At his speech at Davos earlier this year, Musk reiterated his highly provocative prediction that, in the future, the number of robots on Earth will surpass the number of humans.

Clearly, AI and robotics have become the two dominant tech topics worldwide: one is general artificial intelligence steadily approaching the AGI tipping point, and the other is robotics emerging from labs to take over human physical labor. Similarly, beyond the AI concept, the crypto industry’s key focus this year also includes embodied intelligence. Below are notable projects in the robotics赛道.

OpenMind

On August 4, 2025, according to official announcements, OpenMind, a Silicon Valley-based intelligent machine infrastructure company, announced the completion of a $20 million funding round led by Pantera Capital, with participation from multiple institutions including Ribbit, Sequoia China, Coinbase Ventures, DCG, Lightspeed Faction, Anagram, Pi Network Ventures, Topology, Primitive Ventures, and Amber Group, as well as several prominent angel investors.

OpenMind helps robots think, learn, and work by developing open-source software. The native open-source AI robot operating system, OM1, enables the configuration and deployment of AI agents in both digital and physical worlds. Users can create an AI character and run it in the cloud or on a physical robot in the real world.

In simple terms, OpenMind developing OM1 is like building an "AI brain" for a robot. This "AI brain" can have multiple AI agents working together, interact with multiple LLMs, and pull data from various sources to perform tasks—such as posting content on social media for users. Because OM1 is open-source, it’s also a highly adaptable robot operating system, similar to Android for smartphones, independent of underlying hardware.

In addition, OpenMind features a blockchain-based robot identity network called FABRIC, designed to provide a verifiable layer of trust shared by humans and robots. Humans can earn badges by sharing location data via maps, evaluating robot behavior, and contributing to development, while each robot equipped with the OM1 system joins the FABRIC network, gaining a unique verifiable identity that enables on-chain tracking of its commands, operational logs, ownership, and related activities.

In December 2025, OpenMind partnered with stablecoin issuer Circle to launch an autonomous robotic payment system based on the x402 protocol. As robotic capabilities advance, they will no longer merely serve as tools for executing tasks, but will begin to function as autonomous economies—requiring the purchase of compute power, data, skills, and even the hiring of other robots or humans to accomplish complex tasks.

CodecFlow

CodecFlow provides a unified platform that seamlessly runs on cloud, edge, desktop, and robotic hardware, supporting both modern popular APIs and legacy systems. The platform normalizes inputs from diverse robotic sensors into a common format and modularizes complex robotic actions, enabling development teams or users to avoid designing robots from scratch. Perception, decision-making, and control across robots can also interact over a network, rather than being fragmented or tied to hardware-specific platforms.

AI-driven operators perceive and reason in real time to respond to changes in software UIs or robotic environments, addressing the fragility of traditional robotic automation that overly relies on pre-written scripts when faced with even minor changes. In short, they capture screenshots, camera feeds, or sensor data, process these external inputs with AI to interpret observations or instructions, and ultimately execute decisions through user interface interactions.

Peaq

On March 27, 2025, the DePIN Layer 1 protocol Peaq completed a $15 million funding round led by Generative Ventures and Borderless Capital, with participation from Spartan Group, HV Capital, CMCC Global, Animoca Brands, Moonrock Capital, Fundamental Labs, TRGC, DWF Labs, Crit Ventures, Cogitent Ventures, NGC Ventures, Agnostic Fund, and Altana Wealth.

Although the initial narrative focused on DePIN, peaq released its Robotics SDK in September last year, enabling robots to obtain autonomous digital identities, make and receive payments, verify data, and integrate into the on-chain network economy. Now, any robot compatible with the ROS2 system can join the peaq network economy and transact with humans or other robots using its universal standard.

In addition, peaq launched a robot RWA project called "RoboFarm" on DualMint last year, establishing a robot farm in Hong Kong that automates 80% of agricultural production through robotics. The lettuce, spinach, and kale grown are sold in Hong Kong, with NFT holders expecting an annualized return of approximately 18%.

Axis Robotics

Axis Robotics is dedicated to building a distributed infrastructure for scaling embodied intelligence (Physical AI). They believe that a Simulation-First approach is the optimal path to overcoming the bottlenecks of robotic data scarcity and model generalization, achieving triple leaps in data quality, richness, and scale through low-cost, large-scale data collection combined with their proprietary data augmentation engine. Meanwhile, every data asset is equipped with verifiable on-chain provenance, collectively forming the core fuel reservoir driving the evolution of General Robot Intelligence (RGI).

Axis has revolutionized the way training data for robots is provided. Other projects offering "input/provision of robot training data" typically rely on mobilizing users to record and upload videos of themselves performing specified actions in the real world using devices such as smartphones or smart glasses, aiming for low-barrier, global user participation. Although this method reduces data acquisition costs, the physical realism of the captured video data is insufficient, lacking depth information and failing to ensure the consistency and accuracy of 3D data.

Through "simulation emulation," Axis addresses this pain point by enabling models to complete tasks under more demanding virtual conditions—through a wide variety of simulated scenarios (lighting, angles, friction, dynamics, etc.)—thereby achieving strong generalization capabilities. Axis employs a Hybrid Strategy, combining scarce real-world data with vast amounts of synthetic data. Leveraging GPU-accelerated metadata augmentation techniques, it generates extensive variations in lighting, texture, and physical properties for a single scene. Virtual environments are not static or hardcoded; instead, they are dynamically adjustable. Code can generate countless scenarios, challenging robots with more rigorous and comprehensive requirements across diverse conditions. The cost of generating these scenes is low, while the volume of output is enormous—a data-driven approach to convergence toward optimal solutions that has been partially validated by major industry leaders such as Google and NVIDIA.

Axis has completed its first simulated robotic learning project, "Little Prince's Rose," openly released to the community. In the "Little Prince's Rose" project, users successfully guide a robot to perform a watering action within a simulated environment via a web interface. By collecting and analyzing user interactions, the robot learns how to water plants. Users can remotely operate the robot through a web page, maintaining the low cost and low barrier of video upload-based data collection while enabling the robot to develop a native 3D-aware VLA (Vision-Language-Action) foundational model, thereby enhancing its spatial reasoning capabilities beyond what video data alone can provide.

Within just five days of launch, the "Little Prince's Rose" project saw ordinary users worldwide, with no background in robotics, contribute thousands of high-quality, strategy-trainable trajectories through an engaging experience. Leveraging this data, Axis successfully trained a strategy model and achieved real-world replication with the Franka robotic arm. This marks Axis’s full end-to-end closure of the pipeline: "task generation → community collection → data augmentation → model training → real-world deployment."

One hour of real data can be transformed into 1,000 hours of training data, and this efficiency multiplier significantly reduces the cost of generalizing robot models.

During the Lunar New Year Beta test, 18,000 participants without robotics industry backgrounds completed 27 brand-new tasks on Axis in just five days, contributing over 100,000 data trajectories. The test successfully supported high levels of in-task randomization and validated compatibility with diverse assets such as wheeled robots and dual-arm robots.

Axis's core product will be officially launched in late March, and it plans to open-source the world's largest purely simulated dataset based on the Franka robotic arm by the end of April or early May, fully meeting the needs of strategy and model training. Meanwhile, as a robotics project originating from Crypto-AI, Axis has begun exploring and advancing real-world industry applications, accelerating commercialization through partnerships with benchmark clients across multiple niche areas: collaborating with an automotive manufacturer to implement automation solutions in production; reaching consensus with a pre-IPO compute company on virtual assets and world models; and establishing deep partnerships with multiple embodied robotics companies in critical areas such as virtual simulation data collection and model training. These efforts reflect the rare external impact of a Crypto project.

GEODNET

A decentralized network providing centimeter-level real-time kinematic positioning data for drones, robots, and other devices, with over 21,000 active base stations across more than 150 countries. Over the past year, the project generated over $7 million in revenue, showing a quarter-over-quarter growth trend.

Although this project is more commonly classified as DePIN, demand for high-precision, real-time positioning data is expected to grow as robotics become more widely adopted in real-world applications. In February 2025, Multicoin announced it led the acquisition of $8 million worth of $GEDO tokens from the GEODNET Foundation.

BitRobot

The BitRobot Network, jointly developed by FrodoBots Lab and Protocol Labs, aims to enable distributed robotic work and collaboration. Its key components include: Verifiable Robot Work (VRW), a quantifiable metric for defining and verifying robotic tasks and network rewards; Entity Node Tokens (ENT), unique NFT-based identifiers that represent device ownership and network access; and subnets, which serve as the operational layer for task execution and act as resource clusters that create value for the BitRobot Network.

On February 14, 2025, FrodoBots Lab announced the completion of a $6 million seed round, bringing total funding to $8 million.

FrodoBots Lab also sells robots; Earth Rovers, like real-life Mario Kart, cost $249, and players remotely control their robots via browser in the global treasure hunt game ET Fugi, providing data for researchers to deploy and test their latest AI navigation models. ET Fugi is also BitRobot’s first subnet.

Another game bot, Octo Arms, will be launched in the future, allowing players to remotely control robotic arms to complete various 3D puzzle games and competitions.

The concept of a "subnet" in this bot network is somewhat abstract; simply put, any cluster (or specific project/event undertaken by a cluster) that contributes to the overall network ecosystem is considered a subnet—for example, the ETFugi game mentioned above, as well as SeeSaw launched by Virtuals.

SeeSaw

Subnet 5 of BitRobot, a robot training data sharing app launched by Virtuals in October last year. On SeeSaw, users record videos of their daily activities, upload them to complete tasks, and earn rewards. These video datasets—capturing everyday actions like tying shoelaces or folding clothes—from users around the world will be used to train robots.

Open

Auki’s decentralized machine perception network, Posemesh, connects humans, devices, and AI through a DePIN (Decentralized Physical Infrastructure Network) architecture, enabling devices such as robots and AR glasses to share location and sensor data in real time, collectively building a collaborative spatial understanding of the physical world, and providing a shared spatial view for robots, AR, and AI.

Multiple node roles have been designed based on the Posemesh protocol. Compute nodes provide processing power, motion nodes (robotic endpoints) upload location and sensor data, reconstruction nodes generate 3D map models from this data, and domain nodes manage the 3D space. Each node is incentivized with $AUKI tokens based on its contribution, driving a self-evolving machine vision network.

This network emphasizes privacy protection, preventing any single entity from monitoring users' private spaces, and can be applied to multiple scenarios such as retail (product placement optimization), property management (asset tracking), exhibition navigation, and building renovation.

Their Cactus AI spatial computing platform has already engaged in active pilot programs with Toyota Material Handling and the Swedish supermarket Stora Coop.

XMAQUINA

A DAO that enables retail investors to participate in robotics company investments. The DAO raised $10 million by selling its tokens, $DEUS, in batches. Currently, the DAO has used the auction proceeds to acquire stakes in six robotics companies: Apptronik, Figure AI, Agility Robotics, 1X Tech, NEURA Robotics, and Robotico. Some investments have already begun generating profits, with individual returns exceeding 100%.

PrismaX

On June 17, 2025, PrismaX announced the completion of an $11 million funding round, with investors including a16z CSX, Volt Capital, Blockchain Builders Fund, Stanford Blockchain Accelerator, and Virtuals.

PrismaX builds an open coordination layer that connects remote operators, robot users, and robot companies. Operators can connect with users to remotely control robots and complete real-world tasks while collecting valuable data. Users can also request real-world services such as logistics and advertising.

PrismaX also features a protocol for remote robot operators, allowing businesses to find experienced operators capable of handling complex tasks. Operators can choose to stake network tokens to increase trustworthiness and gain access to higher-paying tasks. Rewards earned by stakers are not only based on the amount staked but also on the quality of their work, with additional bonuses awarded for improved efficiency.

Data collected from remote operations will be used to train robots to enhance their autonomy, thereby improving the efficiency of remote operators and ultimately achieving high or even full robot autonomy.

NRN Agents

NRN evolved from the real-time training blockchain game AI Arena, featuring AI Agent battles. On October 28, 2021, developer ArenaX Labs announced the completion of a $5 million seed round led by Paradigm Capital, with participation from Framework Venture Partners. On January 9, 2024, ArenaX Labs announced a new $6 million funding round led by Framework Ventures, with participation from SevenX Ventures, FunPlus/Xterio, and Moore Strategic Ventures.

Although it largely follows the process of collecting data to reinforce robot learning, NRN leverages its extensive experience in gaming to deliver a browser-based experience that turns robot data collection into a game, allowing users to intuitively control simulated robots directly through their browser. During gameplay, the behavioral data generated by user actions is used to train real-world robotic systems.

At this stage, the project will focus on the robotic arm (RME-1) to validate data collection, real-time learning, and adaptability.

Click to learn about the open positions at BlockBeats

Welcome to the official BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group: https://t.me/BlockBeats_App

Official Twitter account: https://twitter.com/BlockBeatsAsia