Data center AI accelerator package power has climbed from about 500W in the A100 generation to roughly 3,500–4,000W for Rubin Ultra-class products expected in 2027, pushing supply current into the thousands of amperes. In conventional frontside power delivery, current must cross the full metal interconnect stack before reaching the transistors, while power and signal lines compete for routing resources. Intel Foundry estimates that raising a single compute package from 1kW to 5kW increases power delivery network I²R losses from 89W to 2,222W, cutting the share of input power reaching the transistors from 66.0% to 41.8%.

When the underlying economics of infrastructure commitments meet the financing muscle of the companies supplying foundational compute, the traditional boundary between supplier and customer begins to blur.

New AI models arrive every few weeks, and the question of which is better often comes down to the leaderboard. Behind those rankings, evaluation can look like an academic exercise in setting rules, running tests, and assigning scores. However, public benchmarks—and strong placement on them—have long been tied to a lucrative business, with money flowing to different players in different eras.

Advanced Micro Devices (AMD) has agreed to acquire World Labs in an all-stock transaction valued at approximately US$8.2 billion. Following the closing, World Labs will operate as a dedicated frontier research organization for large language models (LLMs) and spatial intelligence, with founder Fei-Fei Li joining AMD as executive vice president and chief scientist, reporting directly to Chair and CEO Lisa Su.

South Korea's leading telecommunications operator, SK Telecom (SKT), recently established SK Hyper and SK Horizon in quick succession. The moves mark a pivotal shift for telecommunications AI, signaling a transition from internal operational optimization toward full-scale infrastructure monetization.

Xunfei Healthcare (iFLYTEK Medical), the healthcare AI arm of China's iFlytek, is using its Spark Medical Large Model to support diagnosis and medical record generation while tackling challenges including clinical workflow integration, differences in care settings, and model hallucinations. It also hopes to bring its experience from China to South Korea through local partnerships.

China's continued expansion of DRAM capacity has raised a key question for the global memory market: will supply and demand flip?

Garmin's decision to offer OLED options but no Micro LED version in its new flagship smartwatches, the fenix 9 and fenix 9 Pro, has revived debate over whether the emerging display technology is ready for sustained commercial competition. The shift comes about a year after Garmin launched the world's first Micro LED smartwatch, the fenix 8 Pro Micro LED in September 2025.

Micron unveiled the Micron Ventures Paradigm Fund, a US$250 million venture vehicle, in mid-August 2026, as the memory maker pairs more than US$25 billion in fiscal 2026 capex with a pledge to invest more than US$250 billion in the US by 2035. The fund is designed to give Micron an early look at technology shifts, position it for next-generation AI architectures, and strengthen its strategy to reduce memory-cycle volatility.

Artificial intelligence is rapidly entering enterprise operations, blurring the lines between cybersecurity and supply chain management. In the AI era, Zero Trust is no longer only about who can access a system. Enterprises must also ask whom and what they depend on, how concentrated those dependencies are, and what happens when a critical link fails.

As chip design enters the AI era, heterogeneous integration is driving foundry giants like TSMC, Samsung, and Intel to bring advanced front-end fab capabilities into back-end packaging. In the second part of his interview with DIGITIMES, Dr. Shin-Puu Jeng, chairman of IMAPS and a former TSMC executive, shared how TSMC transformed CoWoS from a niche lab concept into an industry standard.

In a recent interview on the DIGITIMES podcast, Dr. Shin-Puu Jeng, chairman of the International Microelectronics and Packaging Society (IMAPS) and a former longtime TSMC advanced packaging executive, reflected on the early development of CoWoS. In the first of this two-part series, Jeng details how a critical thin-interposer crisis forced TSMC and Xilinx to innovate, laying the foundation for modern advanced packaging.

Global AI has entered a period of dramatic turmoil, with leading researchers warning that top labs are "gambling with our lives" and one Anthropic safety researcher saying there is more than a 10% chance AI will wipe out humanity within the next 10 years. That pressure has pushed Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman into rare agreement: slow model development.

A robot that can map a factory floor still cannot necessarily find the box it has been asked to move, or remember what it saw on its last visit. For autonomous robots, building the map is increasingly just the beginning. What comes next — searching for objects, navigating to specified locations, remembering what the robot has seen, and retrieving that information for model reasoning in complex environments — is where the new applications and challenges lie.

China's two leading memory suppliers are moving to raise capital and expand output as demand for chips grows, a shift that could reshape the country's domestic supply chain. Their efforts reflect a new stage for the sector, but the long-term test will be whether fresh funding can translate into stronger technology, higher yields, and wider customer adoption.