NVIDIA’s push to regain momentum in China with the RTX 6000D workstation GPU isn’t going as planned. Despite being tailored for local compliance, major Chinese AI firms are reportedly steering clear of the card, arguing the performance doesn’t justify the investment—especially when more powerful compute remains accessible through existing inventory and informal channels.
The RTX 6000D was pitched as a breakthrough for China-focused AI workloads. It’s the first Blackwell-based solution offered in the region, pairing GDDR7 memory with TSMC’s 4nm process and delivering roughly 1,100 GB/s of memory bandwidth. Early expectations were ambitious, with projections suggesting sales could reach into the millions to help offset shortfalls from constrained AI accelerators. On the ground, though, adoption is muted. Many local tech giants are prioritizing AI accelerators like the H20 once their orders are cleared, underscoring that the RTX 6000D hasn’t hit its intended mark.
Why the hesitation? In a market defined by raw compute per dollar and per watt, the RTX 6000D appears to land in an awkward middle ground. It’s compliant, but not compelling enough for large-scale AI training or inference when compared to alternative paths Chinese firms are already using. With pressure to maximize throughput, buyers are opting to wait for accelerators that better meet their performance targets or to continue tapping existing compute resources.
Looking ahead, NVIDIA is reportedly lining up a more competitive offering for China: the B30A. Also built on the Blackwell architecture and TSMC’s 4nm node, this part is said to feature 8-Hi HBM3E and a dual-chiplet design, a combination that should deliver a substantial uplift compared to the H20. If approved, it could give NVIDIA a much stronger hand in the region’s AI data center race. The catch is that everything hinges on regulatory clearance. Any green light from U.S. authorities will determine how capable NVIDIA’s China-bound lineup can be and how soon it can reach customers.
NVIDIA’s China story has been complicated for several quarters. Geopolitical uncertainty and export rules have forced the company to dial back features and performance to meet compliance thresholds, creating a moving target for both product planning and customer demand. Securing a foothold in Beijing’s rapidly expanding AI ecosystem remains strategically important, but success now depends on two factors: timely regulatory approvals and the ability to deliver accelerators that offer clear value under current restrictions.
In short, the RTX 6000D hasn’t shifted the market the way NVIDIA hoped. Interest appears to be coalescing around more capable options pending approval, like the B30A, while buyers continue to rely on existing compute where possible. The next few months will reveal whether NVIDIA can align regulatory realities with a product that truly resonates with China’s AI leaders.
