The Revolution of Humanoid Robots: When Will They Arrive in Homes and Where Do Investors See Opportunities

Humanoid robots have ceased to be merely a technological curiosity. In just a few years, they have transitioned from awkward demonstrations, where they barely began to show how they could move, work in a factory, or perform everyday tasks, to becoming one of the main long-term bets of some of the world's most important technology companies.

The advancement has been rapid. These machines can now walk, maintain balance, and perform increasingly complex tasks, while the reduction in manufacturing costs and the aging population begin to provide economic foundations for an industry that until recently seemed reserved for science fiction.

However, technological development is advancing much faster than its adoption. Significant limitations still persist in software, autonomy, safety, and manipulation capabilities that suggest their mass arrival in homes will be much slower than that of technologies like smartphones.

This difference between the potential of humanoids and the time they still need to become a mass consumer product is also key to analyzing the opportunities that the sector is beginning to generate.

Humanoids may reach homes, but not at the pace of smartphones

The incorporation of humanoid robots into everyday life appears as one of the great horizons of the technology industry, although its adoption could be much slower than current enthusiasm anticipates.

For Sergio González, Leader of the Investment Office at Cohen, the fact that humanoid robots coexist with us is not a point of discussion. The adoption of this technology, which just a few years ago seemed like science fiction, has become just a matter of time before it becomes part of our daily lives.

However, he believes that the most reasonable point of comparison to understand how quickly adoption will be does not resemble a technology like, for example, the smartphone, which had a rapid adoption growth and has become an indispensable part of our way of living in society today.

Instead, this new product resembles others that needed decades to achieve a significant presence in homes.

According to the analyst, the case of the robotic vacuum cleaner serves to illustrate this difficulty. Despite costing less than $500, performing a single task, and posing virtually no physical risk to the user, after 23 years, it has only reached between 14% and 17% of American households.

The slow adoption of robotic vacuums shows that even the most accessible technologies can take decades to reach homes on a mass scale.

Even iRobot, the company that created that market, ended up bankrupt last December. In parallel, the automobile took about 17 years to reach half of households.

Therefore, González believes that a humanoid, which is much more expensive, complex, and risky, is unlikely to have a faster adoption curve. In this way, he understands that robots will become part of daily life, "but not at smartphone speed."

Under his assumptions, with a price below $20,000, eight hours of autonomy, and a specific safety standard, the commercial escalation could begin between 2028 and 2030 in logistics and light industry and extend to service companies between 2030 and 2035.

In households, on the other hand, a low double-digit penetration would only appear towards the end of the 2030s, while surpassing half of households would not be possible before 2050 and is not guaranteed.

Thus, it is still difficult for many to imagine, even in a world where artificial intelligence and self-driving cars become part of the new normal, that there will come a point where having a replica that looks similar to oneself, and is much smarter, can coexist in a home.

However, in this process, the so-called "uncanny valley" effect, that is, the feeling of rejection generated by machines that look too similar to humans, would not be the main obstacle.

According to González, the available evidence explains only a small part of the phenomenon, and the industry itself has begun to avoid it through design. The NEO, a domestic humanoid robot developed by the American company 1X, for example, is offered covered in fabric and without a face.

Instead, the restrictions that can really determine the speed of adoption are battery autonomy, hand dexterity, and civil liability in the event of possible accidents.

Human appearance can generate rejection, but the main limits for adoption remain autonomy, dexterity, and safety.

Demographics and costs drive the trend, but technology is still not keeping pace

Behind the development of humanoids, there are three major drivers, although not all have the same degree of predictability. The most solid is the demographic, as population aging and the reduction of the workforce create a structural need to replace human tasks.

The OECD projects an 8% decline in the working-age population by 2060, while Japan could face a shortage of 570,000 caregivers by 2040. For González, this phenomenon constitutes "the anchor of the entire thesis", because it will continue regardless of technological advances or the financial cycle.

The second driver is the reduction of costs. Unitree Robotics, one of the leading Chinese manufacturers of humanoid robots, went from selling its product for around $85,000 in 2023 to $25,000 in 2025.

In turn, Goldman Sachs observed a 40% drop in manufacturing costs in just one year, compared to the 15%-20% it initially expected, and from that acceleration, it multiplied its market estimate for 2035 by six.

However, the third driver, software development, is also one of the main bottlenecks for humanoids. Although mechanical advances already allow these machines to walk and maintain balance, it is still much more difficult to enable them to use their hands to manipulate different objects and adapt to situations that were not previously programmed.

Population aging and falling costs drive humanoids.

A sample of this difficulty appears in Physical Intelligence, a company specialized in developing artificial intelligence models for robots to learn to perform different physical tasks.

Its systems achieve a success rate of only 52% in their own testing bank, showing that they are still far from executing these actions consistently.

In addition, there are physical and regulatory limits. A battery of around 2 kWh allows for approximately two hours of work, so currently, three or four robots would be needed to continuously cover the same position.

The supply chain also adds another vulnerability. China controls about 90% of the processing of neodymium magnets, an essential input for actuators. Thus, while the West seeks to develop its own robotics industry, a critical part of the components continues to depend on Chinese suppliers.

The balances show that today the value is clearer in the components

Beyond the potential of humanoids, the investment thesis still faces a high level of uncertainty. Figure AI, for example, reached a valuation of $39 billion without revealing revenues, while estimates of the market size for 2035 range between $15 billion and $38 billion.

For Sergio González, this dispersion forces a separation between the expected growth of the industry and the value that companies can justify today, as "the investor needs to know that they are buying a very wide distribution of outcomes."

In this context, the second-quarter balances show that, for now, the business appears clearer among component suppliers than among finished robot manufacturers. Sony was the company that most reinforced this thesis, with a 26% growth in image sensors, a record operating result, and a joint investment with TSMC of about $6.3 billion to produce sensors intended, among other applications, for "physical AI."

Samsung is also gaining exposure from semiconductors, driven by memory and HBM4, and received an order to manufacture Tesla's AI5 chip, linked to Optimus. However, Ballie, its domestic robot, showed little progress. Additionally, González warned that the rising cost of memory could pressure the margins of future humanoids by 2027.

Tesla in this context represents the most direct bet. The company dismantled the Model S/X line in Fremont to install the first production line for Optimus there, although initially it will be for internal use and Elon Musk defined it as "the hardest product to scale in Tesla's history". Toyota and Honda, other companies worth keeping an eye on for the future, on the other hand, contributed few innovations.{#p-1786574889265-95221}

Thus, González placed Sony as the company that most reinforced the thesis during the quarter, followed by Samsung and Tesla.{#p-1786577205094-41957}

Tesla advanced with Optimus, although the most concrete value still appears among component suppliers.{#p-1786576235484-67092}
Capture X

US Regulation Strengthens the Bet on Components {#p-1786574786559-73919}

The reading left by the balances also finds support in the new regulatory scenario. On July 28, the United States added mobile robots produced abroad to the FCC's Covered List, which prevents new models from obtaining the necessary authorization to be imported and sold in the country.{#p-1786574786559-7736}

As González explained, the measure does not constitute a tariff nor is it directed exclusively against China. The criterion used requires 65% domestic content and does not contemplate the exceptions that usually apply to allied countries. Thus, a robot produced in Japan or Germany can be excluded from the US market in the same way as a Chinese one.{#p-1786574786559-49254}

However, the restriction does not apply to components sold separately. This again favors the position of companies like Sony and Samsung, which can supply sensors, memory, and semiconductors to manufacturers required to assemble their robots within the United States.{#p-1786574786559-57943}

For González, it is a strategy similar to selling "picks and shovels": gaining exposure to the growth of the industry without depending on which company ends up developing the dominant humanoid.{#p-1786574786559-33132}

Thus, both the latest balances and the new regulatory framework point in the same direction. While manufacturers still need to resolve issues of autonomy, software, costs, and scale, component suppliers are already beginning to capture part of the industry's growth.{#p-1786574786559-57233}

Therefore, although humanoid robotics has strong long-term potential, the investment thesis today appears more consolidated in the infrastructure needed to develop it than in the finished robot.