SoftBank Set to Pass Toyota as Japan's Most Valuable Company

On June 1, 2026, SoftBank Group was on track to overtake Toyota Motor Corporation and become Japan's most valuable company by market capitalization — a milestone that symbolizes the global AI investment boom.

Over the past few years, SoftBank has poured massive investments into AI startups, including OpenAI. The rising valuations of those holdings have now carried it past Toyota, long the symbol of Japan's automotive manufacturing prowess. It is a moment when AI is rewriting the very hierarchy of Japan's industrial structure.

That said, market capitalization reflects market expectations, and it is worth noting that it does not necessarily align with actual earnings or the stability of the underlying business.

US Tightens Restrictions on NVIDIA AI Chip Shipments to Chinese Companies

The US government has taken its restrictions on shipments of NVIDIA AI chips to Chinese companies one step further. While regulations targeting mainland China already existed, the latest move aims to close off circumvention routes through bases outside China as well.

According to Reuters, the measure appears designed to counter efforts by Chinese companies to procure high-performance AI chips via third countries. The US-China standoff over AI semiconductors is moving beyond ordinary trade friction and taking shape as a structural battle for technological supremacy.

MiniMax M3 — A New Model with 1-Million-Token Context Support

China's MiniMax has released its new model, M3. Optimized for coding and agentic use cases, it supports a context length of up to 1 million tokens, along with multimodal input.

Discussion is heating up in the LocalLLaMA community around its long-context stability and coding capability. The 1-million-token figure is a specification ceiling, and real-world performance will have to await future benchmark testing — but this "order-of-magnitude expansion" of context length is emblematic of where the industry is heading.

AI Age Estimation Is Deciding the Fate of Asylum Seekers

According to a report by the UK's Smarter Articles, AI-based age estimation technology is being deployed at border controls, causing serious problems particularly in age determinations for asylum seekers.

With the ruling on whether someone is a child or an adult determining subsequent access to protection and the course of legal procedures, AI estimates still vary widely in accuracy. Critics also point to potential bias based on skin tone and ethnicity. Beyond questions of technical accuracy, there are renewed calls to debate the ethical legitimacy of entrusting AI with decisions that shape human lives.

Benchmarking LFM2.5-8B-A1B on Apple M5 — Where Local LLMs Stand Today

A hands-on review published on Zenn shares real-world measurements of Liquid AI's LFM2.5-8B-A1B running on an Apple M5 (16GB). Despite being a lightweight model with 1B active parameters, it earns praise for striking a balance between Japanese-language quality and inference speed.

Demand for running LLMs locally remains strong, and Apple Silicon's unified memory architecture is particularly well suited to inference workloads. Under a 16GB memory constraint, however, running larger models still has its limits.

Investing in SK Hynix — Shifting Memory Chip Supply and Demand

A top-tier technology fund is planning to invest in SK Hynix, with the view that tight supply will push memory chip prices even higher.

With demand for HBM (high-bandwidth memory) from AI data centers remaining strong, SK Hynix occupies a critical position in NVIDIA's supply chain. The memory chip market is directly tied to the AI semiconductor market, and the investment is seen as a reflection of confidence in the durability of AI demand.

---