Foundation Models & Learning
We explore architectures and learning methods that expand model capability, including reasoning, multimodality, world models, continual learning and self-improvement.
Every technology wave begins outside consensus.
A rare few see what matters before the path becomes obvious.
LinkX backs these early convictions with long-term capital,
helping them become real.
Backing thebuilders ofthe intelligence age
Our North Star
First Principles
Founded in 2021, LinkX Capital invests early in AI and the technologies shaping the next generation of intelligence. We built conviction in AI's long-term potential before the current wave gained broad market recognition, and have backed it with sustained research and investment. We support researcher-founders, believing that deep research expertise and distinctive technical insight can lay the foundations for differentiated products and exceptional companies.Our investments span six areas: Foundation Models & Learning, Chips, AI Infrastructure, AI-Native Applications, Embodied AI, and AI for Science.We work with teams from the early stages of research and company formation, combining long-term capital, incubation support and industry connections to help turn research into products and bring them to market. We look for the next generation of companies that expand the frontiers of intelligence and reshape industries.
Through its long-term collaboration with Tsinghua University in AI incubation, LinkX Capital established and operates LinkX T-ONE Innovation Lab.The lab connects young scientists and research teams with founders, industry partners and long-term capital, helping promising research find a path to products and new companies.
We explore architectures and learning methods that expand model capability, including reasoning, multimodality, world models, continual learning and self-improvement.
We focus on chips and enabling technologies for AI training, inference and on-device intelligence, including compute architectures, memory and interconnects, advanced packaging, and hardware-software co-design.
Compute networks, training and inference systems, data platforms and security make AI more efficient, reliable and usable at scale.
We back AI-native products built around real user needs and workflows, from agents that understand tasks, use tools and improve through feedback to new applications for businesses and consumers.
We explore robotics, embodied models, perception and control systems, and intelligent devices that enable AI to sense, decide and act in the physical world.
We support AI for biology, drug discovery, materials and energy research, connecting computational design with experimental validation to develop new methods and tools for scientific discovery.
From intelligence itself to the systems it transforms.
We explore architectures and learning methods that expand model capability, including reasoning, multimodality, world models, continual learning and self-improvement.
We focus on chips and enabling technologies for AI training, inference and on-device intelligence, including compute architectures, memory and interconnects, advanced packaging, and hardware-software co-design.
Compute networks, training and inference systems, data platforms and security make AI more efficient, reliable and usable at scale.
We back AI-native products built around real user needs and workflows, from agents that understand tasks, use tools and improve through feedback to new applications for businesses and consumers.
We explore robotics, embodied models, perception and control systems, and intelligent devices that enable AI to sense, decide and act in the physical world.
We support AI for biology, drug discovery, materials and energy research, connecting computational design with experimental validation to develop new methods and tools for scientific discovery.