AI robots could cost $13,000 by 2035: Here’s what that means for CFOs

Key number: Bank of America Institute projects humanoid robot material costs falling from $35,000 in 2025 to $13,000–$17,000 by 2035. Deloitte's CFO Guide highlights 'physical AI' and agentic AI as forces reshaping manufacturing, warehouses and supply chains, driving increased hardware investment, higher energy consumption and greater capital intensity. CFOs should update KPIs, strengthen ROI measurement for hybrid human–AI workforces and invest in upskilling finance teams to capture operational benefits and manage new cost structures.

Analysis

Embedding decision-making into physical systems flips several accounting and operating levers simultaneously: expect corporate capex-to-revenue to rise meaningfully in adopters and gross margins to improve after a 1–3 year payback window, but with higher fixed-cost leverage (depreciation + energy). That pushes capital structure risk into the forefront — firms that previously optimized for variable labor will show more volatile cashflow-to-capex profiles, altering covenant stress probabilities and making vendor financing/leasing a likely growth product line.

Supply-side concentration will create asymmetric bottlenecks: a handful of suppliers for precision actuators, vision sensors, and AI accelerators will gain sustained pricing power, while low-margin contract manufacturers and labor arbitrage geographies face structural margin compression. Energy delivery and resilience (on-site storage, microgrids) become a material cost and hidden capex line for factories deploying embodied AI; utility tariffs and electrification policies will therefore be a second-order determinant of ROI.

Regulation, safety incidents, or export controls on advanced chips are the highest probability reversal events over 6–24 months; a single high-profile accident or a tightening of export policy could stall deployments and create impairment cycles. Adoption will be lumpy — rapid in high-value, low-mix industries (auto, logistics) within 2–4 years, slower in high-mix consumer goods (5–10 years) — so position sizing and horizon must match that dispersion.

Key Decisions for Investors

• Long NVDA (NVDA) — buy 12-month at-the-money calls sized as a 1–1.5% portfolio position (or outright shares for lower gamma). Rationale: asymmetric upside from sustained demand for on-prem AI accelerators; downside limited to premium (expect 2–3x upside if adoption curve stays intact). Monitor short-interest and any regulatory guidance on chip exports as main catalyst/risk.
• Pair trade: Long ASML (ASML) / Short Flex (FLEX) — equal notional, 12–24 month horizon. ASML benefits from structural demand for advanced patterning for sensors/accelerators; FLEX is exposed to low-margin EMS work that is most susceptible to reshoring and automation. Target a 200–400 bps outperformance; stop-loss if semicap cycle deteriorates across the board.
• Long Rockwell Automation (ROK) or buy 9–12 month calls — 0.75–1% portfolio. Play systems integrators who capture recurring software/maintenance revenues as factories retrofit; hedge execution risk by shorting Jabil (JBL) 0.5% as a proxy for contract manufacturers losing share to automation.
• Tactical options: Buy 6–12 month calls on Cognex (CGNX) sized as a 0.5% asymmetric bet. Machine-vision wins early in humanoid/precision tasks; this is a higher-volatility, higher-gamma play — cap loss to premium but target 3x on successful adoption signals (large auto/logistics RFPs announced).