Discover Huawei’s Ascend roadmap: from AI chips 950 to 970 and the innovations shaping the future of AI computing. What’s next in Huawei’s AI chip strategy?
Introduction: Why Huawei’s AI Chips Matter
When it comes to the global race in artificial intelligence, Huawei has emerged as a key player—not just in networking and 5G, but increasingly in semiconductor design. Its Ascend AI chip series, part of the company’s AI-focused roadmap, represents an ambitious bid to compete with U.S. giants like Nvidia, AMD, and Intel while also navigating geopolitical restrictions.
From the Ascend 950 to the Ascend 970, Huawei has been incrementally improving chip performance, architecture, and efficiency to power AI applications ranging from cloud computing to edge devices. But beyond just hardware, Huawei is trying to establish a long-term ecosystem around Ascend, creating a comprehensive AI development framework.
So what does the roadmap look like? And more importantly, what comes after the Ascend 970?
This deep dive explores Huawei’s AI chip strategy, its evolution, the challenges it faces, and where it could be headed next.
Section 1: The Global AI Chip Market Landscape
Before diving into Huawei’s roadmap, it’s important to understand the broader market context.
- Nvidia dominates with its GPU-powered AI accelerators, especially in training large models like GPT.
- AMD is pushing with its MI300 series, focusing on energy efficiency and scaling.
- Intel continues to innovate with Gaudi accelerators and its Xeon CPU integration.
- Startups (like Graphcore, Cerebras, and Tenstorrent) are carving out niche spaces with custom AI chips.
Huawei entered this high-stakes environment with the Ascend series, aiming not only at China’s domestic market but also at global enterprises before U.S. sanctions hit.
Section 2: Huawei’s Entry Into AI Chips – The Ascend 950
The Ascend 950 was Huawei’s first major attempt at an AI processor designed for cloud and data center applications.
Key Features of Ascend 950
- Architecture: Huawei’s custom Da Vinci architecture optimized for AI workloads.
- Use Case: Cloud AI services, training, and inference.
- Performance: Competitive within China, though behind Nvidia in global benchmarks.
The 950 marked Huawei’s entry point into the AI hardware race. It was less about beating Nvidia immediately and more about building an in-house foundation to reduce reliance on foreign technology.
Section 3: Ascend 970 – Powering Smartphones with On-Device AI
The Ascend 970 represented Huawei’s pivot to edge AI, embedding powerful neural processing into smartphones.
Key Highlights of Ascend 970
- Integration: Built into Huawei’s Kirin 970 mobile SoC.
- NPU (Neural Processing Unit): First of its kind in a smartphone, enabling real-time image recognition and language translation.
- Efficiency: Balanced power consumption with high AI throughput.
The 970 made waves globally because it brought AI to consumers’ pockets, long before Apple’s Neural Engine became mainstream.
It was here that Huawei started differentiating:
- 950 → Cloud AI
- 970 → Edge AI (smartphones, consumer devices)
Section 4: The Da Vinci Architecture – Huawei’s Secret Weapon
Both 950 and 970 chips rely on Da Vinci architecture, Huawei’s custom AI framework.
Why Da Vinci Matters
- Scalability: From edge devices to cloud servers.
- Flexibility: Supports training and inference.
- Optimization: Fine-tuned for Huawei’s own ecosystem.
This modularity allows Huawei to extend the same architectural DNA across product lines, making development and scaling more efficient.
Section 5: Ascend 950 vs. Ascend 970 – A Comparative Lens
| Feature | Ascend 950 (Cloud) | Ascend 970 (Edge/Mobile) |
|---|---|---|
| Target Use | Data centers, training, inference | Smartphones, edge devices |
| Performance | High compute for cloud AI | Optimized for power efficiency |
| NPU Integration | Yes, large-scale | Yes, mobile-specific |
| Ecosystem Role | Enterprise, cloud AI services | Consumer AI adoption |
This dual-track approach demonstrates Huawei’s strategy: balance enterprise-scale AI with consumer-focused AI adoption.
Section 6: Huawei’s Challenges in AI Chip Development
Despite its innovations, Huawei faces enormous challenges:
- U.S. Sanctions
- Restricted access to advanced chipmaking tools and foundries (e.g., TSMC, ASML).
- Limited partnerships with U.S. software firms.
- Supply Chain Bottlenecks
- Difficulty sourcing advanced lithography equipment.
- Reliance on domestic Chinese fabs still catching up.
- Global Market Access
- Barred from U.S. markets.
- Pressured alliances in Europe.
Yet, these constraints have pushed Huawei toward self-reliance, accelerating China’s ambition for semiconductor independence.
Section 7: Beyond Ascend 970 – What’s Next for Huawei?
The key question: What lies ahead in Huawei’s AI chip roadmap?
Predicted Next Steps
- Ascend 980 and Beyond
- More advanced NPUs, possibly rivaling Nvidia’s H-series accelerators.
- Improved energy efficiency to compete in cloud and mobile simultaneously.
- Deeper Integration with HarmonyOS
- Future Ascend chips will likely integrate seamlessly into Huawei’s HarmonyOS ecosystem, powering everything from smartphones to IoT devices.
- AI Cloud + Edge Synergy
- A focus on end-to-end AI solutions, where Ascend chips in data centers sync with Ascend chips in consumer devices.
- AI for Enterprise & Government
- Huawei is investing in AI for smart cities, healthcare, and national security applications within China and friendly markets.
Section 8: The Future of Huawei’s AI Ecosystem
Huawei isn’t just building chips—it’s building an AI-first ecosystem.
- MindSpore Framework: Huawei’s open-source AI framework optimized for Ascend chips.
- Cloud AI Services: Integration into Huawei Cloud for enterprise-scale applications.
- AI Research Investment: Billions poured into AI labs across China and abroad.
This strategy reflects a broader trend: AI leadership will not just be about hardware, but ecosystems.
Section 9: Global Implications – Competing with the U.S.
Huawei’s AI chip development has geopolitical consequences:
- For the U.S.: Huawei represents a Chinese alternative to Nvidia and AMD, threatening American dominance.
- For Enterprises: Companies outside the U.S. may seek Huawei chips to diversify supply.
- For Policymakers: Raises questions about supply chain security, technology sovereignty, and national competitiveness.
Section 10: Predictions for Huawei’s Next Leap
Where Huawei Might Be Headed
- Custom AI accelerators for LLMs (large language models).
- Specialized chips for robotics and autonomous driving.
- AI-powered telecommunications infrastructure.
- Partnerships with BRICS nations to expand adoption.
Huawei’s roadmap suggests that its next-generation Ascend chips will not just catch up to global players—but potentially set new standards in integrated AI systems.
Conclusion: Huawei’s Ascend Roadmap and the Future of AI Chips
Huawei’s journey from the Ascend 950 to 970 is more than a technological evolution—it’s a statement of intent. Despite sanctions and supply chain challenges, Huawei continues to push boundaries, developing chips that serve both enterprise AI needs and consumer devices.
The big takeaway?
Huawei isn’t simply trying to compete—it’s trying to reshape the AI chip industry around a self-reliant ecosystem. With the coming generations (980, 990, and beyond), the company is betting on a future where its AI chips are not just Chinese alternatives, but global benchmarks.
For the U.S. and its allies, this means Huawei’s roadmap is not just a tech story—it’s a geopolitical one. For enterprises and researchers, it signals more options, more competition, and potentially faster AI innovation worldwide.
The question is no longer whether Huawei can innovate—it’s how far and how fast it will go.