Goldman Sachs' 6x Humanoid Revision: What $38 Billion by 2035 Means If You're Buying Automation Today

The headline, stated correctly

Goldman Sachs Research forecasts the global humanoid robot market at US$38 billion by 2035 — not 2030. That number first appeared in their February 2024 note (Goldman Sachs Research) and has since been revised upward on the unit side.

On 31 August 2026, Goldman raised its humanoid shipment forecast to 6.48 million units by 2035, up from a previous estimate of 1.38 million — roughly a 4.7x increase in units and a six-fold jump versus their earlier base case (Investing.com). The bank singled out automotive manufacturing as the first large-scale deployment domain.

If you are an operator — a restaurant group, a hospital, a factory, a franchise investor — the useful reading of that headline is not "humanoids are here." It's this: the capital markets have decided robotics is a decade-long structural trend, and component costs are going to fall because of it.

Why a 2035 forecast still matters in 2026

Three second-order effects hit buyers long before humanoids do:

1. Actuator and reducer prices fall. Humanoid demand pulls volume through the same harmonic-drive, servo, and force-sensor supply chain that collaborative arms use. Volume drives unit cost down for everyone.
2. Perception software gets cheaper and better. The vision and grasping models being trained for humanoid hands transfer directly to single-arm workcells.
3. Capital becomes available for automation projects. Robotics startups have raised more than US$23 billion globally in 2026, with humanoids alone accounting for around US$8.6 billion (Crunchbase News, reported August 2026). Lenders and landlords are far more comfortable with an automation line item than they were three years ago.

What is actually deployable this year

Here is the honest split.

| Capability | Humanoid, 2026 | Fixed / mobile Cobot, 2026 | |---|---|---| | Repeatable pick-and-place | Pilot scale | Production, thousands of sites | | Working next to people | Emerging, limited certification | ISO 10218 / ISO/TS 15066 certified today | | Total installed cost | Not publicly priced at volume | US$35k–US$140k per cell | | Payback horizon | Unproven | 12–24 months, measurable | | Service network | Vendor-direct, thin | Established integrator networks |

Humanoids in logistics are real but small: deployments at Amazon, Walmart and third-party sites remain pilot-scale across Figure, Apptronik Apollo and Agility Digit (RobotWale, August 2026). Agility has publicly described nine customer sites as it pursues a public listing (Bloomberg Technology, via RuntimeWire, August 2026). Nine sites is a pilot programme, not an installed base.

The buying rule that follows from this

Do not wait for the 2035 machine to solve a 2026 labour problem. The correct move is to automate the task you can define today with equipment that is certified today, and design the cell so the arm, the vision head and the tooling can be re-tasked as software improves.

Concretely:

• Pick one task with a measurable cycle time and a measurable defect rate.
• Buy a payload class one step above what that task needs, so the same arm covers the next two tasks.
• Insist on a normalised gripper interface and a documented safety assessment — that is what preserves the asset's value as the market moves.
• Model payback on labour, throughput, quality and insurance — not headcount alone.

How SMERT fits

Our SMERT Cobots ship today in 5, 10, 20 and 30 kg payload classes with integrated vision, ISO 10218-compliant safe-motion, and a swappable gripper interface, and our 7-axis SMERT H / HX platform covers humanoid-form tasks where they genuinely make sense. We are the OEM — assembled in Hong Kong with our own AI and computer-vision stack — so re-tasking a cell is a software conversation, not a new purchase order.

Explore the lineup and pricing: /robotics · humanoid platform: /robotics/humanoid · order with a 10% deposit: /robotics/shop · model your own payback: /tools/roi-calculator · build a territory business around it: /franchising.

Book a 30-minute deployment review.