.

Forward-Looking Statements

• AI is expected to generate a total addressable market larger than the mobile/cloud cycle ($10T previously), potentially creating significantly more market cap over the next 10 years.
• The firm anticipates continued downward pressure on AI input costs, allowing companies to deliver more value without raising prices.
• A shift is expected toward "agent-based" workflows that replace human labor, fundamentally changing UI/UX design paradigms.
• The firm expects to see a bifurcation in AI applications between "sticky" workflow-integrated tools (e.g., medical scribes, high-end financial analysis) and transient prototyping tools.