On January 6, it was reported that the continuous absorption of advanced graphics processing capabilities by AI workloads has led to supply tightness in the consumer-level GPU market. European retailers indicated that the supply of Nvidia’s RTX 50 series graphics cards is restricted, with mid-to-high-end models such as the RTX 5070 having extremely low inventory levels, and large orders are often canceled. Retailers can only sell a limited range of models, with high-end RTX 50 series products almost out of stock.
Market pressure intensified at the beginning of 2026, with rising GPU prices in Europe and stricter purchasing restrictions imposed by e-commerce platforms on retailers. Although AMD’s RDNA 4 architecture RX 9000 series has intermittent restocking, supply may tighten again due to increased demand from non-consumer markets. Industry analysts pointed out that this shortage reflects structural changes, with AI and data center deployments prioritizing access to advanced process capacities, resulting in longer lead times, higher costs, and lower availability for consumer-level graphics cards.
On January 7, SemiMedia reported that Japan plans to increase its overall budget by about 50% in the 2026 fiscal year, reaching approximately 3.07 trillion yen, with around 1.23 trillion yen specifically allocated for chip manufacturing and AI-related programs. This budget increase marks Japan’s shift from one-time supplementary funding to a more predictable and regular budget allocation, aimed at providing stable financial support for projects with multi-year investment cycles, such as fabs and ecosystem construction.
On January 5, SemiMedia reported that Pulse Electronics (the magnetic component division of Yageo) announced a price increase for ferrite bead products of size 1608 and above, effective from January 1, 2026, to cope with the continuous rise in raw material costs. This price adjustment does not involve existing contracts and ongoing projects.
Previously, Yageo had adjusted the prices of tantalum capacitors under its Kemet brand. Industry insiders noted that this price hike reflects the rising cost pressure in the passive component supply chain, with the significant impact of rising silver prices on the cost structure of magnetic components that rely on silver paste manufacturing. To protect profits, passive component manufacturers are increasingly selectively raising prices, especially in application areas with high supply continuity requirements, such as automotive, industrial, and data centers.
On January 8, Central News Agency reported that market research firm TrendForce indicated that due to Nvidia’s adjustment of the HBM4 specifications for the Rubin platform and changes in product strategy, the mass production of HBM4 is expected to be delayed, with the earliest production start in the first quarter of 2026.
Nvidia increased the per-pin speed of HBM4 on the Rubin platform to 11Gbps, forcing the three major HBM suppliers to revise their designs. Meanwhile, the AI boom has stimulated stronger-than-expected demand for Nvidia’s previous-generation Blackwell products, leading to an adjustment in the mass production schedule for the Rubin platform.
On January 9, Economic Daily reported that Quanta’s consolidated revenue for December 2025 reached 272.495 billion yuan, a monthly increase of 41.2% and a yearly increase of 94.6%, setting a new historical record and breaking through the 200 billion yuan mark for the first time. The total revenue for 2025 was 2.12 trillion yuan, with a yearly increase of 50.5%. Quanta shipped 3.9 million laptops in December, a monthly increase of 11.42% and a yearly decrease of 9.3%; the total annual shipment was 45.9 million units, with a yearly increase of 1.3%.
Despite the decline in laptop shipments, the AI server business made a significant contribution, with a strong performance in December. Nvidia’s GB300 AI servers began to ramp up shipments, and the demand for GB200 continued to grow steadily. Quanta revealed that its AI server business will double in 2026, with order visibility extending to 2027, and AI servers will account for more than 80% of the total server business.
On January 9, Commercial Times reported that with the mass production of Nvidia’s GB300 NVL72 AI servers, Taiwan-based ODM manufacturers maintained double-digit year-on-year growth in December 2025, with record-high performance in the fourth quarter and the entire year. Wistron’s consolidated revenue for December was 255.255 billion yuan, with a yearly increase of 141.6%, and the total annual revenue exceeded 2 trillion yuan; Wiwynn’s monthly revenue was 104.291 billion yuan, with a total annual revenue of 950.663 billion yuan, both setting new historical records. Quanta’s December revenue was 272.495 billion yuan, with a total annual revenue of 2.12 trillion yuan and a yearly increase of 50.5%. Compal’s December revenue was 10.553 billion yuan, with a total annual revenue of 105.58 billion yuan and a yearly increase of 72%.
