ZhiYi Technology has successively completed Series A and Series A+ financing at the 100-million-yuan level, reconstructing the physical AI data base through the revolution of simulation resolution.
New Obstacles for the Implementation of Physical AI: From the Computing Power Crisis to the "Data Gap"
In the past decade, the rise of large models has mainly relied on the leap in computing power brought by GPU clusters. Today, as AI begins to move from the digital world to the physical world, the core bottleneck for its value implementation has shifted to a deep understanding of physical laws.
Whether it's the virtual testing of aero - engines or the digital twin of controllable nuclear fusion, what is truly scarce is no longer just computing chips, but high - quality simulation data in the four - dimensional space - time domain that can support AI training.
However, the engineering capability boundaries of global mainstream simulation tools can no longer meet the requirements of physical AI for "restoring the evolution process of the real physical field in the digital space". If the data volume of a typical steady - state simulation example with 20 million grids is set as the benchmark quantity 1:
l The boundary of mainstream tools: The upper limit of engineering capability is only around 1000.
l High - value industrial scenarios: The demand for composite materials and 3D stacked chips reaches 10⁴~ 10⁷, and the continuous calculation of nuclear fission reactors and aero - engine whole machines reaches 10⁵~10⁷.
l Frontier extreme scenarios: The demand for the whole - process simulation of reusable starships reaches 10⁸ (the single - time data volume is equivalent to the total annual traditional simulation volume globally), and that of controllable nuclear fusion is as high as 10⁹~ 10¹².
Facing this huge data gap, Zhyi Technology has successfully built a core barrier of "algorithm innovation + computing power adaptation + scenario implementation" through breakthroughs in fields such as multi - spatio - temporal scale solving, ultra - long - period simulation, and fully asynchronous heterogeneous computing models.
Disruptive Technological Breakthrough: From "Linearization" of Computing Power Cost to a Comprehensive Breakthrough in First - Principles
Relying on 18 years of continuous research by the founding team, Zhyi Technology, with its independently developed STE - KEP - FR core algorithm, has for the first time globally achieved the linearization of the solver's computing power cost and problem complexity at the engineering practical level — that is, when the resolution doubles, the computing cost only increases linearly, completely breaking the curse of the geometric series climb of traditional methods.
Based on this underlying capability, the company has achieved two disruptive upgrades:
l Practical simulation of hundreds to tens of billions of grid points with a cross - order - of - magnitude reduction in computing power cost: Through fully asynchronous four - dimensional space - time integrated solving, scale adaptability is achieved, avoiding the explosion of computing volume caused by the iterative method and synchronous calculation of traditional software. Currently, the practical simulation capabilities of 250 million grid points on a single GPU card, 2 billion grid points on a single node, tens of billions of grid points in a single cabinet, and hundreds of billions of grid points in a small cluster have been realized. The cost of an example that requires tens of millions of costs on top - level supercomputers for international competitors is compressed to tens of thousands to hundreds of thousands of costs in a small cluster. Under the same cost, the simulation resolution is improved by more than two orders of magnitude.
l Restoring the accuracy of first - principles and producing "golden data" for physical AI: Relying on the unique spatio - temporal locality of the algorithm, the DIMAXER software achieves a solution that is completely close to the evolution process of the physical field, avoiding the turbulence model, geometric simplification, and various empirical assumptions to the greatest extent. There is theoretically no convergence problem, fundamentally solving the defect of error amplification of traditional algorithms under strong nonlinearity and multi - scale coupling. It provides indispensable underlying data for physical AI training with the real evolution process of first - principles accuracy.
Formation of the Data Flywheel: DIMAXER - InField Launches the Infrastructure for Physical AI
The transformation of Zhyi Technology from a CAE software development enterprise to a physical AI data provider is a natural result of the combination of market demand and unique technological advantages. In 2025, the company achieved single - order contracts worth tens of millions of yuan for the purchase of hundreds of GPU card licenses. In 2026, there are multiple contracts worth tens of millions of yuan for the construction of virtual - real integration platforms, virtual laboratories, and ultra - large - scale simulation businesses, with the cumulative process contract amount exceeding 100 million yuan. It is reported that Zhyi Technology will launch the DIMAXER - InField platform next. This platform combines neural networks and improvements to the AI - Parallel core algorithm, aiming to solve the last puzzle of long - period simulation and achieve "the ultimate restoration of the physical process in the digital space". In the future, DIMAXER - InField will integrate high - resolution simulation data, experimental sensor data, and surrogate models to build a closed - loop data flywheel of "simulation as a data producer and AI as a real - time inference engine", form a scalable physical AI infrastructure, and promote the full - scale upgrade of industrial simulation to virtual laboratories.
Views of Investors
Yijing Capital said: Zhyi Technology's original four - dimensional space - time integrated solving algorithm has opened up a new simulation technology path and injected new vitality into the industrial simulation industry. The company provides customers with a one - stop solution for cloud services, high - precision calculation, and ultra - large - scale solving, solving the common industry problems of "ineffective solving in complex scenarios, low efficiency in large - grid calculation, and high cost of commercial software". Combined with the team's concept of continuous pursuit of excellence and the open attitude of building an ecosystem with industrial partners, it is the driving force for Yijing Capital to continue to invest. Facing the coming physical AI revolution, high - precision simulation data is the cornerstone of its implementation. We are optimistic about the company's continuous iterative ability in the core solver algorithm, which will drive the continuous increase in the industry's demand for high - precision simulation data, and will surely build a bridge between the "digital and physical worlds", making it a pioneer in the global implementation of physical AI. Yijing Capital will also continue to empower Zhyi Technology, accompany it on the journey, and become a participant in human progress.
Yida Capital said: Zhyi Technology has long focused on the CAE industrial simulation software track. With more than a decade of algorithm technology accumulation and continuous optimization and iteration, combined with the empowerment of the new - generation AI infrastructure and frontier technologies, it has successfully broken through the technical bottlenecks of traditional industrial simulation software and promoted the industry's evolution towards digital virtual laboratories. The company's self - developed physical AI technology can be widely applied in high - end equipment and strategic industrial fields such as large aircraft, launch vehicles, underwater equipment, embodied intelligent robots, and controllable nuclear fusion, effectively helping customers shorten the R & D cycle, reduce costs, and increase efficiency. Yida Capital highly recognizes the core technical barriers and professional R & D strength built by the company, which will surely lead the technological innovation of physical AI and the large - scale implementation of the industry. In the future, it will also continue to deeply empower and jointly build a new pattern of industrial development.
Biqiu Capital said: The core barrier of the physical AI industry lies not in model iteration, but in the high - quality and high - fidelity physical simulation data infrastructure, which is also the core pain point of current industrial digitalization and intelligence. Zhyi Technology, through breakthroughs in underlying original algorithms, has completely solved the industry pain points of traditional simulation, such as the explosion of computing power cost, accuracy distortion, and limited scale. It produces golden data suitable for high - end industrial and frontier technology scenarios with first - principles simulation capabilities, making up for the key short - board of physical AI implementation. The team has been deeply involved in industrial simulation for many years, with deep technical barriers and solid commercial implementation capabilities, and has formed a clear product matrix and a large - scale order closed - loop. We are optimistic about the long - term explosive potential of the physical AI simulation data track, and also firmly believe that Zhyi Technology can continue to consolidate its leading position in the industry. Relying on its ultimate data production ability, it will continue to empower industrial digitalization, high - end equipment R & D, and frontier scientific research innovation, and promote the leap - forward upgrade of the global physical AI industry.
Dr. Lu Yi, the founder and CEO of Zhyi Technology, said: "In the new industrial revolution driven by AI, intelligent agents liberate engineers from simple and repetitive labor and enable the continuous evolution of expert experience. AI4Science makes it possible that 'calculated examples do not need to be recalculated' and continuously reduces the cost of data reconstruction and regeneration based on the strong generalization ability of massive data. On this basis, the real physical AI revolution will surely continuously pursue higher - quality simulation data and achieve the deep integration of the digital and physical worlds. Zhyi Technology will continue to provide the industry with higher - quality data production capabilities and related services, create incremental value for customers, and take '1000 - second simulation of one trillion grid points by 2030' as a new goal to promote a real productivity revolution in physical AI."