Generative AI applications are expanding rapidly, making computing costs a growing bottleneck to commercial AI deployment. The AI accelerator market, long dominated by graphics processing units (GPUs), has increasingly explored specialised architectures in recent years. Tensor Processing Units (TPUs), optimised for the matrix operations used by AI models, have again drawn industry attention.
Samsung Electronics is considering outsourcing some or all of the back-end design work for an input/output die in Google's reported 10th-generation tensor processing unit, as growing demand for Samsung's 2nm foundry process reportedly stretches its internal engineering resources.
Nvidia unveiled a series of new partnerships in Japan on July 16, 2026, highlighting the growing adoption of AI across manufacturing, robotics, automotive, healthcare and data center infrastructure. The announcements coincided with CEO Jensen Huang's visit to Japan, where the company showcased its latest physical AI technologies and deepened collaborations with several of the country's leading industrial groups.
The IPO prospectus of Changxin Memory Technologies (CXMT), filed ahead of a planned listing on Shanghai's STAR Market, lays bare the international talent base the company has assembled to compete against the established leaders of the global DRAM industry — and raises a subtler question about the residency arrangements of its founder.
Nvidia co-founder and chief executive Jensen Huang used a developer event in Tokyo on July 15 to reject reports that manufacturing problems could delay its next-generation AI accelerator systems, telling reporters the claims were "not true" and that "Vera Rubin is already in production. Giant amounts of production incoming."
CXMT's STAR Market IPO has become more than a fundraising exercise. The strategic placement roster shows how China's largest DRAM maker is using the capital market to connect semiconductor suppliers, AI cloud providers, device brands, automakers, and state-backed investors, reinforcing a domestic memory ecosystem.
Nvidia has laid out a sweeping expansion of its Japanese footprint. The company is moving beyond one-off supercomputer wins to embed its Blackwell-generation chips and software across the country's research labs, banks, hospitals, factories, and automakers. The breadth signals that Japan is being positioned as a full "AI ecosystem" for Nvidia, not a single-sector customer. It's a hedge that spreads the company's growth across sovereign science, industrial automation, and physical AI, even as questions mount over chip pricing and supply.
Japan's companies and research institutions are turning to Nvidia's Nemotron open models to build AI tailored to local language, industry, and public-sector needs. The move highlights how open, customizable systems may shape national AI strategies far beyond Japan, affecting productivity, service delivery, and data control worldwide.
Beijing Approach AI Technology Co., or Approaching.AI, raised more than CNY1 billion within six months by selling AI tokens generated largely on computing infrastructure it does not own.
South Korean President Lee Jae-myung unveiled the country's Three Mega Projects for AI and Semiconductors in late June 2026, an ambitious national strategy designed to strengthen South Korea's global leadership in artificial intelligence and semiconductors. The initiative centers on three pillars—semiconductors, physical AI, and AI data centers—and aims to double the nation's DRAM output within five years while expanding capabilities in high-bandwidth memory (HBM), advanced packaging, AI processors, and next-generation memory technologies. It also seeks to extend South Korea's semiconductor footprint beyond the Seoul metropolitan region.
