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.