Data center power systems are nearing a major transition as GPU rack densities rise toward the 600 kW range. A new report from SemiAnalysis said 800VDC direct-current distribution is moving beyond hyperscale trials and could alter how data centers are built, powered, and regulated.

Memory module maker Goldkey said it plans to raise NT$6 billion (US$191.4 million) to NT$10 billion in working capital in 2026 through multiple channels as tight supply and rising contract prices fuel a memory supercycle. The company also plans to accelerate a shift into higher-value segments such as industrial control, AI, and edge computing after posting a 30% gross margin and 27.4% operating margin in the first quarter of 2026.

Andrea Gallo, CEO of RISC-V International, used his appearance at the MIPS Forum during Computex 2026 to declare that RISC-V has completed its transition from academic project to industrial standard — and that its dominance in physical AI is no longer a future prediction but a present reality.

Innodisk announced a strategic partnership with Qualcomm and Formosa Plastics Group to launch an AI industrial safety vision solution that upgrades surveillance into intelligent video monitoring and decision-making tools. The collaboration combines Qualcomm Insight Platform AI imaging software, Innodisk's edge AI servers and computing platforms, and Formosa Plastics' field deployment and systems integration capabilities to deliver GenAI search and natural-language conversation features for industrial sites.

Andhra Pradesh is making its most concrete move yet in semiconductors, zeroing in on packaging as the immediate entry point into the chip supply chain. Speaking on the sidelines of the Computex technology expo in Taipei, Bhaskar Katamneni, Secretary to the Government for ITE&C, acknowledged that wafer fabrication remains a long-term "seven or eight-year journey," but said packaging work is already well underway — with four PCB manufacturers having already begun operations, according to him.

Qisda Corp. chairman Peter Chen has pledged that the group will not miss out on the AI boom, with AI servers, 1.6T switches, cooling systems, and power solutions already in place. With orders already in hand and some products ready to ship, Qisda expects its AI business to begin taking off in 2026.

As robots and autonomous systems move from factory floors into hospitals, warehouses, and public spaces, NXP Semiconductors CEO Rafael Sotomayor used his Computex 2026 keynote to argue that the defining challenge of physical AI is not raw intelligence, but the ability to act in milliseconds without waiting for instructions from the cloud.

As the artificial intelligence industry moves beyond chatbots and text generation, a new question is emerging: what happens when AI can act on the physical world, not just understand it?

MiniMax and Z.ai both plan to apply for listings in Shanghai's tech-focused STAR Market, months after the two AI companies successfully launched IPOs in Hong Kong's stock market in January, according to Nikkei Asia. This move indicates the pressure the AI companies face to raise capital amid high compute costs to stay ahead of the competition, even as profitability so far remains out of reach.

Global AI growth is increasingly colliding with electricity limits, a shift that could slow data center buildouts and reshape infrastructure planning from the US to Asia. Delta Electronics chairman Ping Cheng said the bottleneck is already delaying projects, pushing operators toward self-owned power systems and off-grid microgrids.

Quanta Computer executive vice president and Quanta Cloud Technology (QCT) president Mike Yang said GPU applications are expanding across training, edge computing, and storage, as well as in the development of application-specific integrated circuits (ASICs). In response to future demand, the company is preparing for power needs and production capacity, planning to add three more plants in the US by the end of 2026.

OpenAI is weighing whether to publicly release software designed to make advanced AI workloads run more easily across chips from multiple providers, a move that could weaken one of Nvidia's most durable advantages: the CUDA software ecosystem that has helped lock developers into its GPUs.

Lightmatter has joined Nvidia's NVLink Fusion ecosystem, a move that could accelerate the rollout of high-performance optical links for AI infrastructure worldwide. The collaboration is aimed at easing data-center bottlenecks, improving bandwidth, and giving global customers more options for building energy-efficient AI systems at scale.

At the opening day of Computex 2026, Lite-On Technology hosted an AI industry summit under the theme "Powering the AI-driven Future," where DIGITIMES chairman Colley Hwang moderated a cross-industry panel with Nvidia, Infineon Technologies, and Gigabyte subsidiary Giga Computing at the exhibition venue.

SK Group is reportedly placing AI and semiconductors at the center of its next round of business restructuring, prompting a fresh internal review of the strategic value of SK Siltron, a major global silicon wafer maker. The planned sale of SK Siltron, once seen as a move to improve SK Group's financial structure, now faces uncertainty as the group reconsiders whether to push ahead with the deal.

The latest funding round for Anthropic produced a headline-grabbing figure: a US$65 billion capital raise that pushed the artificial intelligence (AI) startup's valuation to $965 billion, surpassing OpenAI and making Anthropic the world's most valuable AI startup.

Computex 2026 showcased the industry's latest innovations with its usual fanfare. Yet beneath the spectacle, the event revealed something far more consequential: artificial intelligence is rewriting not only the rules of competition but also the relationships that have long defined the global semiconductor industry.

COMPUTEX drew major AI supply chain players to Taiwan, and industry sources said SK Group chairman Chey Tae-won was set for a private meeting with Foxconn chairman Young Liu, alongside his meetings with TSMC chairman C.C. Wei and Nvidia CEO Jensen Huang. The reported talks could signal broader cooperation on AI infrastructure.

The European Union's (EU) plan to mobilize EUR20 billion (US$23.3 billion) for a network of large-scale AI "gigafactories" is running into delays and funding uncertainty, stoking concerns about Europe's ability to compete in the global AI infrastructure race.

As Computex 2026 opened in Taipei on June 2, Taiwanese President Ching-te Lai said major international tech companies are increasing investment in Taiwan, underscoring confidence in the country's industrial efficiency and democratic system. He said protecting peace and stability across the Taiwan Strait is the government's most responsible commitment to the global supply chain that relies on Taiwan's AI industry ecosystem.

Delta Electronics said at COMPUTEX 2026 that it has launched a prefabricated AI modular data center designed to speed deployment of AI infrastructure by cutting IT build time by about 60%. The move underscores how vendors are racing to support global demand for denser, faster, and more power-efficient AI facilities.

At Microsoft Build 2026 on June 2, Microsoft outlined how its Azure cloud and broader AI portfolio are being re-engineered for agentic AI — software agents that reason, retrieve knowledge, take actions, and run continuously rather than responding to one-off prompts. Spanning silicon, databases, runtimes, developer frameworks, and governance standards, the announcements describe a stack designed to enable organizations to build agents, deploy them in production, and keep them under control.

Ampak Technology unveiled plans to further pivot into AIoT and Edge AI integrated solutions at COMPUTEX 2026, announcing product launches including an AI system-on-module, 5G Reduced Capability (RedCap) offerings, and High Power Wi-Fi 6 drone communication modules. The wireless communication module designer said the moves respond to an industry shift from cloud data centers to endpoint and edge computing and aim to capture demand from enterprise and industrial applications in the US and other markets.