+First time on the top traders desk?
The desk runs its own numbers and publishes its calls next to the account that trades them. The short route:
Physical AI & Robotics Research
Robotics and physical AI are the next AI trade the desk does not want to sleep through: VC put roughly $9.4B into robotics in 2025 (+41% YoY), and 1X, Figure and Tesla are all racing to ship humanoids in 2026. But it is not a bet on a robot; it is a bet on the supply chain: compute, sensors, actuators, power and the handful of names shipping into real deployments. Demos are easy; ops are the moat. Start with is physical AI investable.
Market metrics
Sources: State of Robotics 2026, company disclosures. Long-run TAM forecasts vary wildly ($200B-$5T).
When software gains a body, the scarce layers become edge inference, sensors, actuators, control systems and real deployment distribution, not the humanoid shell everyone films for the demo. The durable trade is the picks and shovels: foundries, GPUs, EDA, power and cooling, and the industrial robotics with real installed bases, companies that take a cut of the whole physical-AI economy regardless of which robot wins.
Who leads — shipping vs demoing
| Player? | Status? | Trigger? |
|---|---|---|
| 1X (NEO) | OpenAI-backed home humanoid; 10,000+ pre-orders, deliveries targeted end-2026 | $20K / $499mo |
| Figure (03) | Hayward factory live; BMW Leipzig deployment; shipments planned by year-end | pre-revenue |
| Tesla (Optimus) | Gen 3 lines at Fremont, $20B 2026 capex — but zero external customers, not in material factory use | scale-up |
| Apptronik | Google + Mercedes backing; industrial focus | $520M raised |
| Unitree | Cheap Chinese humanoids actually shipping in volume — the price-floor threat | shipping |
| Supply chain | Nvidia, TSMC, semis, power & cooling — paid regardless of which robot wins | the trade |
How to read a row → Figure (03): factory is live and BMW Leipzig is a real deployment, but it is still pre-revenue: watch whether the paid pilot turns into shipments, not the demo reel.
Who to watch
- 1X NEO deliveries: the first consumer humanoid; does it ship end-2026 and actually work in a home? (the NEO hands problem)
- Figure paid pilots: BMW Leipzig is the tell: paid industrial deployment vs another demo reel
- Tesla Optimus external customers: until there is one, it is a capex story, not a product
- Nvidia · TSMC · power & cooling: the supply chain that takes a cut of the whole physical-AI economy
- Machine payments: if robots become economic actors they need wallets, identity and settlement, crypto rails
What's broken
- Narrative outruns reality: the gap between robot videos and shipped, serviced units is the widest in tech; most humanoid startups have no real deployments
- Unit economics unproven: nobody has shown a humanoid that ships, services and retains users without costs exploding
- No external customers: even Tesla concedes Optimus is not in material factory use; announced customers, not demos, are the bar
- Forecasts are fantasy-wide: "market size" runs from ~$6B now to $200B-$5T later; underwrite the supply chain, not the endpoint
- China price floor: Unitree undercutting on price could gut Western humanoid margins before they scale
The desk methodology — what we track
Robotics is where narrative outruns reality by the widest margin, and where being early and right pays the most. The desk buys the supply chain that takes a cut of the whole physical-AI economy and treats humanoid makers as show-me stories. We track paid pilots and external customers (not demo reels); on-device reliability and service economics; and whether machine-driven volume ever needs crypto rails for identity, wallets and settlement. What would change our mind on a humanoid name: recurring paid deployments with unit economics that hold. Public picks-and-shovels core, selective private upside; track record since 2017.
follow the robotics calls on x →Related coverage
- AI × Crypto Financial Infrastructure
- Crypto × TradFi Research
- Alpha — research and thesis breakdowns