In the context of global technology giants competing to lay out a new round of strategic highlands, the robot track is becoming the most concerned technological frontier after the AI model. On June 12, 2025, a blockbuster news that detonated the industry quickly swept the screen in the field of semiconductors and smart manufacturing: Skild AI, a California-based AI robotics start-up, received more than $100 million in Series B financing led by the second phase of the SoftBank Vision Fund, including $25 million from Nvidia and $10 million from Samsung Electronics. Skild AI's latest valuation reached about $4.5 billion. This not only highlights the market's high recognition of the prospect of AI + robot integration, but also indicates that a technology and ecological "card war" around the future of "general robots" is accelerating.
Ⅰ NVIDIA: Bet on "physical AI" and lay out a new ecosystem of robots
As a leader in AI chips, NVIDIA's investment is not an isolated incident, but a continuation of its "Physical AI" strategy. The so-called "physical AI" refers to the extension of AI algorithms from the virtual world to the real world, so that robots, self-driving cars, and other systems can perceive, understand, and operate autonomously in the physical space. This echoes NVIDIA's layout of Jetson embedded platform, Isaac robot simulator, and AI large model in embedded inference in recent years.
According to market research firm Tractica (acquired by Guidehouse Insights), the global robotics and unmanned systems market is expected to grow from $21 billion in 2019 to $71 billion in 2025, with a compound annual growth rate of more than 22%. This means that robots have become an important incremental market for AI chip manufacturers.
The core of Skild AI's technology is to provide "human-like perception" and "efficient decision-making" for robots Its R&D team comes from leading companies such as OpenAI, Waymo, and Boston Dynamics, and has solid capabilities in deep reinforcement learning, visual-language joint modeling, and multimodal sensor fusion. NVIDIA's strategic intention to invest in Skild AI is obviously not only about financial returns, but also to incorporate it into its own AI robotics ecosystem, complementing the closed loop from GPU hardware, simulation platforms to high-level decision-making software. Through this vertical integration, NVIDIA is expected to create an industrial closed loop of "from algorithms to chips to robots", and occupy the initiative in the future era of AI robots.
Pictured: Skild AI, an AI robotics start-up, has received investment from Nvidia, Samsung, and others
Ⅱ Samsung: The dual abacus of technical intelligence and ecological synergy
As one of the world's largest semiconductor manufacturers, Samsung Electronics' investment in Skild AI is small ($10 million), but its strategic implications cannot be ignored.
On the one hand, Samsung has continued to explore in the field of robotics in recent years, including the launch of service robots "Bot Care", "Bot Handy" and other products, but the overall pace is much slower than that of American technology companies such as Amazon and Google. In this context, through strategic investment in cutting-edge companies such as Skild AI, Samsung can "bind" cutting-edge technology teams at low cost, grasp the cutting-edge evolution path of AI robots in a timely manner, and find new landing scenarios for its own hardware platforms (such as SoCs and sensor modules).
On the other hand, South Korea's local industry is also actively entering the robot track. Large conglomerates such as LG, Hanwha Group, and Mirae Asset participated in this round of financing, with investments ranging from $5 million to $10 million. In this context of "intra-city confrontation", Samsung must maintain control of key resources to prevent the outflow of technology and market information.
It is worth noting that, according to multiple people familiar with the matter, Samsung's internal technical team still has reservations about some of Skild AI's solutions, believing that its path planning in semantic understanding and highly dynamic environments still needs to be improved. This strategy of "betting and watching" reflects Samsung's prudence and pragmatism in the field of robot investment.
Ⅲ Global tech giants compete for the robot track: the pattern is emerging
Skild AI isn't the only robotics business that has garnered capital and strategic attention. From Apple and Meta to Amazon, Google, Tesla, Huawei, and BYD, almost all tech giants are investing resources in different dimensions to seize the high ground of robotics.
* Apple: Although public information is limited, according to a 2024 report by Bloomberg, Apple's R&D team has launched a research and development plan for domestic robots, with an initial focus on "mobile intelligent assistants".
* Meta: In 2023, Meta will establish a humanoid robot project department, with the goal of combining large language models with physical robots to explore AI embodied AI.
* Amazon: The Kiva system it acquired has created a barrier in the field of warehouse automation, and is now developing a new generation of mobile collaborative robots.
* Google: Its DeepMind and Google Brain teams have long been working on robot perception and action generation algorithms, and will open-source its RT-2 (Vision-Language-Action) multimodal model in 2024.
* Tesla: In 2023, the beta version of the humanoid robot "Optimus" will be released, and Musk has publicly stated that he hopes to enter the home application scenario within 5 years, and the price will be controlled within $20,000.
* Huawei: Focusing on Ascend chips, HarmonyOS systems, and RoboStack cloud platform, Huawei is building a full-stack robot ecosystem of "AI + OS + industrial applications", focusing on service robots and intelligent manufacturing.
* BYD: Relying on its vertically integrated manufacturing advantages and the accumulation of motors and electronic control systems, BYD is actively developing collaborative robots suitable for smart factories, and plans to expand its robot technology to new energy workshops and overseas bases.
What these companies have in common is that the deep integration of AI models has become a core capability for the next stage of robotics breakthroughs, not just mechanical execution or embedded hardware integration.
Ⅳ What is the value of Skild AI's core technology?
Skild AI was valued at $4.5 billion because of the highly scalable and transfer-learning capabilities of its software platform. At present, its platform mainly focuses on three core modules:
1. Perception system: Integrate high-resolution cameras, LiDAR, multi-axis IMUs and other sensors, and combine neural networks for semantic segmentation, target recognition and depth estimation;
2. Decision engine: Based on the multi-modal Transformer architecture, it supports input of multi-dimensional information such as language, images, videos, and environmental status to achieve task-level control.
3. Action planning and control: The multi-task control strategy trained by imitation learning and reinforcement learning algorithms enables the robot to have the ability to perform high-complexity tasks.
These technologies enable the platform developed by Skild AI to be used not only for humanoid robots, but also for security robots, manufacturing robots and other forms, with the characteristics of "model versatility", which is exactly in line with the ecological platform compatibility and general computing deployment logic pursued by NVIDIA and Samsung.
Ⅴ The distance between technology convergence and commercial implementation
Although capital and technology are injecting a boost into the robot track, the commercialization of the robot industry still faces challenges:
* Cost bottleneck: At present, the unit cost of mainstream humanoid robots such as Optimus and Agility Robotics' Digit is generally more than 30,000 US dollars. Although Skild AI is committed to optimizing the software stack and reducing system complexity, the cost reduction on the hardware side relies on broader supply chain collaboration.
* Application scenario limitations: At present, robots are gradually implemented in B-end scenarios such as manufacturing, logistics, and medical care, but there are still problems such as insufficient interactive intelligence and high threshold for use in C-end scenarios such as home, education, and retail.
* Non-uniform standards: Different countries have different standards in terms of underlying interfaces, communication protocols, and safety specifications of robots, resulting in high technology migration costs and poor compatibility.
These practical problems remind us that the robot industry will continue to focus on technology accumulation and ecological construction in the short term, and it will take time to truly achieve large-scale commercialization.
Ⅵ Conclusion: From the "card war" of capital to the "ecological war" of technology
Nvidia and Samsung's joint bet on Skild AI is not only a forward-looking layout of the AI + robotics technology trend, but also marks that the global technology giant has officially entered a new stage of "embodied intelligence" from the AI model war.
If generative AI has reshaped the information processing capabilities of virtual worlds in the past five years, embodied intelligence will be a key force in transforming the real world in the next decade. From large models to robots, from algorithms to physical entities, we are on the eve of the second wave of intelligent technology.
Skild AI may be just the "vanguard sentinel" of this change, but the signal behind it is clear: the next round of technology dominance will move from "computing power" to "ecosystem battle". Whoever can build a closed loop on chips, operating systems, AI models, and robot platforms is likely to lead the future path of general intelligence.
