newsfilter.io

Chips, Gigawatts, AI, & Agents: Coatue's Public Markets Update | CIO Jaimin Rangwalla

  • Private AI companies are achieving valuations and revenue scales previously unseen before an IPO, breaking into the world's top 25 companies while remaining private.
    • OpenAI's most recent round reached over $800 billion.
    • Anthropic's last round was in the high $300 billions.
    • SpaceX's transaction with XAI valued the company at $1.25 trillion.
    • These figures contrast sharply with the 2012 Meta (Facebook) IPO, which was the largest at the time at approximately $100 billion.
  • New AI players are achieving $25 billion in annualized revenue at one-third to one-half the speed of traditional hyperscalers and the original "Mag 7" technology giants.
    • OpenAI and Anthropic are adding approximately $10 billion in revenue monthly, equating to nearly $2.5 billion weekly.
    • This weekly revenue addition exceeds the total annual revenue of most companies in the SaaS universe.
    • Combined, OpenAI and Anthropic are approaching $60 billion in ARR, surpassing major legacy SaaS platforms like Salesforce and ServiceNow in a fraction of the time.
  • KOTU is redefining its investment framework to follow the "Gigawatt" rather than specific hardware like GPUs, focusing on power as the atomic unit of AI growth.
    • The firm is identifying specific bottlenecks and opportunities within the power supply chain, including generation, transmission, and distribution.
    • Investment focus is shifting to the inputs required to deliver gigawatts and the users of the token output generated by these power sources.
  • Supply constraints in the AI sector have persisted longer than historical norms, with some memory supply agreements now extending commitments through 2030.
    • KOTU CIO Jamon notes that the "tightness" is intensifying rather than easing, driven by trillion-dollar buyers making long-term capital commitments.
    • The firm identifies multiple simultaneous bottlenecks across power, memory, optical components, and skilled labor, creating a complex supply environment.
  • Architectural shifts in AI hardware are driving a massive increase in semiconductor demand, with projections of a CPU-to-GPU ratio flip from 1:4 to 4:1.
    • The rise of autonomous agents, which perform serial processing tasks, is increasing demand for CPUs relative to GPUs.
    • This shift is modeled as a 16x expansion in the market size for certain semiconductor categories.
    • Individual users' technology footprints are expected to expand 10x as they host hundreds of virtual agents requiring persistent memory and processing power.
  • KOTU distinguishes between "sellers of shortage" and "buyers of shortage" to evaluate market performance and stock valuations.
    • Sellers of the shortage (memory, optical, power infrastructure, labor) are seeing significant margin expansion and earnings growth of 3x to 7x due to pricing power.
    • Buyers of the shortage (hyperscalers like Microsoft, Amazon, Meta) are facing compressed multiples as high CapEx costs outpace near-term cash flow generation.
    • Hyperscalers with vertical integration, such as Google (TPUs) and Amazon (Tranium), are viewed as potential hybrids benefiting from both sides of the equation.
  • Public market performance remains decoupled from negative news sentiment, driven by strong fundamentals rather than sentiment analysis.
    • S&P 500 earnings growth is projected to accelerate to 18% through 2026, supporting market resilience despite headlines.
    • KOTU analysis shows no correlation between negative news sentiment and market direction, noting that fundamentals and earnings revisions are the primary drivers.
    • Specific sector leaders are consistently beating earnings estimates by 30% to 40% per quarter.
  • The AI ecosystem is evolving from chatbots to autonomous agents capable of spawning sub-agents and executing complex tasks without human intervention.
    • OpenClaw and similar "harness" technologies allow AI to control external devices, send messages, book reservations, and manage calendars autonomously.
    • This shift reduces the "human in the loop" constraint, enabling exponential scaling of work output through concurrent agent pools.
    • Persistent memory in next-generation accelerators is solving the "amnesia" problem, allowing agents to retain context across sessions.
  • KOTU anticipates that the IPOs of major private AI firms will not disrupt the public market but rather expand the available investable universe for retail investors.
    • The firm views the addition of three $1 trillion companies as a normal market evolution, adding to the S&P 60-70 trillion total market cap.
    • New entrants are expected to refresh the leadership of the top 25 tech companies, which typically reorders every five years.
  • KOTU identifies potential risks centered on technology breakthroughs that could render current hardware requirements obsolete or regulatory interventions.
    • A breakthrough in model efficiency (e.g., doing more with less power) could disrupt current supply chain valuations, though the firm views this as beneficial for long-term adoption.
    • Regulatory environments are expected to support national security interests in AI rather than stifle growth.
  • The firm's internal operations and investment thesis are adapting to the unprecedented pace of AI innovation.
    • KOTU has established a "bullpen" style office to foster deep collaboration between data scientists, analysts, and AI specialists.
    • The firm hired Frank, described as an "AI mad scientist," to research and implement cutting-edge tools using the firm's 20-year data history.
    • The investment approach emphasizes sprinting within a marathon mindset, prioritizing high-conviction ideas in a rapidly changing landscape.