Inside the Hype: Humanoid Robots in the Spotlight

I walked into the Humanoids Summit last year expecting the usual mix of venture bravado and sleek product demos. What I heard, instead, were engineers openly urging restraint. A founder compared humanoid robots to the Apple Newton. A promising idea, sure — just not one ready for prime time.

That tension—between marketing optimism and engineering realism—defines the state of humanoid robots today. While newsfeeds fill with polished videos of human-shaped machines moving with uncanny precision, the story on the factory floor is different.

Humanoid robots are attracting significant attention from some of the world’s biggest companies. Amazon, Mercedes-Benz, BMW, and Tesla are all publicly engaged in pilots or internal development. Figure AI, Agility Robotics, and Apptronik are just some of the companies benefiting from a capital influx that surpassed $5 billion in 2024, according to industry reporting from the Wall Street Journal.

Yet, beyond the labs and pilots, these machines have yet to deliver consistently in uncontrolled environments.

What the Numbers Actually Say

With a projection that up to one million of the humanoid robots could be deployed worldwide by 2035, the revised market assessment by Goldman Sachs estimates almost $38 billion for the humanoid robot market, showing a forecast extended almost to 1.4 million units of shipments.

Most robots currently in deployment are performing rudimentary tasks in highly structured settings. Amazon’s trial with Agility Robotics’ Digit, for example, involves the robot walking upright and transporting totes across a designated warehouse space. The robot looks human in shape but is supervised closely. Full autonomy is not yet viable.

At Mercedes-Benz facilities, Apptronik’s Apollo is undergoing limited testing, mostly focused on moving assembly kits and inspecting components at its Berlin-Marienfelde campus. These are narrow applications aimed at integration into existing workflows.

BMW’s partnership with Figure AI began with teleoperation via VR. In 2025, the company completed a pilot at its Spartanburg plant, where Figure 02 robots handled over 90,000 parts across 1,250 hours with a 99% task success rate. While the systems are still supervised, autonomy has advanced in specific roles such as sheet metal handling.

Tesla’s approach, via the Optimus robot, is more tightly held. CEO Elon Musk has claimed production could reach one million units annually by 2030. At this stage, Optimus remains in internal research and testing within Tesla facilities, with no verifiable commercial deployment.

Function Over Form: The Design Dilemma

The choice to build robots in a humanoid form raises more questions than it answers. Engineers have long acknowledged the complexities of bipedal locomotion, dexterous hands, and touch-sensitive feedback systems. Walking like a human is a hard problem. Gripping objects like a human is harder still.

Some insiders have begun to question whether mimicking human form is useful at all. Alternatives like wheeled bases, robotic arms, or task-specific manipulators may offer more dependable performance in the short term. But they don’t capture the imagination the way a humanoid does. And in a sector driven partly by perception, that matters.

Why Global Brands Are Still Investing

Despite the limitations, investment continues. This isn’t driven by novelty alone. There are structural pressures behind the bets being made.

Labour shortages across manufacturing hubs, increasing demand for reshoring production, and rising wage floors in developed markets are all pushing brands toward automation. Add to that the acceleration of AI capabilities — particularly in perception, movement, and task planning — and the promise of humanoid robots becomes more attractive.

Strategic partnerships between brands and robotics firms are not just tech trials; they’re insurance policies for a future workforce. What’s being tested today may become part of global operations in a decade, assuming the tech matures.

Real-World Use Is Still Controlled and Experimental

Despite big claims and bigger funding rounds, the reality on the ground is cautious. Robots deployed today are doing the equivalent of internships: highly guided, with minimal room for failure.

Every robot working in a factory today is accompanied by a detailed list of restrictions. Movement is confined. Objects are standardised. Safety systems are layered in. Industry analysts note that the cost of deploying a single humanoid robot is mostly absorbed not in the hardware but in infrastructure, software integration, and human oversight.

In most pilot programmes, the robot’s autonomy is closer to semi-scripted performance than general-purpose intelligence. Engineers report spending months training robots to complete what might seem like trivial tasks — opening a door, placing a box, or navigating a corner.

Signals From Inside the Industry

Industry voices remain candid about the limitations. During the Humanoids Summit, CTOs, founders, and senior engineers alike acknowledged that timelines for true autonomy remain uncertain. One engineer remarked, “The hardware is improving faster than our ability to make it useful.”

Nicolaus Radford, founder of Persona AI and a veteran of NASA, has consistently stated that while the physical systems can now be built, fine manipulation and economically viable use cases remain a major hurdle.

At Weave Robotics, Kaan Dogrusoz, whose company is developing a domestic robot named Isaac, notes that balancing feature sets with affordability is the toughest challenge. A working demo, he says, is far from a shippable product.

What the Industry Is Actually Doing

Global brands aren’t rolling these machines out across operations. They’re studying. Piloting. Gathering data.

Amazon has committed to ongoing testing in logistics. BMW and Mercedes-Benz are focusing on narrow task assignments that integrate with existing workflows. Tesla continues its internal development under tight control. These are not deployments in the traditional sense. They are live research environments.

The goal is to understand what humanoid robots can actually do — and, equally important, what they cannot.

What’s emerging is a clearer picture of the limitations:

• Robots are not yet reliable in unpredictable environments.
• Tasks often need to be redesigned for the robot, not the other way around.
• Safety regulations and operational standards for humanoid robots are still in the early stages of development.

The industry is still in a pre-adoption phase. Investment continues because the long-term potential remains compelling, but nobody working closely with the technology is suggesting that broad deployment is imminent.

What’s Worth Watching Next

The clearest sign of progress will be robots functioning autonomously for extended periods without constant supervision. If a machine can complete a shift in a warehouse or factory without human intervention, that would mark a milestone.

Other indicators include the publication of ROI studies by brands like BMW or Amazon, the appearance of third-party safety standards, and hardware breakthroughs that simplify core actions like gripping, walking, or sensing objects in cluttered environments.

Until then, humanoid robots remain a bold experiment — well-funded, strategically aligned, and far from resolved.