News Summaries

China’s Chip and Storage Wins Spark Home‑Made AI Superpowers

Since 2026 China’s home‑grown AI models have been called on more and more, driving a surge in demand for computing power and storage. At the same time, global players like NVIDIA and Google are rolling out next‑generation AI processors, creating a rare window for Chinese semiconductor firms to catch up. In the worldwide race, GPUs have long ruled AI training, but their high energy use and cost are becoming a bottleneck as models grow larger. Google’s custom TPU chips, built solely for deep‑learning math, now deliver several times the performance of GPUs at the same power level, and are expected to ship millions of units this year. China is not watching from the sidelines. Companies such as Moore Threads, MetaX and Biren are advancing GPU designs, while Huawei, Cambricon and Baidu’s Kunlun are mass‑producing ASIC chips tailored for AI. A breakthrough came from Zhonghao Core Intelligence, which has built China’s first commercial‑grade TPU, putting the country among the few that can design and produce these specialized processors. On the storage side, Changjiang Storage has begun mass‑producing 294‑layer 3D NAND flash, and CXMT is rolling out DDR4, LPDDR4X and DDR5 DRAM chips, closing the gap with overseas leaders. These twin advances in chips and storage give China a more diversified, self‑reliant AI computing stack, reducing dependence on a single technology and bolstering the digital transformation of countless industries.

New ‘Narwhal’ Light Waves Could Revolutionize Optics—Scientists Reveal How They Trap Light Beyond Known Limits

A team of physicists has uncovered a bizarre new type of electromagnetic wave they’ve nicknamed “narwhal” waves because of their twisted, horn‑like shape. Unlike ordinary light that spreads out as it travels, these waves can confine and guide photons along a narrow path, effectively trapping light in a way that defies the usual diffraction limits. The researchers created the effect by arranging ultra‑thin metallic structures in concentric circles on a chip and then firing a laser through them, forming a self‑sustaining holographic pattern that steers the light. The discovery opens the door to a host of practical applications. In telecommunications, narwhal waves could dramatically boost data‑transfer rates by keeping signals tightly packed, reducing loss and interference. In medical imaging, the ability to focus light beyond conventional limits may enable sharper, deeper scans without invasive procedures. The team also sees potential for ultra‑compact photonic circuits, where light replaces electrons to process information faster and with far less heat. While still in early experimental stages, the work demonstrates a new way to manipulate light at the nanoscale, challenging long‑standing assumptions about how photons behave. If scaled up, these “narwhal” waves could become a cornerstone of next‑generation optical technologies, from faster internet to more precise scientific instruments.

New Roadmap Shows Three Ways to Build Room‑Temperature Quantum Materials for Faster, Cooler Computers

A team of scientists from the University of Ottawa and MIT has just laid out a clear plan for creating quantum materials that work at everyday temperatures—a breakthrough that could make future computers far more powerful and energy‑efficient. Their roadmap, published in the journal *Newton*, focuses on magnetic topological materials, a special class that blends magnetism with the exotic rules of quantum physics. These materials act like tiny highways for electrons, allowing electric current to glide along their edges without losing energy, even without an external magnetic field—a phenomenon known as the quantum anomalous Hall effect. The researchers identified three distinct research paths to achieve room‑temperature operation, ranging from tweaking the atomic structure of existing compounds to designing entirely new crystal lattices. By following these routes, engineers hope to replace the ultra‑cold environments currently required for quantum devices, opening the door to practical, low‑power quantum computers that could run in ordinary data centers or even personal devices. The roadmap not only maps the scientific challenges ahead but also offers a hopeful glimpse of a future where quantum technology becomes a daily reality.

AI Revolutionizes Finance and Puts China on Track to Lead Global Robot Production, Says Nobel Laureate

At a high‑profile forum in Chengdu, leaders agreed that artificial intelligence is now a core engine reshaping the world’s financial system. AI is not just a tool for faster trading; it is driving a shift from pure efficiency toward a balance of speed, security and resilience. Nobel‑winning economist Michael Spence warned that rising geopolitical shocks and soaring debt levels make AI‑boosted productivity essential for keeping economies stable. Spence praised China’s strategy of embedding AI into real‑world manufacturing and services, predicting the country could soon become the world’s largest producer of intelligent robots. Former IMF deputy chief Zhu Min contrasted the U.S. focus on foundational AI research with China’s push to apply AI across factories, banks and everyday life. Bank executives highlighted how AI is redesigning risk control, customer service and product development, while also flagging legal and supervisory gaps that need new standards. The Ministry of Commerce announced plans for an “AI + Consumption” policy to spur smart‑home and retail innovation, and industry groups released technical standards for AI‑driven marketing platforms. Overall, the consensus was clear: AI is set to power the next wave of economic growth, but only if governments, firms and regulators work together to manage its risks and ensure fair, transparent deployment.

Diamond Breakthrough: One Chip Could Blend Superconductors and Semiconductors for Faster Quantum Devices

A team of scientists from Penn State, the University of Chicago’s Pritzker School of Molecular Engineering, and the U.S. Department of Energy’s Q‑NEXT research center—working under Argonne National Laboratory—has uncovered how to turn diamond into a versatile quantum material. By growing ultra‑pure diamond crystals and carefully filtering out background noise, the researchers were able to see the hidden electronic signals that make the material act both like a superconductor (carrying electricity without resistance) and a semiconductor (controlling electronic flow). Their findings, published in the Proceedings of the National Academy of Sciences, map out a clear path to building quantum chips that can perform several functions at once, rather than needing separate components for each task. "This offers a new way of thinking by integrating superconducting and semiconductor behavior to create opportunities for multifunction quantum devices," said David Awschalom, a leading quantum scientist at the University of Chicago. If the approach works at scale, future quantum computers could become more compact, energy‑efficient, and easier to connect with today’s classical computers, accelerating the rollout of real‑world quantum technologies.

Harbin Hosts China’s Top Materials Scientists for Cutting‑Edge Forum

From May 15‑17, Harbin became the hub of China’s materials research as the Chinese Academy of Sciences held its 211th Science and Technology Frontier Forum, themed “Frontiers and Applications of Materials Science.” The three‑day event gathered 35 academicians and 78 leading scholars to discuss hot topics such as AI‑driven material design, next‑generation energy materials, semiconductor breakthroughs, and interdisciplinary collaborations. Key figures—including Academician Yang Wei, director of the Academy’s Technical Sciences Division, Harbin Vice‑Mayor Yang Shupeng, and Academician Chen Jie of Harbin Institute of Technology—opened the forum with speeches highlighting the strategic importance of materials science for China’s modern industrial system. Panels and interactive sessions featured experts from aerospace, mechanics, chemistry, and energy sectors, who presented cutting‑edge research and debated the nation’s strategic needs. The forum’s purpose was threefold: showcase the latest scientific advances, critically assess China’s strengths and gaps, and build a lasting network among researchers to boost innovation capacity. Organized by the Academy’s Technical Sciences Division and co‑hosted by Science China: Technical Sciences and Harbin Institute of Technology, the event underscored materials science as the foundation for new productive forces and a catalyst for higher‑quality economic development in China.

Hidden Treasure from the First Atomic Bomb Test: Scientists Uncover a Never‑Seen‑Before Crystal

When the United States detonated the world’s first atomic bomb at the Trinity site in New Mexico on July 16, 1945, the blast created a mushroom cloud and ushered in the nuclear age. Decades later, a team of researchers led by geologist Luca Bindi from the University of Florence made a surprising discovery in the desert sand that survived that historic explosion. Under the extreme heat and pressure of the blast, atoms rearranged themselves into a brand‑new type of crystal—a clathrate made of calcium, copper and silicon—that has never been found in nature or synthesized in a lab before. The scientists identified the material by carefully analyzing rock samples collected from the test site, using modern imaging and spectroscopy tools. Even more intriguing, the same explosion also produced a rare silicon‑rich quasicrystal, a structure that defies ordinary crystal rules and was reported by Bindi’s group in earlier work. These findings show that the most violent events can forge exotic substances, expanding our understanding of chemistry under extreme conditions and hinting at new materials that could one day have practical applications in technology or industry.

Turning AI ‘Black Boxes’ into Doctor’s Tools: New Method Makes Medical AI Transparent

A research team at the Shenzhen Institute of Advanced Technology has unveiled a breakthrough that could make artificial‑intelligence (AI) systems in healthcare far easier for doctors to understand. Their new technique, called Class‑Association Manifold Learning (CAML), translates the hidden decision rules of complex AI models into simple, visual maps that can be displayed in just a few dimensions. By breaking down the data into two parts—a low‑dimensional “decision” space and a high‑dimensional “personal” space—CAML lets clinicians see exactly how a model arrives at a diagnosis and even generate realistic, “what‑if” patient examples to explore how changes in data affect outcomes. In tests on retinal scans, chest X‑rays, brain MRIs, ECGs and gene data, the method compressed model knowledge into an eight‑dimensional space while losing only 1‑3 % of accuracy—far better than existing explainable‑AI tools. Doctors who reviewed the generated explanation maps said they were far more intuitive and trustworthy. The work, published in *Nature Biomedical Engineering*, could help meet new regulatory demands for AI transparency, improve confidence in AI‑assisted diagnoses, and speed up medical discoveries by turning opaque algorithms into clear, actionable insights.