The Fourth Ticket
On April 30, 2026, news from Japanese tech media sent shockwaves through the GPU industry: Chinese GPU manufacturer Lisuan (formerly Sophgo) has obtained Microsoft WHQL (Windows Hardware Quality Labs) certification.
This marks the first time a Chinese GPU company has secured this certification, making it the fourth GPU vendor globally to pass the test—following NVIDIA, AMD, and Intel.
The WHQL certification is more than just a piece of paper. It means that Lisuan’s GPU drivers have passed Microsoft’s full compatibility test suite, qualifying them for the “Designed for Windows” logo on Windows systems. System stability and blue screen risk control now meet Microsoft’s standards.
For a Chinese GPU startup, this is a rare signal in a field notorious for its high ecosystem barriers.
Cracks in the Oligopoly
The concentration in the GPU market is suffocating. NVIDIA holds approximately 80-92% of the AI training GPU market share, AMD around 10%, and Intel about 5%. Combined, the three companies account for over 95% of the market.
This landscape has persisted for over a decade. New entrants face not only technological barriers but also formidable ecosystem barriers:
- Driver Compatibility: Without WHQL certification, driver installations trigger security warnings on Windows systems, making enterprise IT departments highly unlikely to purchase.
- Software Ecosystem: The three major ecosystems—CUDA, ROCm, and oneAPI—cover nearly all AI frameworks, requiring new architectures to undergo extensive re-adaptation.
- Developer Mindshare: Engineers are deeply accustomed to these three toolchains, making migration costs prohibitively high.
By securing WHQL certification, Lisuan has effectively chiseled the first crack into this ecosystem barrier.
Who is Lisuan?
Lisuan (Sophgo) spun out of Bitmain and initially started with AI ASIC chips before pivoting to general-purpose GPU R&D. Its core product lineup includes:
| Product Line | Positioning | Key Features |
|---|---|---|
| SG Series | AI Inference Accelerator Cards | Designed for cloud inference, supports mainstream AI frameworks |
| BM Series | Edge Computing Chips | Low power consumption, suitable for IoT and edge scenarios |
| General-Purpose GPU | Desktop/Data Center | Compatible with DirectX/OpenGL/Vulkan, targeting consumer and server markets |
Notably, Lisuan’s GPU strategy does not directly compete with NVIDIA’s CUDA ecosystem. Instead, it has adopted an open standards-first approach—prioritizing support for open APIs like Vulkan, OpenCL, and DirectX to lower the barrier for developer migration.
Why WHQL?
For GPU vendors, WHQL certification is an “entry ticket,” but far from the finish line:
Short-Term Value:
- Windows system drivers install without pop-up warnings
- Drivers can be automatically distributed via Windows Update
- Lowers compliance barriers for enterprise procurement
Medium-Term Value:
- Paves the way for entering the PC OEM market (vendors like Lenovo and Dell require WHQL certification)
- Game and creative software compatibility receives Microsoft’s endorsement
- Developers can commit to porting with greater confidence
Long-Term Challenges:
- Ecosystem building is far more complex than certification—NVIDIA’s CUDA moat cannot be bridged by WHQL alone
- Performance gaps will still require several product generations to close
- The AI training market remains locked by CUDA, while the inference market is relatively open but highly competitive
Signals to the Industry
Lisuan’s WHQL certification conveys three key signals:
1. Chinese GPUs Are Transitioning from “Functional” to “User-Friendly”
Early Chinese GPUs could mostly only run specific AI inference tasks in Linux environments, with near-zero Windows desktop compatibility. WHQL certification means Lisuan’s GPUs have now met Microsoft’s stability standards for general-purpose scenarios.
2. The Open Standards Route May Offer a Differentiated Path
While NVIDIA locks in the AI training market with CUDA and AMD struggles to catch up with ROCm, Lisuan’s choice of the Vulkan + OpenCL open standards route targets developers who “refuse to be locked into any single ecosystem.” This path is harder, but if successful, the resulting barriers will be even more formidable.
3. Demand for Supply Chain Diversification Amid Geopolitics
Against the backdrop of ongoing US-China tech competition, the demand from Chinese tech companies for “non-US GPUs” is very real. Even if performance temporarily lags, supply chain security alone constitutes a purchasing motive. WHQL certification makes this purchasing decision much easier for corporate IT departments to accept.
Recommendations for Readers
If you are considering Chinese GPU solutions:
- Inference scenarios (especially those already adapted to Vulkan/OpenCL) can begin testing
- For training scenarios, it is advisable to wait and see; the CUDA ecosystem remains irreplaceable in the short term
- Watch for whether Lisuan will subsequently release optimized support for mainstream AI frameworks (PyTorch, TensorFlow)
If you are tracking the GPU industry landscape:
- WHQL certification is a prerequisite for Lisuan to enter the PC OEM market; next, watch for announcements of partnerships with complete system manufacturers
- In China’s GPU sector, players like Moore Threads and Biren Technology also compete alongside Lisuan, but currently, only Lisuan holds WHQL certification
The oligopolistic structure of the GPU market will not be overturned by a single certification. However, once a crack appears, subsequent changes often unfold faster than anticipated.