The 5th Inning | Citrini on Phase II of AI, Semis, Robotics, Healthcare, and Teradyne

Episode Summary

Executive Summary: Citrini argues AI has moved from easy “phase one” semiconductor picks into a harder but still attractive phase centered on robotics, automation, and real-world implementation. He favors cyclically beaten-down beneficiaries with improving secular tailwinds—especially Teradyne, power chips, and select software/data names—while remaining cautious on Chinese equities, health insurance, and tariff risk until markets react.

Main Topics: AI moves from phase one semis to phase two implementation (Priority: 5/5): Citrini says the easy money in AI semiconductors has passed, and investors must now look for beneficiaries of actual AI deployment, especially where valuations are still reasonable and the business can compound from real-world adoption. Robotics as the next major AI trade (Priority: 5/5): He frames robotics—especially warehouse automation, industrial automation, and humanoid-adjacent systems—as the next phase of AI, driven by better vision-language-action models and cheaper sensing/hardware. Teradyne as an underappreciated robotics and cyclical recovery play (Priority: 5/5): Teradyne is highlighted as a prime example of a cyclically depressed company with robotics upside, especially via Universal Robots and Amazon warehouse automation, with a sizable revenue/EPS uplift possible. Selective semiconductor exposure beyond NVIDIA (Priority: 4/5): While still bullish on NVIDIA, he prefers shorter-horizon opportunities in wide-bandgap power chips, lagging-edge names, and contract manufacturers that may benefit from data center, EV, and robotics demand. Healthcare, GLP-1s, and health insurance under pressure (Priority: 4/5): He sees opportunities in medical devices, biopharma, and second-order GLP-1 beneficiaries, but is not bullish on health insurers due to regulatory and cost pressures. China robotics and selective Chinese equities (Priority: 4/5): After visiting China, he sees strong robotics momentum there but remains cautious on Chinese equities overall due to macro volatility, anti-U.S. momentum, and valuation expansion without enough earnings support. Tariffs as a risk but not yet a market-imposed priority (Priority: 3/5): He thinks tariffs could matter for inflation and growth, but says he will stay cautious rather than aggressive until markets begin to price the threat more seriously.

Key Arguments: The AI trade is no longer about obvious first-order semis alone; the next edge is identifying companies that monetize AI through actual implementation. Robotics is the clearest “phase two” of AI because it translates digital intelligence into physical-world productivity gains. Teradyne is attractive because it combines cyclical recovery potential, robotics upside, and an underappreciated market position. Wide-bandgap power semis like SiC and GaN may outperform over the next year because power delivery bottlenecks matter more as AI/robotics/data centers scale. NVIDIA remains structurally strong, but its size means investors should increasingly seek adjacent beneficiaries rather than simply concentration in the mega-cap leader. Health insurance is too complicated and politically exposed to be a clean bet; medical devices and some smaller healthcare names look more actionable. GLP-1 adoption is creating second-order winners in aesthetics, medical devices, and possibly healthcare cost reduction over time. China is compelling in robotics hardware and supply chain depth, but macro and policy uncertainty make broad Chinese equity exposure less attractive right now. Tariffs are a real risk, but he prefers to wait for market confirmation before repositioning aggressively. AI will create winners among companies with proprietary data, mission-critical workflows, or strong user/data flywheels, not just flashy agent builders.

Data Points: Teradyne Amazon revenue uplift: about $400 million to $500 million - Estimated added revenue from Amazon’s Vulcan warehouse robot deployment across the network Teradyne EPS uplift: about $1 per share - Back-of-the-napkin estimate from the Amazon robotics contract Amazon robot unit cost: $150,000 to $250,000 each - Estimated cost per warehouse robot used in the deployment math Amazon robot count estimate: about 5,100 robots - Estimated number needed to handle Amazon’s warehouse workload Amazon handling throughput: 300 items an hour, 20 hours a day, 365 days a year - Inputs used to estimate robot deployment needs Items per robot per year: about 2.2 million - Calculated throughput used in the Teradyne/Amazon sizing argument Teradyne market positioning: around 20% off 52-week or 3-year lows - Cited as part of the bullish setup for the stock NVIDIA market cap: $4 trillion - Used to illustrate that NVIDIA is now too large to be treated like a normal niche AI winner SMH AUM: over $23 billion - VanEck Semiconductor ETF promotion describing the fund’s scale Counter-UAB basket performance: 50% year-to-date / since inception - Best-performing thematic basket in Citrini’s 2025 trade list Ukraine-Russia normalization basket: 47% - Second-best performing thematic basket in the 25 trades for 2025 set Passive electrical components basket: up 14% - Recovered strongly after being down about 30% earlier in the year Passive electrical components basket drawdown: down 30% year to date at the low - Used to show how sentiment and cyclicality can reverse Remittances basket: worst performing - Long idea failed because fear of deportation reduced work and remittance capacity Rare diseases basket: second worst performing - Biotech/rates/FDA and political sentiment weighed on the theme Lilly combination-drug phase 2 result: 10% body-weight loss and 5% lean muscle gain - Discussed as a potentially transformative obesity treatment advance Lilly Direct GLP-1 price: about $500 per month - Used to explain why access remains limited for most consumers China Unitree humanoid robot price: about $16,000 - Example of cheap Chinese robotics hardware scaling rapidly China robot dog price: about $4,000 - What they bought during the China visit to illustrate falling hardware costs Universal Robots acquisition: 2015 - Teradyne’s acquisition that underpins its robotics exposure AI market opportunity framing: roughly a trillion-dollar annual revenue opportunity - Used to describe the broader generative AI/application stack opportunity Chegg AI disruption example: first AI name to fall on its sword - Illustrates how obvious AI losers/winners can reveal market changes

Pivotal Quotes: "I would say we're in like the fifth or sixth inning" — Citrini: Describing the maturity of the AI investment cycle and why the easy gains are gone "The second phase of AI is in robotics" — Citrini: His central thesis for the next leg of the AI trade "I think Tesla might've been a documentary" — Citrini: Referring to the unsettling capabilities of Unitree’s robot dog and the speed of robotics progress

Implications: Listeners should expect AI investing to get more selective: favor robotics, power infrastructure, and beaten-down cyclical beneficiaries with real monetization paths. Broad AI/semis exposure still works, but the next edge likely comes from implementation, data moats, and second-order effects.