Gasgoo Munich-PEEK (polyether ether ketone) is a specialty engineering plastic characterized by high-temperature resistance, self-lubricating properties, and high specific strength. Thanks to these attributes, PEEK has been widely hailed over the past two years as a "master key" to overcoming bottlenecks in robotic motion performance; it has garnered intense interest across the industry supply chain, and related market concepts have repeatedly surged in popularity.
Yu Fei, General Manager of Robotics Technology at Kingfa Tech., offers a rational assessment based on the realities of industrial implementation: regarding PEEK applications in humanoid robotics and embodied AI, the hype surrounding the concept has significantly outpaced its actual practical value. Scalable application scenarios have yet to materialize, and the industry harbors overly optimistic market expectations.
Constrained by the limitations of overall mass production capacity, humanoid robotics currently faces multiple hurdles regarding materials—including a lack of standards, a reliance on highly customized solutions, and the ongoing tension between cost and reliability. The industry is currently at a critical juncture, struggling to transition from small-batch prototyping to large-scale mass production.
Yu Fei, Technical General Manager of the Robotics Industry at Kingfa Tech
Humanoid Robot Materials: 80%–90% Highly Customized
According to the latest statistics from Counterpoint Research, global humanoid robot shipments have exceeded 22,000 units in the first half of 2026, an increase of nearly 300% compared with the same period in 2025, and continue to maintain rapid growth.
Based on the current situation, Gasgoo partner and vice president of the research institute Wang Xianbin predicts that global humanoid robot shipments will reach 60,000 units this year, of which about 50,000 units will be in the Chinese market, an increase of more than 2 times from 18,000 units in 2025.
Yu Fei is optimistic. He believes that the overall shipment of humanoid robots is expected to reach about 100,000 units this year and further increase to 200,000-300,000 units next year. The industry may enter a "concentrated explosion window" from 2028 to 2030.
"In the second half of this year, according to my observation, many projects are being implemented, the market situation is much better than in the first half of the year, and the overall installed capacity will increase significantly." Yu Fei said.
However, from the perspective of enterprise implementation, there are still obvious shortcomings in industry maturity.
According to Yu Fei, there are currently close to 400 domestic companies involved in humanoid robots, but only a handful of manufacturers can truly achieve an annual production capacity of over 1,000 units. Except for a few companies such as Yushu, Zhiyuan, and Youbi, the vast majority are still in the laboratory iteration or small batch trial production stage.
It is this kind of industry maturity that has caused the material side to fall into the double dilemma of "high customization" and "standard blank".
First of all, the small scale of downstream mass production means that upstream demand will be extremely fragmented and non-standardized.
"Currently, 80%–90% of humanoid robot material solutions are highly customized, making reuse across customers extremely difficult," Yu Fei pointed out.
For example, even if two different complete machine manufacturers target the same joint parts, due to different structural designs, load conditions, life requirements, etc., the corresponding material formulas are also different, making it difficult to provide one solution to multiple customers at the same time.
Especially at this stage, all ontology manufacturers generally tend to develop their own proprietary structures, and seldom consider material versatility in the front-end design stage. Upstream suppliers can often only follow the structure for customized development, and cannot form a standardized supply system.
Image source: Beijing E-Town
Secondly, the absence of industry standards acts as another constraint.
Unlike mature sectors such as home appliances and automotive manufacturing—which possess relatively comprehensive systems of industry and corporate standards—humanoid robotics is an emerging industry that has yet to establish unified standards.
In the absence of standards, manufacturers of complete units are often forced to test multiple solutions in parallel.
"Clients might provide anywhere from one to five material options for testing at once; we then have to determine whether a failure stems from the structural design, the material itself, or a combination of both," explains Yu Fei.
Consequently, the process—from prototyping to finalizing the design for mass production—often requires three to five rounds of iterative adjustments to both material formulations and the unit's structure before a solution is locked in.
Crucially, these two issues reinforce each other: the lack of standards drives bespoke customization, while such customization makes it even harder to establish standards. Ultimately, this drives up both the development timeline and costs across the entire value chain.
According to Yu Fei, a single humanoid robot currently requires approximately 6–10 kilograms of plastic. Based on market prices for modified plastics, the material cost per robot exceeds 1,000 RMB.
However, because order volumes from individual clients are small and many products require unique formulation development and validation, it is difficult to effectively amortize the per-unit development costs.
So, how can the industry break free from the high-cost trap associated with customization?
In Yu Fei's view, the key to a breakthrough lies not with material suppliers acting alone, but with a significant scaling up of downstream robot manufacturing.
"As the industry scales up in the future, manufacturing methods will gradually shift from CNC machining to mold opening and injection molding, causing supply chain and material costs to decline." Cost reduction through economies of scale is an inevitable outcome of mass-market expansion.
PEEK offers outstanding performance advantages, but large-scale implementation remains immature.
Over the past year, the hottest buzzword in the humanoid robotics materials sector has undoubtedly been PEEK.
With a density of approximately 1.3 g/cm³—roughly half that of aluminum alloy—PEEK boasts properties such as self-lubrication, wear resistance, high-temperature tolerance, and a high strength-to-weight ratio. It perfectly meets the multifaceted requirements—lightweight, wear-resistant, and heat-resistant—of critical components like robot gears, speed reducers, bearings, and dexterous hands.
Image source: Tinci Materials
Consequently, many humanoid robot OEMs—such as Tesla, Unitree, and Agibot—have begun incorporating PEEK materials into their products.
The supporting supply chain is keeping pace; alongside material suppliers like Kingfa Tech, Wote, Youju New Materials, and Tinci Materials, component manufacturers such as Chaojie, Eoplex (Changying Precision), and Ningbo Huaxiang are also accelerating their strategic positioning in this sector.
For instance, Huaxiang Qiyuan has independently developed a planetary gear reducer made of modified PEEK. With the exception of core metal transmission components, the main body utilizes modified PEEK; the assembly weighs just 270 grams—a reduction of approximately 70% compared to metal reducers of the same specifications. By applying PEEK structural components throughout the unit, Huaxiang Qiyuan has achieved a total weight reduction of 5.3 kilograms per humanoid robot.
Forecasts suggest that, driven by industrial upgrades and the trend toward lightweighting, the global PEEK market will grow rapidly from its 2024 valuation of 6.1 billion yuan, with the humanoid robot sector alone projected to generate demand exceeding 4 billion yuan by 2027.
Yu Fei maintains a rational and prudent stance regarding the industry's intense enthusiasm for PEEK.
His assessment is based on two main factors.
First, the gap in performance suitability.
"There are significant differences in dimensional characteristics between PEEK and metal, and a difference of at least two orders of magnitude in fatigue performance," Yu Fei noted.
Fatigue performance refers to a material's ability to withstand millions of stress cycles without failure; since robot joints require high-frequency reciprocating motion, this limitation is difficult to circumvent.
Second, the cost-benefit mismatch.
Often dubbed "plastic gold," PEEK currently commands a price of up to 1,000 yuan per kilogram, placing it at the pinnacle of the materials hierarchy. This implies that, in current application scenarios, using PEEK not only raises costs for customers but does not necessarily yield significant benefits.
"Therefore, when serving customers, we prefer to recommend comprehensive material solutions that offer better suitability and cost-effectiveness, tailored to the specific operating conditions and mass production requirements of the enterprise," Yu Fei stated.
Despite his cautious outlook on the short-term application of PEEK, Kingfa has not halted its investment in the technical development of PEEK materials. If the structural design of humanoid robots undergoes a major overhaul in the future—aligning material properties closely with specific operational demands—PEEK is poised to enter a critical window for large-scale adoption.
Automotive Experience Offers Lessons but Cannot Be Directly Replicated
Amid the current humanoid robot boom, the trend of companies from other sectors—particularly the automotive industry—entering the field is a key topic, with the convergence of automotive supply chains and embodied AI warranting special attention.
"Over 90% of the OEMs and Tier 1 suppliers I've engaged with are positioning themselves in the embodied AI space," notes Yu Fei. "Some have made a complete pivot, abandoning their automotive businesses entirely, while others have made more modest adjustments but are still establishing a foothold in this sector."
As a global leader in modified plastics, Kingfa Tech. has itself benefited significantly from the automotive supply chain. In the first half of 2026, the company sold 1.4744 million tons of modified plastic products—a year-on-year increase of 12.65%—and generated 18.167 billion yuan in revenue, up 10.28%. Notably, global sales of automotive-grade materials reached 616,700 tons, a rise of 10.07%.
The logic driving this influx of automotive supply chain companies into embodied AI is consistent: smart vehicles and embodied robots share a common lineage in terms of technical architecture, supply chain systems, and engineering methodologies. Both rely on modules such as sensors, computing platforms, motors, and batteries, creating opportunities for cross-application across multiple dimensions.
"For instance, existing experience and methodologies from the automotive industry, as well as the established supply chain itself, offer areas for reuse," Yu Fei explains.
However, the materials themselves cannot simply be transferred over.
Yu Fei estimates that the overlap in material types between automotive applications and embodied robots is roughly 20% to 30%.
The underlying reason for this lies in the fundamental differences in operating conditions. Automotive applications typically involve low-frequency usage, whereas humanoid robots operate under high-frequency conditions. Their joints require continuous, rapid oscillation, placing higher demands on strength, impact resistance, and thermal conductivity—requirements that necessitate distinctly different materials.
For example, while the automotive sector makes extensive use of standard polyamides like PA6 and PA66, humanoid robots rely more heavily on higher-grade specialty engineering plastics such as LCP, PPS, and PEEK. "Even for the 20%–30% of material categories that overlap, direct reuse is quite difficult because the operational dynamics differ significantly," Yu Fei points out. In other words, this kind of reuse isn't a simple matter of copying and pasting; it requires a "two-way convergence" of improved material performance and optimized product structure.
Lightweighting is a classic example of something that "looks easy to copy but is actually hard to replicate."
In the automotive industry, lightweighting has been thoroughly proven as a way to extend vehicle range; today, that same logic is playing out in the realm of humanoid robots. Yet, even with the benefit of automotive industry experience, Yu Fei believes the path to lightweighting humanoid robots remains a long and challenging one.
The reason is simple: lightweighting is not merely a matter of swapping materials—such as switching directly from steel or aluminum-magnesium alloys to plastics—but involves redesigning the entire product structure. Currently, the industry—including both OEMs and material suppliers—lacks a sufficient body of accumulated engineering case studies.
"So, the general direction of lightweighting is certainly sound; it's just that the specific pathways for implementation and execution will take another three to five years to figure out—after all, much of the existing experience cannot simply be copied over. While large models and AI can certainly accelerate progress, the necessary trial-and-error processes cannot be skipped," Yu Fei says.
It is also worth noting that, beyond the reuse of technology and experience, the question of whether humanoid robots should directly adopt mature automotive-grade standards is a hot topic in the industry.
Yu Fei's assessment is that, in the short term, directly applying the mature automotive-grade system is the right general direction and carries the lowest risk, given that these standards have undergone extensive validation and offer proven reliability.
However, in the long run, automobiles and robots are distinct industries; automotive-grade specifications may result in performance redundancy for robots and offer limited suitability.
Therefore, a more realistic approach is to "adopt first, then optimize," though the industry will ultimately need to establish its own dedicated standards for humanoid robot materials and products.
Four Hurdles Stalling the Mass Production of Humanoid Robots
In the field of materials for humanoid robots, the challenges—whether regarding the difficulties of custom development or the limitations of PEEK applications—all ultimately point to the practical hurdles inherent in developing humanoid hardware.
Image source: Kingfa Tech
The first hurdle is weight. Currently, the structures of humanoid robots are predominantly made of metals such as magnesium and aluminum alloys; while these offer reliable load-bearing capabilities and mature manufacturing processes, their density is higher than that of engineering plastics, keeping the robot's overall weight high.
Weight reduction is not only crucial for enabling robots to walk more stably, run faster, and carry heavier loads, but for industrial robots, it also translates to faster cycle times and higher production efficiency.
However, as previously mentioned, "replacing steel with plastic" is not merely a matter of swapping materials; it requires a comprehensive redesign of the robot's structure.
The second hurdle is heat dissipation.
A prime example of this issue occurred during a humanoid robot marathon in Beijing's Yizhuang district, where many robots required dry ice cooling halfway through the race.
High-power joints engage in continuous, high-frequency movement, concentrating heat within the compact body. Heat dissipation remains a common weakness across the industry, and material thermal conductivity is a key factor in solving this problem.
The third hurdle is cost.
The humanoid robot industry is still in its infancy, with limited development experience, meaning structural designs require multiple iterations before being finalized. Particularly under current small-batch production models, the development costs for components and materials cannot be amortized through scale, further compounding the financial pressure associated with customization.
The fourth hurdle is standardization.
There is a lack of industry and corporate standards. Without a unified testing and evaluation system, defining material performance and determining compliance requires OEMs and material suppliers to coordinate on a project-by-project basis, which in turn exacerbates the burden of customization.
Kingfa is actively positioning itself to address these challenges.
Kingfa Tech. has established a dedicated R&D project team for robotics materials and is collaborating with leading companies; some of its material products have already entered mass supply.
Yu Fei emphasized that Kingfa's core competitiveness lies not merely in selling materials, but in providing comprehensive solutions that cover everything from material selection, structural design, and mold making to final molding.
For instance, in response to the "four trends" in humanoid robotics—lightweighting, aesthetic appeal, temperature resistance, and safety/intelligence—Kingfa is advancing the "steel-to-plastic" transition using a "top-down" approach. This means prioritizing the replacement of non-load-bearing components such as faceplates, upper arms, and chest housings first. Regarding structural components that bear the load of the torso and lower body, Yu Fei frankly admits that "the global industry has not yet achieved large-scale implementation of plastic-replacing-steel solutions; this remains a key area for overcoming core technical challenges."
In terms of thermal management, the high-thermal-conductivity materials developed by Kingfa have achieved a thermal conductivity of 15 W/(mK). When combined with optimized structural design for the robot itself, this "can basically resolve most heat dissipation issues."
However, beneath the potential of this massive growth market lie tangible risks for material suppliers.
On one hand, the humanoid robot business represents a forward-looking strategic move for Kingfa; it is unlikely to generate significant revenue growth in the short term. On the other hand, while the supply side of the industry is booming, downstream robot manufacturing has yet to cross the threshold for mass production, and technical pathways have not yet converged. Consequently, material suppliers cannot simply apply the standardized mass-production logic used in traditional industries.
Therefore, even if material suppliers are fully prepared with production capacity and formulations, there remains considerable uncertainty regarding whether this will ultimately translate into actual orders.
Conclusion
The hype surrounding PEEK reflects the clash between ideals and reality in the humanoid robot materials sector: the market often hopes for a "universal material" to provide an instant breakthrough, yet industrial implementation is never that simple.
While PEEK boasts impressive performance, it suffers from drawbacks related to cost and fatigue resistance. Although engineering expertise from the automotive supply chain offers valuable lessons, material formulations cannot simply be copied directly. Furthermore, achieving industry-wide cost reductions, efficiency gains, and technological breakthroughs is not something material companies can accomplish in isolation; success ultimately depends on the downstream robot industry achieving mass-market scale and technical standardization.
Thus, before the industry truly scales up, upstream material companies should focus on patiently refining cost-effective solutions tailored to real-world operating conditions—a strategy that offers the most reliable path to steady growth.
