AI‑Powered Satellite Turns Itself Around in Space – A Leap Toward Fully Autonomous Orbiters
A team of researchers at Germany’s Julius‑Maximilians‑Universität Würzburg (JMU) has proved that a satellite can steer itself without any human commands. Using a high‑fidelity computer simulator, they trained an artificial‑intelligence model in the lab and then uploaded it to the flight controller of the InnoCube nanosatellite, which is already circling the Earth in low‑orbit. During a series of on‑orbit tests, the AI successfully re‑oriented the satellite’s attitude—essentially turning the spacecraft to point where it needed to—demonstrating true autonomous control in the harsh environment of space.
This breakthrough joins a growing list of AI achievements in orbit. NASA’s Jet Propulsion Laboratory recently used an automated system to aim a satellite’s camera around clouds, while the U.S. Naval Research Laboratory is preparing its Autosat system to let a satellite calibrate its own signals and handle data links on its own. Researchers at the University of California, Davis and the startup Proteus Space are also gearing up to launch AI‑enabled satellites.
Together, these milestones signal a shift from ground‑controlled spacecraft to self‑sufficient satellites that can make real‑time decisions, reduce operational costs, and open new possibilities for deep‑space missions and large constellations.
Read more
China Unveils Quantum Supercomputer That’s 4.5 Billion Times Faster Than Today’s Fastest Machine
China Telecom’s Quantum Research Institute announced on November 14 that its new superconducting quantum computer, named Tianyan‑287, has reached "quantum supremacy" – it can solve certain problems about 4.5 billion times faster than the world’s most powerful classical supercomputer. The machine uses the same 105‑qubit chip family as the celebrated Zhu Chongzhi No. 3 prototype and adds 182 coupling qubits, making it the most advanced superconducting quantum processor built in China.
Tianyan‑287 will soon be linked to the Tianyan quantum‑computing cloud platform, offering the first globally accessible quantum‑cloud service that boasts true supremacy. Remarkably, every major component – from the dilution refrigerator to the control electronics – was designed and manufactured domestically, showcasing a fully home‑grown supply chain. The team also deployed an AI‑driven calibration system that automatically tunes the delicate quantum hardware, dramatically improving precision and uptime.
While experts caution that practical, everyday applications are still 5‑10 years away, the breakthrough marks a pivotal step toward the next information revolution. Investors see the race to commercialize quantum technologies heating up, and China’s achievement positions it as a serious contender in the emerging quantum economy.
Read more
China Joins Global Race to Build AI Supercomputers in Space
In early November, the world’s tech giants announced bold steps toward putting artificial‑intelligence power in orbit: Nvidia shipped its H100 GPUs to space, SpaceX’s Elon Musk promised 100 GW of space‑based data centers via Starship, and Google unveiled “Project Suncatcher,” which will launch two prototype satellites carrying its TPU chips by 2027. China is keeping pace. Guoxing Aerospace’s home‑grown “Zero‑Carbon Space Computing Center” was highlighted as a top innovation at the 2025 World Internet Conference, and in May the country launched the first ever space‑computing satellite constellation – twelve satellites placed into a 700‑km sun‑synchronous orbit on a single Long March 2D rocket.
Why go to space? The Sun offers abundant energy, and putting AI hardware above the atmosphere can sidestep ground‑based power limits and latency. China’s emerging space‑computing ecosystem spans satellite manufacturers (Guoxing, Galaxy Space, MicroNano), domestic chip makers developing radiation‑hardened FPGAs and AI accelerators, and downstream users targeting remote‑sensing, maritime IoT and other data‑intensive services. While high‑performance, radiation‑tolerant GPUs remain largely foreign, Chinese firms are closing the gap. With cheaper launches and maturing on‑orbit computing, experts expect space‑based AI infrastructure to become a mainstream part of the global tech landscape within the next five to ten years.
Read more
New Technique Slashes AI Memory Use by Half, Boosting Speed and Efficiency
Artificial intelligence is getting smarter every day, but the biggest language models that power chatbots and virtual assistants also gobble up massive amounts of computer memory. That memory drain slows down performance and makes it harder to run these models on everyday devices. A research team led by Professor Wang Huan at Westlake University, together with partners at McGill University and the Mila Institute, has just announced a breakthrough that could change the game. By redesigning the way the model stores and retrieves intermediate reasoning steps, the team managed to cut the memory footprint of AI reasoning tasks by roughly 50 percent without sacrificing accuracy. The new approach, described as a "lightweight memory scheduler," dynamically compresses data on the fly and re‑uses freed space, allowing the same model to run faster and on cheaper hardware. Early tests show that large‑scale language models can now handle complex queries with half the RAM they previously needed, opening the door for more responsive AI assistants on smartphones, laptops, and edge devices. The researchers say the method is compatible with existing architectures, so developers can adopt it without a complete redesign. This advance could accelerate the rollout of powerful AI tools in everyday applications, making them more accessible and environmentally friendly.
Read more
Atomic‑Scale Chips: How 2‑D Materials and Single‑Atom Tools Are Shaping the Post‑Moore Future
The race to build computer chips that work at the scale of individual atoms is heating up worldwide. Researchers say that ultra‑thin, two‑dimensional (2‑D) semiconductor materials—just a single atom thick—could be the key to keeping Moore’s Law alive beyond 2030. He Jun, dean at Wuhan University, explains that making these atom‑thin layers uniformly across a whole wafer and then stitching different materials together is the biggest technical hurdle.
In a breakthrough, Professor Zhang Zheng from the University of Science and Technology Beijing grew a high‑quality, single‑crystal monolayer of MoS₂ on a 2‑inch wafer and scaled the process to a 6‑inch platform, proving that 2‑D chips can be manufactured reliably. Meanwhile, Shen Dezhen at the Chinese Academy of Sciences demonstrated a Raman spectroscopy technique that can “see” vibrations of individual atoms, offering a way to detect tiny defects that would otherwise ruin a chip’s performance.
Together, these advances—precise wafer‑scale growth of 2‑D semiconductors and single‑atom‑level inspection—bring us closer to atomically precise manufacturing of next‑generation information devices, from ultra‑fast processors to quantum‑grade components.
Read more
Humanoid Robots Poised for a Breakthrough Year – The ‘iPhone Moment’ Arrives in 2026
A wave of Chinese manufacturers is gearing up for a landmark year in humanoid robotics. Zoomlion says it has been building fully‑integrated humanoid robots since 2024, rolling out several self‑designed prototypes and even a dedicated training campus with a hundred workstations. Its robots now operate on factory floors, thanks to in‑house hardware like planetary joints, reducers and custom AI models.
Changan Automobile is teaming up with top partners to crack the robot’s "brain," power and drive systems, while Midea Group’s industrial robot "Melo" is already handling inspections, maintenance and material‑handling tasks at its washing‑machine plant. Midea also unveiled a home‑use robot, "Mira," at the 2025 World AI Conference.
Tech giants such as Tesla, Yu Shu Technology and UBTECH are racing to perfect dexterous robot hands, and component specialists like Xingdong Jiyuan and Magic Atom are supplying the parts. Analysts forecast the global embodied‑intelligent humanoid market to exceed 200 billion RMB by 2030, and many see 2026 as the industry’s "iPhone moment" when mass production finally takes off. This surge promises a boom not only for robot makers but also for the entire supply chain, from sensors to software platforms.
Read more