In September 2026, Goldman Sachs sharply revised upward its forecast for the humanoid robotics market, predicting 6.5 million humanoid robots deployed by 2035 and a market size of $138 billion. The revised estimate represents a roughly fivefold increase from the bank’s earlier projections and reflects the rapid acceleration of investment, manufacturing, and policy support across the US and China. But a forecast is not a fact — it is a model output built on assumptions that deserve scrutiny, especially by the product teams, educators, and technology leaders who would need to translate these numbers into real decisions.

In Brief

  • Goldman Sachs raised its humanoid robot shipment forecast to 6.5 million units by 2035, up from approximately 1.2 million in its previous estimate — a fivefold increase
  • The bank values the 2035 market at $138 billion, with logistics, warehousing, and automotive production lines expected to be the earliest adopters at scale
  • The forecast reflects surging investment in China’s humanoid robotics sector, including XPENG’s production line commissioning and Unitree’s STAR Market listing plans
  • Goldman Sachs identifies cost reduction, AI model capability, and labor shortages as key demand drivers
  • Analyst estimates for humanoid robots vary widely — from cautious timelines of 10+ years to aggressive projections of millions of units — making it critical to understand the assumptions behind each forecast

What the Forecast Actually Says

According to reporting by 24/7 Wall St. and Investing.com on September 13, 2026, Goldman Sachs revised its humanoid robot forecast sharply upward. The bank now expects 6.5 million humanoid robots to be shipped by 2035, with a total addressable market of $138 billion. The previous forecast, issued earlier in 2026, had projected approximately 1.2 million units.

Goldman Sachs reportedly identified logistics and warehousing as the sectors most likely to drive initial scale adoption, followed by automotive manufacturing. This aligns with the pattern seen in 2026: XPENG’s IRON robot is being deployed first at the company’s own stores and campuses, while Chinese robotics companies more broadly have been targeting warehouse and factory environments for their first commercial deployments, as Reuters reported in August 2026.

The bank’s forecast also reflects the funding environment. According to The Robot Report, XPENG’s robotics unit Dogotix raised over $900 million in August 2026 at a valuation exceeding $6.3 billion. Unitree, the company behind the G1 EDU robot and the NVIDIA Isaac GR00T reference platform, is planning a STAR Market listing in Shanghai. And in September 2026, TechCrunch reported that Mecka AI, a robot training data company, was nearing a $500 million valuation in a Sequoia-led deal — evidence that the capital is flowing not just into hardware but into the data infrastructure that enables robot learning.

Why Goldman Sachs Raised the Forecast

Several factors appear to have driven the upward revision:

Manufacturing milestones. XPENG’s September 2026 announcement that its IRON humanoid robot walked off an automated production line was a first for the industry. As CnEVPost reported, more than 80% of core processes on the line are automated. This demonstrated that at least one company had moved from prototype-building to line manufacturing — a prerequisite for volume shipments.

Policy support in China. The Chinese government has designated humanoid robotics as a strategic industry. According to Chatham House, China’s humanoid robot industry has received substantial state support, though the think tank also raised questions about whether the sector represents a bubble. Goldman Sachs’s forecast appears to assume that this policy support will translate into sustained production and deployment.

Labor shortages in logistics and manufacturing. The Goldman Sachs report reportedly identifies persistent labor shortages in warehousing and manufacturing as a structural demand driver. This is a plausible claim — warehousing operators in the US and Europe have reported difficulty filling positions since the pandemic, and the logistics sector has already adopted significant automation through collaborative robots and autonomous mobile robots.

AI model advances. The rapid improvement of vision-language-action models, including NVIDIA’s GR00T N1.6, has shortened the perceived timeline for useful humanoid robot capabilities. Goldman Sachs apparently models these advances as enabling broader deployment scenarios.

The Assumptions Worth Questioning

A market forecast is only as good as its underlying assumptions. Several deserve examination.

Unit Cost Trajectory

Goldman Sachs’s forecast implies a dramatic reduction in unit cost over the next decade. Today, a single humanoid robot from companies like Figure, Unitree, or XPENG costs anywhere from $30,000 to $300,000. Reaching 6.5 million units by 2035 implies that prices must fall to a level where mass commercial deployment is economically rational — likely below $50,000 per unit for warehouse applications, and potentially lower for high-volume deployments.

This is not impossible — comparable cost curves have occurred in other hardware categories, including electric vehicles and solar panels. But those curves required sustained investment in manufacturing scale, component standardization, and supply chain maturation. Humanoid robots are at an earlier stage than either EVs or solar were when their cost declines accelerated. The Microduck at $399 demonstrates that small-scale humanoid robots can be cheap, but a 25 cm toy-scale robot is fundamentally different from a full-size humanoid capable of warehouse work.

Capability Versus Deployment

The gap between a robot that can demonstrate a task in a controlled environment and a robot that can perform that task reliably, at production speed, in a variable real-world environment is enormous. As Reuters reported in August 2026, Chinese humanoid robots can perform impressive demonstrations — kung fu, dancing, running races — but struggle with the sustained, variable work required in factory and warehouse settings.

Goldman Sachs’s forecast implies that this capability gap will close substantially within the next decade. Unitree’s founder, Wang Xingxing, told CNBC in August 2026 that humanoid robots’ “ChatGPT moment” could be 10 years away. If the person building some of the most capable humanoid robots in the world says useful general deployment is a decade out, a forecast predicting 6.5 million deployments by 2035 (nine years from now) is optimistic by implication.

ROI and the Adoption Decision

For a warehouse operator or manufacturer, the decision to deploy humanoid robots depends on return on investment. A humanoid robot must be cheaper or more productive than the alternatives — which include traditional industrial robotic arms, collaborative robots (cobots), and human labor.

The cobot market is already mature. Companies like Universal Robots have been selling collaborative robots for over a decade, with well-understood ROI models, safety certification, and integration support. A humanoid robot competing against a $25,000 cobot with proven deployment history needs to offer capabilities that a cobot cannot — primarily mobility and multi-task flexibility.

Goldman Sachs apparently models this transition as accelerating in the late 2020s. Whether the ROI math works by then depends on assumptions about labor cost inflation, robot capability improvement, and the cost of supporting infrastructure (charging, maintenance, fleet management software) that remains immature.

The Bubble Question

Not everyone agrees that the humanoid robotics market will grow as predicted. An Oregon-based robotics company, interviewed by OregonLive in August 2026, said the industry needs a reality check: “I really want to pop that bubble.” The Chatham House analysis from September 2026 similarly questioned whether China’s humanoid robot investment is sustainable at current levels.

Goldman Sachs’s forecast, by contrast, implies that the current investment wave is justified and will lead to sustained growth. The bank’s incentives should also be noted — investment banks benefit from the capital raising activity that forecasts like this stimulate. This does not make the forecast wrong, but it is a reason to read it with appropriate skepticism.

Why This Matters for Product Builders

From a product-building perspective, the Goldman Sachs forecast is useful as a directional signal, not as a precise prediction. Here is what matters:

The supply chain is maturing. Whether or not 6.5 million humanoid robots are deployed by 2035, the investment flowing into the sector is real and is building manufacturing infrastructure, component supply chains, and software platforms. For teams building robotics products — including educational robotics — this maturing supply chain means better, cheaper components and more standardized software tools.

The talent pipeline is forming. NVIDIA’s Isaac GR00T reference platform and the Hugging Face LeRobot library are making humanoid robotics research more accessible. Universities that previously could not afford to work with humanoid robots now have reference platforms and open-source tools. This is directly relevant to the STEAM education work that LearnOBots does — the pipeline of students who will build the next generation of robots is being expanded by these platforms.

The data infrastructure is being built. The Mecka AI funding and Kinetic Blocks’ marketplace for humanoid robot training data, reported by Robotics & Automation News in September 2026, signal that the industry is beginning to address one of its hardest problems: generating enough diverse, high-quality training data. This is a positive indicator for capability improvement.

Relevance for Pakistan and Emerging Markets

For Pakistani technology teams and educators, the Goldman Sachs forecast has several implications worth considering.

First, if humanoid robots achieve even a fraction of the forecasted deployment scale, the manufacturing and integration opportunities will not be limited to the US and China. Countries with existing automotive assembly, electronics manufacturing, or industrial automation capabilities could participate in the supply chain. Pakistan’s manufacturing sector has assembly capability, and the LearnOSTEAM platform is already building the educational foundation for robotics-literate graduates.

Second, the forecast underscores the importance of robotics education. If Goldman Sachs expects 6.5 million humanoid robots by 2035, the world will need engineers, technicians, and operators who can deploy, maintain, and improve them. For educators building STEAM programs in Pakistan and similar markets, this is a strong argument for investing in robotics curricula now — not because humanoid robots will appear in Pakistani warehouses next year, but because the graduates of these programs will enter a global workforce where robotics literacy is increasingly valuable.

Third, the forecast’s focus on logistics and warehousing is relevant to Pakistan, where the logistics sector is growing rapidly and faces many of the same labor cost and availability pressures seen globally. While humanoid robots may not be the first automation solution adopted in Pakistani warehouses — traditional automation and cobots are more likely — the broader trend toward warehouse automation is one that Pakistani logistics companies and technology providers should be watching.

What to Watch Next

  • Goldman Sachs’s next revision. Track whether the bank raises or lowers its estimate over the next 12–18 months. A downward revision would signal that the initial hype cycle is cooling.
  • Unit cost disclosures. Watch for humanoid robot pricing from XPENG, Unitree, Figure, and others. If prices approach $50,000 for capable units, the forecast’s volume assumptions become more credible.
  • Real-world deployment results. Initial deployments at XPENG’s stores, Amazon warehouses (where Figure robots are being tested), and similar sites will provide the first independent evidence of whether humanoid robots can perform useful work at production speed.
  • Cobot market response. Companies like Universal Robots, which just unveiled its Gen 7 platform at IMTS 2026 according to Automation World, are not standing still. If cobots absorb the near-term demand for warehouse and factory automation, humanoid robots may need to wait longer for their market.
  • Regulatory and safety frameworks. Volume deployment of humanoid robots in human environments will require safety standards that do not yet exist. Watch for progress from ISO, ANSI, and Chinese regulators.
  • Pakistan and South Asia robotics policy. Monitor whether Pakistan, India, or other South Asian markets develop specific policies or incentives for robotics adoption or manufacturing. Karnataka’s decision to introduce STEAM education in government schools including robotics and coding, reported in June 2026, is an early signal from the region.

Product Builder’s Perspective

The Goldman Sachs forecast is a useful planning tool if treated as a scenario, not a certainty. For product teams, the key question is not “will there be 6.5 million humanoid robots by 2035?” but “what should we build if there are?”

The answer, from a product-building perspective, is to invest in the layers that benefit regardless of the exact unit count: robot training data infrastructure, fleet management software, safety and compliance frameworks, and the educational pipelines that produce robotics-literate talent. These are the picks-and-shovels of the humanoid robotics gold rush, and they are less sensitive to the exact timing of volume adoption than the robot hardware itself.

For educators, the signal is clearer: robotics education is not speculative. The investment, the manufacturing infrastructure, and the policy support are all real and growing. Whether the 2035 market is $138 billion or half that, the graduates who understand robotics, AI, and physical computing will have opportunities.

Conclusion

Goldman Sachs’s revised forecast is a meaningful data point — one of the largest investment banks in the world is betting that humanoid robotics will become a significant market within a decade. The forecast reflects real progress in manufacturing, AI capabilities, and capital availability. But it also rests on assumptions about cost reduction, capability improvement, and ROI that remain unproven at scale.

For builders, educators, and technology teams, the right response is neither blind optimism nor dismissive skepticism. It is to invest in the infrastructure, education, and tools that will be needed if humanoid robotics grows — while watching the evidence of real-world deployment closely enough to adjust course when the data arrives.

The 6.5 million number may prove prescient, optimistic, or conservative. What is certain is that the humanoid robotics industry is no longer a research curiosity. It is a manufacturing industry with billions of dollars of investment, production lines being commissioned, and a maturing supply chain. The next decade will determine whether that industry becomes transformative — or whether it becomes the next category to fall short of its own hype.

What would it take for you to deploy a humanoid robot in your operation — and what is stopping you today? That question, answered honestly by the people who would actually use these robots, is more useful than any forecast.

Sources

Shamyl Bin Mansoor

Shamyl Bin Mansoor is a technology entrepreneur, product builder and STEAM educator. He is Co-founder and CTO of LearnOBots and works on educational technology, robotics and interactive learning products. Learn more about him.

Related Articles

XPENG's IRON Humanoid Robot Walks Off the Production Line: What It Means for Robotics Manufacturing

XPENG has commissioned the world's first automated production line for advanced humanoid robots, with its IRON robot walking off the line autonomously. Here is what this means for the humanoid robotics industry, product builders, and educators.

World Labs Atlas: What Spatial Intelligence World Models Mean for Robotics, Simulation, and Education

World Labs has introduced Atlas, a multimodal world model that generates, reconstructs, and simulates 3D environments from sparse images. Here is what it means for robotics teams, simulation builders, and educators.

Humanoid Robot Security: What the Unitree G1 EDU Vulnerabilities Mean for Robotics Builders and Educators

Two root remote code execution vulnerabilities in the Unitree G1 EDU humanoid robot — one exploitable over Bluetooth — reveal how security is becoming a critical concern for educational and research robotics platforms.