One billion! Largest Robot Chip Financing This Year

Just now, Shenzhen-based robotics chip company Groundhog Robotics announced that it had taken theUS$150 million (about RMB 1,024 million)New B2 roundfinancing(math.) genusA retail tech and supply chain giant, DripThe company has been invested by industry giants and first-tier capitals such as Prosperity7 Venture Capital Fund, High Tide Venture Capital, Temasek's Vertex Growth, and Five Sources Capital.

Just 20 days ago, Groundnut Robotics just completed a 120 million U.S. dollars (about 820 million yuan) B1 round of financing, and in May last year, it also completed a 100 million U.S. dollars (about 680 million yuan) A round of financing. Superimposed on this latest round of financing, its total publicly disclosed financing has exceededUS$370 million (about RMB 2.53 billion)The investor lineup coversDDT, Meituan Longzhu, BAIC Venturesand other head manufacturers as well as many well-known investment institutions such as Jinqiu Fund, Wuyuan Capital and Huangpu River Capital.

Formerly the Horizon AIoT division, Groundhog Robotics was officially spun off in January 2024 toChip shipments increased by more than a million orders of magnitude per year, 70% of the body with intelligent enterprise coverage, more than 400 robotics enterprise customers, more than 100,000 robotics developersThe report card, standing on the double wind of capital and market. Groundnut robot CEO Wang Cong said that this also makes them highly favored by the capital, financing is almost watertight.

In the consumer robotics market, the Groundhog Robotics Rising Sun series of smart computing chips have shipped more than as of 20245 million tablets(math.) genusCorvus, Cloud Whale, Shadowstone Insta360are clients; oriented towards the humanoid robotics circuit.Fourier, Accelerated Evolution, Self-Variable Robotics, Star Motion Epoch, Interdynamics by Interdynamics, GWGHeadline companies in the industry have taken the lead in becoming the first strategic customers of S600 globally.

By the end of this year, the Groundhog Robot will be usingDomestic current most advanced process technology, launched three new chipsIt is positioned at the flagship level of the industry.

In this robotics revolution sweeping the world, this chip company, which has been madly snatched by first-tier capital, why in the end has it become the capital's favorite? Wise things into the ground melon robot office in Beijing, and founder Wang Cong started a deep dialog. Wang Cong believes that a single point of technology is not a moat, but a comprehensive lead is the real barrier, compared to the startups specializing in this track, Diguabot has beenAchieving full leadership.

First, the robot chip is the “boss decision” business, the most difficult is how to define the chip

The prototype of the groundnut robot is Horizon's AIoT department, at that time, Wang Cong led this department's main business includes smart home, urban transportation, intelligent parks, etc., but its pain point is that most of the AIoT application scenarios have a low degree of intelligence, do not have a high degree of reliance on AI, and it is difficult to form a continuous climb in the demand for computing power.

At the beginning of 2020, the idea of transforming into a robot chip sprouted in Wang Cong's mind. But the reality is a stick in the head, in the epidemic impact, team turmoil, has not yet established a market reputation and other multiple pressures, they started unusually difficult.

In the industry's general view, the core driving force of hard science and technology entrepreneurship is the product itself. But Wang Cong see this very thoroughly: robotics chip track, the essence of the CEO decision-making level of long-term business, a single product is just a ticket, whether to go far.At the core, it depends on whether they can become a long-term value community with each other.

Therefore, what pried open the door of the first customer was not a particular product of the groundnut robot. He recalled the earliest cooperation with a robotics company, frankly, the communication between the two sides at that time can be said to be out of the heart - from the algorithm training, technology evolution path, infrastructure construction to the chip future route planning, the direction of the two sides of the long-term strategic cooperation, all the key thinking is unreservedly and dish out.

Then the early team of groundnut robot gradually gained a firm foothold. 2024 around the time, with the body intelligence industry on the wind mouth, the industry heat soared, in January of the same year, there were already sixty or seventy people in size of the horizon robotics division was formally split, set up for the groundnut robotics. So far, more than 2 years, the number of employees of the ground melon robot has grown three times, reaching more than 200 people.

Horizon's genetic strengths give it a combination of capabilities that are scarce in the marketplace when it cuts into the robotics track - theThe Trinity of Chips, Algorithms and Robotics TalentIt is also reusableUnderlying Neural Network BPU Architecture.

However, it is still necessary to start from scratch in terms of chip definition, architecture design and new algorithm development for robotics customers. That's why there is a clear consensus within Groundnut Robotics:It's not hard to build a chip, what's hard is to define a chip.

The mutual constraints between chip computing power, power consumption and cost is the objective law of the industry, and the most challenging thing is to find the optimal balance between the three. What Gourd Robotics is trying to do is to find the optimal balance between the three for the robotics industry.“greatest common divisor”The R&D process is a matter of strategic decision-making. During the R&D process, its every trade-off is a matter of strategic decision-making, both in terms of prioritizing the chip's capabilities and pinpointing the target customers and defining the chip around their real needs.

It is in this harsh background, today's groundnut robot, has been among the core camp of the upstream of the robot industry chain.

Second, can support more than a hundred kinds of robot form, the end of the year will send three chips

In terms of product layout, the Groundhog Robot has a clear sense of boundaries:Only making chips, not solutions, only targeting general needs, not taking customized needs.

Based on this position, its team has constructed a business matrix that horizontally expands the ecology, perfects the software and hardware synergy capabilities, and vertically covers diversified arithmetic and development needs of complete products.

Specifically, for pan-robotics and embodied intelligence buildingmicrochipProvidesRDK Robotics Developer KitIn addition, on the basis of theEstablishment of robot operating system, robot algorithm center, data solution including data annotation to simulation verification, through such a combination to smooth out the underlying difficulty of robotics development, so that more manufacturers and developers do not have to build wheels from scratch to serve robotics companies.

Its chips and developer kits provide5~560 TOPSMulti-layered intelligent computing platform covering humanoid robots, wheel-footed robots, quadruped robotic dogs, service companion robots, logistics AMR, floor sweeping robots, lawn mowing robots, etc.More than 100 robot forms.

The team of Groundnut Robotics will integrate the common requirements of pan-robotics and humanoid robots in the chip design stage, and output the standard products directly after the definition is completed, so as to realize one-time R&D and multi-scene adaptation. “The standard is not to fight, is accompanied by the development of the law of the industry slowly grow out of the”, Wang Cong broke the product logic, the current development of the robotics industry is in the early stages, the standard has not yet been formed, the need to follow the laws of the industry to update the iterative products.

Specifically, its chips are laid outX and S seriesThe X series is mainly used to satisfy the pan-robot industry and consumer robots, and the S series is a big computing power chip for body intelligence.

For the current needs of enterprises in different industries, Wang Cong believes that the consumer electronics industry depends on the performance of the product up to standard, the price is appropriate, power consumption is reasonable, in order to make the product more competitive to gain revenue, and with the body of the intelligence is in the early stage, the customer selects the chip is more concerned about the algorithm will be better to run out of the short term will not go to the fight price, spell the cost.

At present, its Xuri 3 (X3), Xuri 5 (X5) have realized large-scale mass production, Xuri 7 (X7) will be launched at the end of this year; last November, the groundnut robot released S600, INT8 peak arithmetic up to 560 TOPS, and is advancing with the head of the customer in the application of the actual scene cooperation. Groundnut Robotics will be launched at the end of this year Sunrise 7 (X7) and three other new chips.All adopt the most advanced process technology in China, positioning the flagship level in the industry..

For product recognition in the market, Wang Cong is quite proud, “Our products have a good reputation among customers, faster and more efficient in the algorithms that customers need to run, and there will be a generational upgrade in the performance of the new chip.”

The RDK Developer Kit is another major business layout for Groundhog Robotics.

Before it was spun off from Horizon, Groundhog Robotics released the flexible and efficient RDK X3 module, and now it has launched the cost-effective, efficient and easy-to-use mainRDK X5and for the full-scenario arithmetic core of embodied intelligenceRDK S100.

Each of these developer kits has its own advantages: RDK X5 has 10 TOPS arithmetic power, which is the best robotics development kit within 1,000 yuan, and the 4G version is priced at 549 yuan and the 8G version is priced at 699 yuan; the RDK S100 arithmetic power reaches the level of 100 TOPS, which can satisfy the higher-order needs of humanoid robots, bionic robots, and other body-intelligent application scenarios for perception accuracy and generalization ability.

▲Magic Box Multimodal Intelligent Robot Development Platform Based on RDK X5

From mature scenarios to cutting-edge robotics tracks, developers can find a balance between cost-effectiveness and performance.

In terms of broader business building, Groundnut Robotics has also assembled its own set of robot operating systems, provided a robot algorithm center, and even offered a set of data solutions that include data annotation to simulation and verification, tying together tools for collaborative development of robot hardware and software.

At the same time, in order to attract developers, 2022 Groundnut Robotics researchers will also write their own courses to go to the university to open classes. Wang Cong talked about, this is because they combed the pyramid portrait of developers, the bottom of the pyramid is the student developers, from the foundation to start, through the competition to pry the college robotics students and teachers, and now at the ecological level has covered theMore than 500 colleges and universities.

▲ Developer Pyramid

Third, a year to invest hundreds of millions of dollars to engage in research and development, robotics track unique stand so that investors rush to invest in

Groundnut Robotics is located in, both robotics and chip dual track, from the beginning to build a very high financial and resource barriers.

Wang Cong revealed that Groundnut Robotics' R&D investment this year will reachseveral hundred million dollarsIn the past, the proportion of chip R&D and developer platform investment was basically equal, and this year, driven by the planning of new chip releases, resources will focus on tilting towards chip R&D.

Along with the robotics track continues to heat up, the groundnut robot's growth is remarkable: over the past year, its shipments grew 180% year-on-year, and the number of customers grew 200% year-on-year. from the perspective of the market structure, the sweeping robots and other mature consumer electronics scene contributes to a larger total revenue, the humanoid robotics track orders steadily climbed, the unit price is high, the potential is large, although not yet scaled but the future is promising.

Behind the huge ongoing investment always comes the reality check of commercial returns.

But Wang Cong believes that the robotics industry has not converged, no need to dwell on short-term investment returns, the core is to recognize the direction of heavy research and development, in the process of doing the best, “when the industry outbreak, the head of the enterprise can naturally get a good return”.

Such layout ideas make the core advantages of the groundnut robot in this track stand out, capital has flocked to the robotics track to grab the layout, and theGroundhog Robotics“ investment is almost a ”closed-eye search."On the one hand, the groundnut robotics previous commitment to investors are honored. On the one hand, the groundnut robot previous commitment to investors are honored, on the other hand, the capital is optimistic about the robotics track, it is inevitable that the need for upstream core enterprises, and the groundnut robot is the current optimal choice.

Wang Cong believes that true innovation tracks require more of a long-term perspective, especially in an investment-heavy, long-cycle industry like robotics.This industry is betting on the future..

Fourth, a single point of technology is never a moat, the groundnut robot has a comprehensive lead over domestic enterprises

Groundhog Robotics wants to bring“Maximizing the Robotics Era.”It is destined to be a long and difficult road to get this thing done.

However, in Wang Cong's view, any single point of technical advantage may be replaced in the future. The real moat, can only rely on long-term accumulation and sustained investment in order to construct, in each technical node, each service experience, each step of ecological construction than the opponent is a little bit ahead of the superposition of sustained capital investment and rapid iteration, and ultimately converge into an impenetrable barrier.

The algorithmic form of the current robotics industry has not yet converged, but its underlying logic is the same. Based on this, the groundnut robot anchors the core part of the industry's most practical needs.

Conclusion: Not chasing short-term commercialization, Groundnut Robot wants to be a long-termist of embodied intelligence

Now, humanoid robots are about to enter the critical node of large-scale landing, the chip as the core “arithmetic heart” or will become the key to determine the direction of the industry. Gourd Robotics chip has been able to accurately adapt to more than a hundred types of robots, to find the robotics industry arithmetic needs of the “greatest common denominator”.

However, in Wang Cong's view, the development of embodied intelligence is still in the early stages, commercialization is not a proposition at this stage, from the breakthrough entrepreneurship to deep plowing products, and then layout of the ecological, groundnut robotics goal is to do a good job of robotics intelligence evolution of the underlying infrastructure to long-termism waiting for the industry to explode.