Researchers throughout Taiwan are tackling advanced challenges in AI growth, local weather science and quantum computing.
Their work will quickly be boosted by a brand new supercomputer at Taiwan’s Nationwide Middle for Excessive-Efficiency Computing that’s set to ship over eight occasions extra AI efficiency than the middle’s earlier Taiwania 2 system, Nvidia CEO Jensen Huang introduced in a keynote at Computex 2025 in Taipei at present.
The AI supercomputer at NCHC is slated to function Nvidia HGX H200 techniques with over 1,700 GPUs, two Nvidia GB200 NVL72 rack-scale techniques and an Nvidia HGX B300 system constructed on the Nvidia Blackwell Extremely platform — interconnected by Nvidia Quantum InfiniBand networking. Introduced at present at Computex, it’s anticipated to go stay later this 12 months.
NCHC additionally plans to deploy a set of Nvidia DGX Spark private AI supercomputers and a cluster of NVIDIA HGX techniques within the cloud.
Researchers from tutorial establishments, authorities companies and small companies in Taiwan will be capable to apply for entry to the brand new system to speed up revolutionary tasks.
“The new NCHC supercomputer will drive breakthroughs in sovereign AI, quantum computing and advanced scientific computation,” mentioned Chau-Lyan Chang, director basic of NCHC, in a press release. “It’s designed to empower Taiwan’s technological autonomy, fostering cross-domain collaboration and global AI leadership.”
Creating native language fashions for Sovereign AI
The brand new supercomputer will help tasks like Taiwan AI RAP, a generative AI utility growth platform. Taiwan AI RAP goals to help the fast growth of AI merchandise by providing startups, researchers and enterprises entry to personalized fashions that mirror native cultural and linguistic nuances.
Among the many platform’s choices are fashions created by Taiwan’s Reliable AI Dialogue Engine, or TAIDE — a public sector initiative to construct Taiwanese massive language fashions (LLMs) for duties together with pure language processing, clever customer support and translation.
Collaborators offering textual content, pictures, audio and video knowledge for the initiative embrace native governments, information organizations and public departments such because the Ministry of Training and Ministry of Tradition.
To help the creation of sovereign AI functions, TAIDE presently affords builders entry to a group of Llama3.1-TAIDE basis fashions. The staff is constructing further sovereign AI LLM providers utilizing NVIDIA Nemotron fashions.
A professor at Nationwide Tainan College is utilizing the TAIDE mannequin to energy a conversational AI robotic that speaks Taiwanese and English with elementary and center college college students. It’s been utilized by over 2,000 college students, academics and oldsters so far. One other professor tapped the mannequin to generate high-quality academic supplies, shortening lesson preparation time for academics.
In healthcare, analysis groups in Taiwan used the TAIDE mannequin to develop an AI chatbot with retrieval-augmented technology that helps case managers ship well timed, correct medical info to sufferers with main accidents and sicknesses. And the Epidemic Prevention Middle at Taiwan’s Facilities for Illness Management is coaching the mannequin to generate information summaries to help the monitoring and prevention of how ailments unfold.
Rushing scientific analysis in local weather and past
In local weather analysis, NCHC helps researchers utilizing the Nvidia Earth-2 platform to advance atmospheric science. These researchers are tapping Earth-2’s CorrDiff AI mannequin to sharpen the precision of coarse-resolution climate fashions, and DeepMind’s GraphCast mannequin in Nvidia PhysicsNeMo for international climate forecasting.
They’re additionally adopting the Nvidia NIM microservice for FourCastNet, an Nvidia mannequin that predicts international atmospheric dynamics of climate and local weather variables, and utilizing Nvidia GPUs to speed up the simulation of numerical climate prediction fashions.
With the brand new supercomputer, the researchers will be capable to run extra advanced simulations and speed up the tempo of AI coaching and inference.
Advancing quantum innovation
NCHC researchers are additionally advancing quantum analysis utilizing the Nvidia CUDA-Q platform and Nvidia cuQuantum library concentrating on functions in quantum machine studying, chemistry, finance, cryptography and extra.
The analysis institute has developed Quantum Molecular Generator, a instrument that generates legitimate chemical molecules, utilizing quantum circuits and the CUDA-Q platform. It’s additionally created cuTN-QSVM, an open-source instrument constructed on the cuQuantum library that accelerates large-scale quantum circuit simulations.
The instrument permits researchers to sort out extra advanced issues, providing linear scalability and supporting hybrid quantum computing techniques to assist speed up the event of large-scale quantum algorithms.
NCHC researchers lately used cuTN-QSVM to carry out a record-breaking 784-qubit simulation for a quantum machine studying algorithm. The institute additionally plans to construct a hybrid quantum-accelerated computing system by integrating Nvidia DGX Quantum techniques.
