Nvidia is reportedly working on a new AI chip specifically for China. This comes after recent regulatory hurdles with the H20 chip. The upcoming chip, rumored to be called B30A, aims to offer stronger performance while staying compliant with U.S. export rules.
Let’s dive into what this means for Nvidia, China, and the global AI industry.
Background: The H20 Chip and Export Restrictions
Earlier in 2025, the U.S. government eased some restrictions on AI chip exports. This allowed Nvidia to sell the H20 chip to China. However, the H20 was still tightly regulated. Chinese authorities flagged security concerns, causing some companies to pause purchases.
In response, Nvidia reportedly instructed suppliers, including Foxconn and Amkor, to halt H20 production. This pause paved the way for a new, more advanced chip designed to satisfy both regulatory and market demands.
Introducing the B30A: A Next-Generation AI Solution
Built on Blackwell Architecture
The new chip is said to use Nvidia’s latest Blackwell architecture. This promises higher efficiency and better AI performance compared to H20.
Unlike dual-die chips like the B300, the B30A uses a single-die design, making it simpler to produce and easier to approve for export.
Key features include:
-
High-bandwidth memory (HBM) for rapid data transfer.
-
NVLink for faster interconnectivity between GPUs.
-
Optimized for AI workloads, including deep learning and large-scale data processing.
Reports suggest it could outperform the H100 in certain AI tasks, even though it may deliver roughly half the total compute power of Nvidia’s top-tier B300 accelerator.
Why Nvidia Needs a China-Specific Chip
Market Access
China is a key growth market for Nvidia. In recent years, it has contributed roughly 13% of Nvidia’s revenue. Maintaining a strong presence is critical for the company’s bottom line and global market share.
Regulatory Compliance
U.S. export regulations remain strict, particularly around advanced AI capabilities. The B30A is reportedly engineered to comply with these rules while still delivering competitive performance in China.
Political Sensitivity
The H20 faced pushback from Beijing after U.S. officials publicly downplayed its strategic importance. Companies were discouraged from using it. A new chip, designed to meet both regulatory and political expectations, is now necessary.
Technical Details of the B30A
Though Nvidia has not officially confirmed all specifications, sources indicate:
-
Single-die architecture for streamlined production.
-
Blackwell-based enhancements for higher AI performance.
-
HBM memory support and NVLink integration.
-
Compliance-focused design, avoiding features restricted under U.S. law.
-
Sampling for Chinese clients could begin in late 2025, with wider availability in early 2026.
Compared to the H20, the B30A is likely more energy-efficient and easier for Chinese companies to deploy at scale.
How the B30A Compares to Existing Chips
| Feature | H20 | B30A (Rumored) |
|---|---|---|
| Architecture | Hopper | Blackwell |
| Design | Multi-die | Single-die |
| Memory | HBM3 | HBM3/Blackwell-enhanced |
| Interconnect | NVLink | NVLink |
| Compute Power | Lower than B300 | ~50% of B300, better than H100 |
| Production Status | Halted due to Chinese restrictions | Samples expected late 2025 |
| Regulatory Compliance | Controversial | Designed to comply |
Industry Implications
For Nvidia
The B30A allows Nvidia to retain a foothold in China. It also sets a precedent for creating region-specific AI chips that balance power with export regulations.
For Chinese Companies
The chip could fill the void left by the H20’s pause. However, domestic AI chipmakers like Huawei, Moore Threads, and Biren continue to grow, creating competition.
For Global AI Competition
This development highlights how geopolitical tensions affect semiconductor supply chains. Companies must now navigate both technical performance and international regulations to succeed.
Advantages of the B30A
-
Regulatory Compliant – Built to meet U.S. export restrictions.
-
Higher AI Performance – Improved speed and efficiency over the H20.
-
Faster Deployment – Single-die design simplifies manufacturing and supply.
-
Market-Specific – Tailored to meet China’s AI infrastructure needs.
-
Supports Growth – Enables Chinese companies to continue AI development without violating restrictions.
Challenges and Limitations
-
Not as powerful as top-tier global chips like the B300.
-
Political uncertainty could still delay adoption.
-
Competition from domestic Chinese firms could limit Nvidia’s market share.
-
Limited flexibility compared to Nvidia’s unrestricted chips.
Future Outlook
The B30A may mark the beginning of a trend: customized AI chips for specific regions. Companies like Nvidia and AMD may continue to develop chips that balance performance with regulatory constraints.
China’s domestic AI chip market is growing rapidly. Even with the B30A, Chinese firms are investing heavily in alternatives. Nvidia will need to continuously innovate to maintain relevance.