Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesThe May 29, 2025 report that Nvidia and AMD might soon sell China-specific AI chips was a forecast, not a confirmed launch announcement. Nvidia was reportedly preparing a Blackwell-derived product called the B20, while AMD was linked to the Radeon AI PRO R9700. The expected July 2025 sales window did not produce unrestricted access: H20 and MI308 exports later required licenses, and newer products such as H200 and MI325X became subject to case-by-case review.
The practical question is no longer simply whether a chip was engineered below a U.S. performance threshold. Each shipment can also depend on the product, customer, end use, U.S. approval, Chinese import policy, supply availability and the buyer’s willingness to rely on a politically exposed supplier.
What was reported in May 2025?
On May 29, 2025, TechCrunch, citing DigiTimes supply-chain reporting, said Nvidia was preparing a lower-performance, China-focused Blackwell derivative known as the B20. The same report linked AMD’s Radeon AI PRO R9700 workstation GPU to Chinese AI demand. It described a possible launch around July 2025, subject to production and regulatory approval.
Neither claim should be treated as a formal Nvidia or AMD announcement of a China-specific product. The B20 and the R9700 were reported development and market plans, not confirmed commercial launches in the source material. Nvidia’s report also said the company faced a $4.5 billion charge related to H20 licensing restrictions and forecast an $8 billion second-quarter revenue impact; those were 2025 figures, not current estimates. TechCrunch’s May 29, 2025 report is the source for those details.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minute#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
Why make a China-specific chip?
U.S. export controls restrict the transfer to China of advanced computing semiconductors and related technology. Rules can consider computing performance, memory and interconnect characteristics, system-level capability, end users and the transaction structure.
That creates four separate hurdles:
- Engineering: a company designs a product intended to fit the technical limits in force at that time.
- Legal eligibility: the design may fall outside an automatic prohibition, but that does not guarantee shipment.
- Licensing: the U.S. government may still require a license, limit it to named customers or deny it.
- Market acceptance: Chinese regulators and customers must permit and want the product, and the supplier must have inventory.
A chip can therefore be designed for the legal export envelope and still be unavailable to a particular Chinese buyer. The Bureau of Industry and Security says its advanced-computing controls address national-security and foreign-policy risks and can cover U.S.-origin and certain foreign-produced items, software, technology and activities by U.S. persons. BIS background on the controls explains the broader authority.
The products at the center of the dispute
Nvidia: H20, reported Blackwell variants and H200
H20 was developed for the China market around earlier U.S. restrictions. In April 2025, the U.S. government told Nvidia that H20 exports to China required a license. Nvidia later said licenses granted in August 2025 allowed certain H20 products to ship to some China-based customers. Its filing also described a multibillion-dollar financial impact and warned that export controls could materially affect its China business. Nvidia’s SEC filing provides those disclosures.
The B20, and later names such as B20A, B30A and RTX 6000D, have appeared in reporting about possible China-focused Blackwell-related designs. Nvidia has not publicly confirmed all of those names as China-specific products in the evidence available here, so they should be treated as reported programs rather than established product specifications. A lower-performance design might reduce throughput, bandwidth or scalability, but exact reductions, pricing and customer lists have not been established.
Rank #2
- NVIDIA Ampere Architecture-based CUDA Cores - Double-speed processing for single-precision floating point (FP32) operations and improved power efficiency provide significant performance improvements for graphics and simulation workflows, such as complex 3D computer-aided design (CAD) and computer-aided engineering (CAE), on the desktop.
- Second-Generation RT Cores - With up to 2X the throughput over the previous generation and the ability to concurrently run ray tracing with either shading or denoising capabilities, second-generation RT Cores deliver massive speedups for workloads like photorealistic rendering of movie content, architectural design evaluations, and virtual prototyping of product designs. This technology also speeds up the rendering of ray-traced motion blur for faster results with greater visual accuracy.
- Third-Generation Tensor Cores - New Tensor Float 32 (TF32) precision provides up to 5X the training throughput over the previous generation to accelerate AI and data science model training without requiring any code changes. Hardware support for structural sparsity doubles the throughput for inferencing. Tensor Cores also bring AI to graphics with capabilities like DLSS, AI denoising, and enhanced editing for select applications.
- Third-Generation NVIDIA NVLink - Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate graphics and compute workloads and tackle larger datasets.
- 48 Gigabytes (GB) of GPU Memory - Ultra-fast GDDR6 memory, scalable up to 96 GB with NVLink, gives data scientists, engineers, and creative professionals the large memory necessary to work with massive datasets and workloads like data science and simulation.
H200 is more powerful than the earlier China-oriented products. It became potentially exportable to selected Chinese customers only after a later U.S. policy change and individual licensing, not through unrestricted market access.
AMD: MI308, MI325 and Radeon AI PRO R9700
MI308 is a China-targeted or China-compliant derivative of AMD’s Instinct accelerator family. AMD reported that it began shipping some MI308 products to China after receiving licenses. The company also said MI308 sales depend on customer demand, Chinese import controls and its ability to obtain U.S. licenses.
MI325 and MI325X became relevant under the later licensing policy. AMD said it received licenses in February 2026 authorizing shipments of some MI325 products to certain China-based customers. The Radeon AI PRO R9700 was reported as a workstation-oriented product that could address Chinese AI workloads, but the source did not establish that AMD marketed it specifically as a China export-control response. AMD’s filing covers its licensing and shipment disclosures.
How the “soon” prediction changed
| Date | Development | What it means |
|---|---|---|
| May 29, 2025 | Reports described Nvidia’s B20 and AMD’s Radeon AI PRO R9700, with possible July sales. | A supply-chain forecast, not a confirmed launch. |
| April 2025 | The U.S. required licenses for Nvidia H20 exports; AMD’s MI308 was also affected by new licensing restrictions. | Earlier technical compliance did not ensure continued automatic exports. |
| July 2025 | The administration indicated that some H20 and MI308 shipments could resume as licenses were approved. | Access became shipment- and customer-specific. |
| August 2025 | AP reported that Nvidia and AMD agreed to provide the U.S. government 15% of revenue from certain China chip sales as part of licensing arrangements. | This unusual reported arrangement should not be described casually as an ordinary sales tax. |
| January 13, 2026 | BIS announced case-by-case review for H200, MI325X and similar chips, subject to security conditions. | Some higher-end exports could be considered, but none were guaranteed. |
| July 14, 2026 | Reporting said a limited number of H200 chips had reached China and additional Chinese customers had received approval to buy advanced Nvidia and AMD chips. | Licenses had translated into some shipments, not broad unrestricted availability. |
Sources: TechCrunch, Reuters report on the April licensing requirements, AP report on July 2025 approvals, AP report on the 15% arrangement, BIS policy announcement and Reuters reporting carried by Investing.com.
Rank #3
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
Nvidia and AMD face different commercial calculations
| Company | Relevant products | Main commercial issue |
|---|---|---|
| Nvidia | H20; reported B20 and other Blackwell-related variants; H200 under licensing | A large installed software and customer base, but substantial exposure to changing rules, licensing delays and Chinese substitution. |
| AMD | MI308; MI325/MI325X under licensing; reported R9700 relevance | A smaller accelerator footprint creates room to gain share, but sales remain dependent on U.S. licenses, Chinese import rules and demand. |
For Nvidia, repeated redesigns can create engineering, inventory and forecasting costs. A restricted product may preserve relationships but cannot fully replace unrestricted sales of top-tier accelerators. Customers may also migrate workloads to Huawei or other domestic suppliers if supply is unreliable.
AMD has an opportunity to expand Instinct adoption, yet the same policy uncertainty applies. A license does not guarantee that Chinese customers will order, that imports will clear, or that a product will fit state-funded projects.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why Chinese approval matters as much as U.S. approval
Washington’s permission is only one condition. Chinese authorities may impose import or procurement restrictions, particularly for state-funded data centers. Customers may prefer domestic accelerators to reduce political and supply risk even when a U.S. chip is legally available.
- Foreign chips can offer familiar software compatibility or strong performance for existing workloads.
- Domestic sourcing can reduce exposure to future U.S. licensing changes.
- Porting models and software from Nvidia’s CUDA ecosystem or AMD’s ROCm stack can impose engineering costs.
- Uncertain support, replacement supply and resale rights can outweigh benchmark performance.
That is why “approved,” “available” and “competitive” describe different commercial outcomes.
Rank #4
- Discrete graphics card memory 40 GB
- Memory bandwidth (max) 1555 GB/s
- Graphics processor family NVIDIA
- Graphics processor A100
What the policy is trying to prevent
The U.S. government frames the controls as national-security measures intended to limit China’s access to advanced computing useful for artificial intelligence, supercomputing and military applications. The policy uses several tools, including performance thresholds, licensing requirements, entity listings, end-user restrictions, foreign-direct-product rules and limits on some U.S.-person support.
The policy has changed repeatedly. A product that fit an earlier threshold can later require a license after a new rule or agency interpretation. That moving target is the central reason a “compliance chip” is not a permanent solution.
What remains unconfirmed
- Whether the reported B20 entered volume production.
- Whether the R9700 was sold in China specifically as an export-control response.
- The exact specifications, price and customer list for any China-specific Nvidia Blackwell product.
- Whether reported licenses produced substantial commercial shipments beyond the disclosed examples.
- The final legal structure and administration of the reported 15% revenue-sharing arrangement.
- Whether Chinese state-funded data centers will use these products at scale.
- Whether Nvidia and AMD will keep developing China-specific products if U.S. rules change again.
What investors and enterprise buyers should watch
- New BIS licensing rules, advisories and end-user restrictions.
- Nvidia and AMD SEC filings that identify products, customers, licenses and financial charges.
- Actual shipment volumes, rather than headlines announcing approvals.
- Chinese procurement rules for government-backed data centers.
- Huawei and other domestic accelerator progress, including software support.
- Whether products remain available long enough to justify model-porting and infrastructure investments.
Enterprise accelerators are generally quote-based through vendors, server manufacturers, distributors or cloud providers. Customers with China-linked ownership should not assume that buying through a third country avoids U.S. controls. Export eligibility must be checked for the exact product, destination, end user and transaction.
The Bottom Line
The original story was about engineering weaker chips to fit the rules then in force. By 2026, the decisive issue was broader: licenses, named customers, Chinese procurement policy, supply continuity and geopolitical risk. Nvidia and AMD may still sell selected AI accelerators into China, but compliance design alone no longer guarantees access.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
