Global GPU Computing News【20260110】

1. NVIDIA officially unveiled its next-generation AI platform, Vera Rubin, at CES 2026. The core Rubin GPU delivers 5x the inference performance of the previous Blackwell platform, and the company announced the platform has entered full-scale production.

2. Simultaneously, NVIDIA introduced the DGX Spark, a personal supercomputer designed for local AI computing. Based on the Grace Blackwell architecture, it aims to enable developers to run or fine-tune hundred-billion-parameter large language models locally, lowering the barrier to using high-end AI computing power.

3. The U.S. policy on AI chip exports to China has shown fluctuations. Reports indicate that after blocking the export of the specially tailored H20 chip, the Trump administration approved the sale of the more powerful H200 chip to China, a decision that has sparked discussion within the industry.

4. Domestic GPU company Tianshu Zhixin announced it will disclose the detailed roadmap for its next three generations of products on the 26th of this month. The products target NVIDIA's H200/B200 series, with plans to compete directly between 2026 and 2028.

5. UBS released an industry report expressing optimism about the 2026 AI GPU market. It believes AI GPUs and custom ASICs will jointly drive market growth and predicts NVIDIA will deliver approximately 60,000 server racks based on the Blackwell platform within the year.

6. Market analysis suggests the AI chip market is expected to maintain strong growth in 2026. In a related move, NVIDIA signed a $20 billion licensing agreement with chip design company Grok to jointly enhance AI inference speed.

7. AMD launched its new Ryzen AI Embedded processor series, the P100/X100, at CES. Integrating RDNA 3.5 architecture GPUs and NPUs with up to 50 TOPS of computing power, they are primarily aimed at edge AI applications such as automotive and industrial sectors.

8. AMD also introduced a cabinet-scale AI computing solution named Helios, featuring a fully liquid-cooled design. This signals a shift in industry competition from single-chip performance to the delivery of complete system-level computing power.

9. Industry reports summarized the cutting-edge developments in AI computing at this year's CES, noting that NVIDIA, AMD, and Intel all launched next-generation computing platforms. High-performance GPUs, CPUs, and high-speed interconnect chips remain the core focus for future development.

10. Loongson Technology disclosed progress on its GPU products. Its first GPU, the 9A1000, already possesses basic display functionality and is undergoing driver adaptation, laying the groundwork for the subsequent development of more advanced products like the 9A2000. The company plans to strive for internationally advanced levels by 2026.