Goldman shares latest updates on Tesla’s Optimus project

Goldman Sachs reports Tesla is making meaningful advances on its Optimus humanoid robot—notably on complex components like hands and forearms—while emphasizing improvements in reliability, manufacturability and AI training, and plans to deploy the robot internally before broader commercialization. Goldman cautions that commercialization and large-scale adoption depend on software validation, lower hardware costs and proven real-world reliability, with production likely to begin gradually and scale later in the decade.

Analysis

Tesla’s push into humanoid robotics, pursued with a vertically integrated engineering model, creates asymmetric pressure across the ecosystem: capital-intensive compute and semiconductor suppliers stand to capture the bulk of near-term value while many niche mechanical suppliers face margin compression or consolidation. Expect a re-rating where system-level compute winners (GPU/accelerator vendors) expand gross margins faster than traditional actuator/gearbox vendors over a 12–36 month window, driven by software-defined value capture.

Commercialization economics are the real gate: for widespread factory adoption the all-in unit cost must fall into a low five-figure range and validation overheads must drop by multiples — that implies multi-year scale manufacturing and simulated training pipelines that consume billions in datacenter cycles. If compute remains the dominant incremental cost, incremental robot deployments will correlate strongly with enterprise AI spend cycles and datacenter capex, making robotics adoption lumpy and highly sensitive to macro capex trends.

Key tail risks are non-linear: a single high-profile safety incident, IP litigation over motion capture or control stacks, or a sustained shortage in custom precision motors could pause deployments and crater sentiment within weeks. Conversely, an early internal productivity win announced with quantified labor-replacement metrics (e.g., % reduction in cycle time or headcount) would materially shorten the adoption curve and re-rate the stock within a quarter.

The consensus is underweight operational drag on core automotive margins and overindexes to headline potential rather than unit economics; the market is likely to punish missed internal KPIs more than it rewards prototype milestones. That asymmetry favors tactical option structures that capture upside from durable compute demand while limiting exposure to tech execution risk in robotics hardware scale-up.

Key Decisions for Investors

• Long NVDA (6–12 months): buy NVDA shares or 6–12 month call spread (buy 1× ATM call, sell a higher strike) sized 2–4% portfolio to play sustained datacenter/accelerator demand from robotics training. Reward: exposure to high-margin compute growth; Risk: GPU cycle down 25%+ if macro capex slows — cap position with a 20% stop-loss or sell-to-close on 10–15% gain to trim volatility.
• LEAP-style bullish on TSLA (12–36 months): buy TSLA Jan 2027 LEAP calls funded by selling shorter-dated calls (calendar/diagonal) to manage theta. Position size 1–3% of portfolio; R/R asymmetric — large upside if commercialization converts to revenue, limited carry from sold calls if near-term execution disappoints.
• Pair trade (immediate): long NVDA / short ROBO ETF (or selected small-cap robotics stocks) 1:1 notional for 3–9 months. Rationale: capture divergence between compute winners and valuation-rich hardware/robotics names that may underperform during early internal deployments. Risk: broad sector rally lifts both legs — hedge by sizing so max drawdown on pair is ~2–3% of portfolio.
• Event hedge (30–90 days): buy TSLA tail protection (OTM puts) sized to cover 3–5% portfolio exposure in case of a safety/validation incident or negative regulatory news that could cause >20% downside. Take profit or roll if implied volatility spikes post-news.