Jensen Huang's recent China visit yielded significant announcements: the resumption of H20 AI chip sales and the imminent release of a new China-specific RTX PRO workstation GPU. Dubbed RTX 6000D, this specialized card targets compute graphics, digital twin applications, and AI workloads. This development coincides with AMD lifting its MI308 AI chip export restrictions to China.
While attention often focuses on China-specific AI accelerators like A800/H800/H20 or gaming cards like RTX 4090D, workstation GPUs face equally stringent controls. NVIDIA's previous flagship RTX 5000 Ada was barred from Chinese markets, leading to the creation of the tailored RTX 5880. The newly launched RTX PRO 6000—released globally in March—boasts formidable specifications: 24,064 CUDA cores, 2.6GHz clock speed, 125 TFLOPS single-precision performance, plus 96GB of 512-bit GDDR7 memory delivering 1,792GB/s bandwidth at 600W TDP.
The RTX 6000D will inevitably feature curtailed capabilities to comply with U.S. export controls. While exact computing power reductions remain undisclosed, its GDDR7 memory bandwidth drops sharply to 1,100GB/s—likely achieved through reduced 384-bit bus width and lowered 23GHz effective frequency. This represents a more severe downgrade than the gaming-focused RTX 5090D v2, which retained 28GHz memory frequency despite similar bus-width reductions.
Production plans indicate substantial scale: NVIDIA anticipates shipping 1-2 million RTX 6000D units starting Q3 2025, potentially generating up to $10 billion in revenue. This push follows a turbulent period where China (including Hong Kong and Macau) saw its contribution to NVIDIA's revenue plummet from over $17 billion (13% of global total) in FY2025 to below 5%, with export bans causing $2.5 billion in lost sales and $4.5 billion in asset impairment charges.
Comments