Goldman Sachs raises its 2035 humanoid forecast to 6.5 million units and $138bn

Goldman Sachs has raised its long-range humanoid robot forecast to about 6.5 million units (6.48 million) in 2035, up from a prior 1.38–1.4 million, and now sizes the 2035 market at $138 billion rather than $38 billion, according to 24/7 Wall St. coverage of its 80-page Physical AI research report, published September 13 and 14, 2026. The unit figure is roughly 4.7 times the old one.

Why it matters: Goldman’s February 2024 figure of $38 billion was one of the most widely cited humanoid market sizes. The revision lifts that reference point, but it remains a bank forecast, not evidence of demand.

What exactly did Goldman change?

Per 24/7 Wall St. and humanoid.guide, the revised path runs in three steps: 75,000 units in 2026 (previously 51,000), 890,000 in 2030 (previously 256,000) and 6.48 million in 2035 (previously 1.38 million). The 2035 market value moves from $38 billion to $138 billion, an increase of about 3.6 times. Units rise faster than dollars, which implies a lower average value per robot than the earlier model assumed.

The 2030 number is about 3.5 times the old one, and the 2026 number is about 47% higher. The near-term figure is the first one that can be checked: 75,000 is a full-year 2026 estimate, so shipment counts published by vendors and trackers over the coming months will show whether the ramp is on the path Goldman describes.

The full report is available to Goldman clients; the figures here come from published coverage of it, not from the report itself.

How does it compare with Goldman’s 2024 forecast?

In a Goldman Sachs Research note dated February 27, 2024, the bank projected a $38 billion humanoid market and 1.4 million unit shipments by 2035, up from an earlier $6 billion estimate, and more than 250,000 shipments by 2030. At the time it cited faster-than-expected AI progress and falling costs: its analysts said manufacturing costs had dropped about 40% in a year, with unit costs falling from a $50,000–$250,000 range to $30,000–$150,000. The 2026 revision is therefore the second large upgrade in about two and a half years.

Which industries does Goldman expect to adopt humanoids first?

Coverage of the report names logistics and warehousing as the first large-scale adopters, with automotive manufacturing next. 24/7 Wall St. also reports Goldman’s estimate that each humanoid carries from $3,000 to more than $6,000 of semiconductor content, which is why much of the coverage frames the forecast as a supplier story rather than a bet on individual robot makers.

What the forecast does not show is committed demand. As humanoid.guide notes, the projection is a high-growth scenario that depends on technical and manufacturing progress, and no published order book approaches these volumes. For comparison with where capital and IP are concentrated today, see our analysis of physical AI funding in H1 2026 and the LexisNexis humanoid patent ranking.