Japanese semiconductor firms are pivoting toward advanced packaging and alternative lithography as TSMC extends its scale advantage in artificial intelligence chips. TSMC's 3nm and 2nm capacity is largely locked in. Advanced packaging now contributes a growing share of sales. The company is on track to see annual revenue exceed US$250 billion around 2029 or 2030. That widening gap has pushed Japanese companies to focus on panel-level packaging and nanoimprint technologies. Both emerged at SEMICON Japan 2025 as key levers to bypass manufacturing bottlenecks and secure a foothold in the fast-expanding AI accelerator supply chain

Artificial intelligence continues to be the central force behind the upcoming productivity revolution. Yet in the US, foundational energy constraints threaten to stall progress. The primary obstacle is not a shortage of semiconductor chips or inadequate computing capacity. It is a more fundamental resource: electricity. Increasing demand from AI data centers (AIDCs) is straining the nation's electric grid. It is testing the limits of social tolerance

The global semiconductor market is projected to reach US$1 trillion as early as 2026, significantly ahead of previous industry forecasts targeting 2030. The World Semiconductor Trade Statistics (WSTS) forecast, released in December 2024, expects growth to be driven predominantly by logic chips, including GPUs and AI accelerators. Memory markets are poised for the steepest increase

The biggest recent news in AI chip development is Nvidia's non-exclusive technology licensing agreement with Groq. Nvidia invested US$20 billion to acquire Groq's technology license and onboard its core engineering team

When the Manhattan Project mobilized the full weight of the American state in 1945 to unlock atomic energy, it revealed something humanity had not fully grasped before: once a scientific breakthrough is absorbed into national strategy, its impact can far exceed any single industry or technology. Eighty years later, the US is attempting to recreate that logic—this time around artificial intelligence (AI)

Facing US restrictions on high-end computing products, China is restructuring its AI chip industry by advancing GPU, TPU, and NPU technologies simultaneously. Domestic firms struggle to match Nvidia's software ecosystem but seek breakthroughs with TPUs for efficiency and NPUs for edge applications

TSMC is navigating a complex landscape of export controls and domestic competition as speculation grows around its second Kumamoto factory. The key question: will the Japanese facility produce advanced 2nm chips to fuel the nation's artificial intelligence ambitions

After briefly approaching a US$5 trillion market capitalisation, Nvidia spent 2025 deploying capital at an unprecedented pace, backing Groq, OpenAI, Nokia, Synopsys, and Intel through technology deals, equity stakes, and strategic partnerships. The objective is straightforward: convert AI-driven cash inflows into a durable, structural influence across the AI ecosystem

In recent years, Foxconn has steadily expanded its ambitions in electric vehicles, moving beyond its traditional role as a platform provider and contract manufacturer toward the consumer-facing end of the market. Through its subsidiary Foxtron—operating under the formal name Hon Hai Advanced Industry—the group has begun directly running an EV business in Taiwan. Earlier this month, Foxtron hosted an online launch event, unveiling three new EV models simultaneously

The long-rumored transfer of Luxgen, the automotive brand under Yulon Motor, to Foxtron Vehicle Technologies—a joint venture between Foxconn and Yulon—was formally confirmed on the evening of December 19. The announcement has drawn intense scrutiny from both financial markets and the manufacturing sector, as its ripple effects across Taiwan's automotive supply chain continue to unfold

China's GPU start-ups are back at the center of a capital-market frenzy. Following Moore Threads, five-year-old MetaX has listed on Shanghai's STAR Market, with its shares surging more than sevenfold on debut. At its peak, the company's market capitalization briefly hit CNY330 billion (US$46.9 billion)—a striking number for a business generating just CNY1.5–2 billion in annual revenue, and one that would raise eyebrows even by global semiconductor standards

As humanoid robots emerge as a key technology riding the wave of AI, the US and China are rapidly forming two competing camps. Industry insiders analyzing their strengths say the US holds an edge mainly in software and foundational models, while China slightly outpaces the US in hardware innovation and iteration speed, with costs reportedly just one-fifth those of American products, making it difficult for other countries to catch up in the short term

Intel is reportedly seeking to acquire SambaNova, a prominent AI inference chip specialist. The deal has a high probability of closing, primarily due to a significant governance alignment: Intel's current CEO, Lip-Bu Tan, also serves as SambaNova's Executive Chairman

Broadcom reported strong AI-related order growth but noted lower gross margins for ASIC products compared to non-AI items, causing market uncertainty about 2026 revenue. Industry experts explained that lower ASIC margins are common due to customer-driven design changes and production timing

As the US–China tech confrontation drags on, discussions consistently return to chips, semiconductors, rare earths, and tariffs. Despite accelerating supply-chain decoupling and overseas clients pushing for "Out-of-China" capacity, China remains the world's largest PCB manufacturing base and has fully captured both the volume and value gains from the AI server cycle. The country continues to demonstrate cost and efficiency advantages over production in Taiwan and Southeast Asia