Huawei has released its Data Storage 2030 white paper, setting out a technology roadmap for the global storage industry over the next five to 10 years, as AI large language models drive data creation into what the company calls the yottabyte era.
Following the conclusion of the Trump-Xi meeting and amid continued delays in China approving imports of Nvidia H20 GPUs, China's National Development and Reform Commission (NDRC) on May 22 sent a strong policy signal on artificial intelligence (AI) self-sufficiency, explicitly calling for greater efforts to pair domestic large language models with domestically developed computing chips.
Taiwan's state-funded quantum technology research team will enter a second phase running from 2027 to 2031, with a focus on accelerating the transfer of research results to leading companies, according to Shangjr Gwo, who heads the National Science and Technology Council's (NSTC) quantum systems task force. Speaking at the 13th Taiwan-Finland Business Forum in Taipei on May 25, Gwo, convener of the Taiwan Quantum Program Office, said Finland is expected to be a key partner in advancing HPQC hybrid computing and integrating QPU, CPU, and GPU systems more effectively.
AI computing's massive demand for infrastructure is making interconnect and laser technologies key to overcoming power and bandwidth limits, prompting silicon photonics unicorn Lightmatter to unveil its latest laser product, Guide DR. To advance its 3D-stacked silicon photonics engine, Lightmatter is also working closely with TSMC using its Compact Universal Photonic Engine (COUPE) platform.
NTT updated its lightweight Japanese large language model, tsuzumi 2, to better understand charts, graphs, and other visuals in business documents, thereby strengthening numerical and logical reasoning. The change is aimed at on-premises and private-cloud deployments that handle sensitive data, with implications for multinational firms and organizations that need secure, accurate Japanese document processing.
Advantech has deepened partnerships with Nvidia, AMD, and South Korean startups to push a hybrid edge architecture for manufacturing AI, promising faster on-site inference and continuous cloud-driven model updates. Global manufacturers could use this approach to scale Physical AI across factories while managing compute, protocol, and sensor challenges for worldwide adoption.
J&V Energy Technology announced the launch of a subsidiary focused on supercomputing. It said its system-level energy storage unit, Recharge Power, will list on the emerging stock board on May 27 as the two firms target energy storage infrastructure for AI computing centers. The move responds to rising GPU power demands and aims to capture growth from the convergence of AI and energy in Taiwan and abroad.
Castrol is expanding from supplying cooling fluids to providing liquid-cooling testing and lifecycle services for AI data centers, the firm announced, as demand for faster deployment and reliable operation grows. The company said its Silicon Valley laboratory opened in 2026 to deliver load bank testing and simulation of power and liquid-cooling infrastructure before customer site deployment, targeting containerized data centers and hyperscale environments.
Chun Yuan Steel outlined strategic moves on May 25 as it pursues new growth opportunities tied to Toyota's plan to establish a Taiwan production base and to rising demand from China's southern low-altitude economy and robotics sectors. Executives said the company will prioritize automation and targeted investments rather than rapid capacity expansion, and flagged the second half of 2026 as a period when automotive materials demand should improve.
Longwell, a cable and connector supplier, said on May 25 that it has entered Nvidia's supply chain for the next-generation AI server platform after its high-power cables passed the latest chip platform certification and moved into market qualification and pilot production. Shipments were expected to begin in the second half of 2026, with volumes set to expand further in 2027, the company announced.
TECO Electric & Machinery announced on the 25th that it signed an agreement to acquire about 78% of Malaysian engineering firm Dynaciate Engineering Sdn. Bhd. for roughly 200 million ringgit (about NT$1.6 billion). The closing was targeted by the end of August and the firm said the move is intended to expand TECO's data center infrastructure footprint in Southeast Asia to serve global cloud service provider customers.
More coverage
